If you read the Thesis Brief, you know the business. You know the flywheel, you know the Estonian tax shock, you know that a 78% EBITDA margin on a classifieds portfolio is genuinely extraordinary. You know the thesis.

This piece goes underneath it.

I want to be upfront about something before we start. I bought a position in BCG after writing the Thesis Brief. But by the time I finished this Deep Dive, I had sold it.

Not because the business disappointed me. The business is everything the Brief said it was, and more. But going through every corner of a company sometimes teaches you something the headline numbers cannot.

It’s about what you are actually paying, and about what you are taking on that does not show up in any financial model.

That lesson is in here, in this deep dive. It comes near the end. I think it is worth the walk to get there.

In this deep dive we are going to:

• look at each of BCG’s four segments individually, not just the revenue number but what is actually driving it and whether it sustains.

• look at the two people running this business and whether they have earned the trust their shareholders are placing in them.

• examine the one acquisition BCG has made and what it says about their capital discipline.

• build the valuation from scratch, assumption by assumption, rather than just quoting a range.

(Put this on listen. It is built for the walk, the commute, and the gym.)

Let’s get into it.

Subscribe now

The four segments, one at a time

BCG reports four business lines:

1. Auto

2. Real Estate

3. Jobs and Services

4. Generalist

Together they produced €44.8m in revenue in the first half of BCG’s 2026 financial year — the six months to the end of October 2025.

But the aggregate number hides something interesting.

Three of the four segments are telling a story of genuine momentum. The fourth is telling a very different story, and understanding the difference is the key to understanding whether this thesis holds.

Real Estate

Let’s start with Real Estate, because Real Estate is where BCG is doing something that deserves more attention than it has received.

In the six months to October 2025, Real Estate revenue was €13.2m. That is up 20% from the year before. Twenty percent revenue growth from a business that already dominates its market. The interesting part is where that growth came from. It was not just price increases, though there were price increases.

The number of real estate brokers paying BCG a monthly subscription grew 4%. The monthly fee each broker pays (the ARPU) grew 16%. And on the consumer side, private sellers paying per listing saw their per-listing yield increase 27%.

Three levers pulling in the same direction at the same time. That does not happen by accident.

Share

What is driving it is a combination of two things. The first is the Baltic real estate market itself, which is genuinely hot right now. Real estate transactions across Lithuania, Estonia, and Latvia grew 7% in the period. Average apartment prices in the three capital cities are up 4%.

When the underlying market is active, brokers are busier, developers are more willing to pay for visibility, and private sellers are more motivated to list at a premium tier because faster sales justify the cost. BCG benefits from a healthy market.

But the second driver is more interesting and more durable. BCG has been deepening the Real Estate product in ways that are changing the relationship between the platform and its professional users. Property Price Compass, the tool that integrates national land registry transaction data with listing history to produce a professional pricing report, was introduced at Aruodas.lt in Lithuania over the last two years.

It is now something broker clients expect. At Untu.lt, the recently acquired platform, agents are handling leads through an AI-assisted call tracking service that logs conversations, suggests follow-up actions, and gives brokers visibility into what happens after a lead is acquired.

These are not cosmetic features. They are workflow tools. And workflow tools create a completely different switching cost profile than listing services. A broker who uses Property Price Compass at every vendor meeting cannot easily switch to a competitor platform without losing the tool her clients have come to expect. BCG is not just the place where listings happen. It is becoming the software layer through which Baltic real estate professionals run their business.

The honest question to ask about Real Estate is whether this growth is repeatable.

The market tailwind (active transactions, rising prices) could cool. Interest rates in the Baltics have been falling, which has supported the market, but that tailwind is not permanent. And the ARPU expansion of 16% reflects pricing changes implemented in September and October 2024.

Those changes have now fully annualised. The next set of pricing changes was implemented in September and October 2025, and they have not yet fully fed through. So the near-term revenue picture for Real Estate looks strong. The medium-term picture depends on whether the Baltic real estate market stays active and whether BCG can continue deepening the product quickly enough to justify continued pricing power.

My read is that Real Estate is the segment carrying the thesis right now. It is growing at a rate that suggests the moat is widening rather than just holding.

Auto

Now Auto, which is the segment everyone has been focused on for the wrong reasons.

Auto produced €16.0m in revenue in the half. Exactly flat with the year before.

Not collapsing, not growing. Just flat.

The Brief explained why; the Estonian vehicle tax halved car transactions, listing volumes fell 29%, but dealer subscriptions held and per-listing prices rose enough to compensate.

What the Brief did not go into is what Auto actually looks like at the unit level, and the unit level picture is more interesting than the flat headline suggests.

On the B2C side (the professional dealer subscriptions) the number of dealers paying BCG fell by just 1% despite the Estonian market collapsing. One percent. In a country where car transactions were down between 45% and 66% every single month. Dealers stayed on the platform because where else were they going to advertise? And the monthly fee each dealer pays grew 13%. In Auto, in a terrible market, BCG still managed to raise prices and keep almost every dealer on the subscription.

On the C2C side (private sellers) the volume picture is ugly. Listed ads down 29%, active ads down 26%. That is the Estonian tax at work. But the per-listing yield grew 29%. The sellers who did list paid more. Again, because there is nowhere else to go.

Subscribe now

Now here is the thing about Auto that makes me genuinely optimistic about H2. The terrible Estonian comparison base rolls off in January 2026. The tax was introduced in December 2024, so from January 2026 onwards you are comparing current Auto performance against already-collapsed prior-year numbers. The year-on-year percentage will flip from negative to positive almost mechanically. Management has confirmed this explicitly.

Lithuania, which is the larger Auto market for BCG, grew car transactions 8% in the same period. BCG’s Autoplusas platform in Lithuania leads the nearest competitor by 6 times on time spent on site. That market is healthy and BCG’s position in it is strong. Auto is not broken. One country implemented a severe tax that hurt one year’s results. The business underneath is intact.

Jobs and Services

Jobs and Services grew 7%, driven by a combination of more customers and better yields. The Jobs side, primarily CVbankas in Lithuania, grew both customer numbers and the monthly fee per customer. The unemployment rate in Lithuania ticked up slightly to 7.1%, which sounds like a headwind but actually reflects a growing labour force rather than a shrinking economy. Lithuanian wages grew 8.5% in the year. Companies are still hiring and still willing to pay for CVbankas’s recruitment tools, including the salary estimator that gives them a data advantage in setting competitive offers.

Services — the C2C classifieds for tradespeople and service providers — grew active ads 11%. More plumbers, cleaners, and builders are listing on BCG’s services portals. This is a steadily growing vertical that does not get much attention because it is modest in size but the trend is consistent.

Generalist

Generalist grew 4%, which looks unimpressive until you understand what it is.

Skelbiu.lt, BCG’s main generalist portal in Lithuania, is the sixth most visited website in the entire country. It competes partly with BCG’s own vertical platforms through cross-listing.

A private seller who lists a car on Skelbiu.lt may also get redirected to Autoplusas. The generalist portal is as much a traffic acquisition tool for the verticals as it is a standalone business. Growth of 4% driven by yield improvement, with active inventory holding steady, is the right outcome for this segment.

The synthesis is straightforward. Real Estate is carrying the thesis, growing at 20% with unit economics that suggest the moat is deepening. Auto is temporarily impaired by policy, structurally intact, and set to show positive year-on-year growth from January 2026 onwards. Jobs and Services is growing steadily with no signs of stress. Generalist is modest, useful, and not the point.

Share

The people running this business

BCG listed on the London Stock Exchange in July 2021. Justinas Šimkus has been CEO since before the IPO. Lina Mačienė is CFO.

They have been running this business together through the IPO, through the post-COVID normalisation, through the Estonian tax shock, and through a sustained period of annual price increases that have pushed ARPU and yields consistently higher across every segment.

The way to evaluate management is not to read their investor presentations. It is to compare what they said with what happened.

At the IPO, BCG’s stated capital allocation policy was to return one third of adjusted net income annually through dividends, with the preference for returning excess cash through share buybacks. Three and a half years later, that is exactly what they have done.

The dividend policy has been adhered to with mechanical consistency. Buybacks have continued through good periods and bad ones.

In the half-year to October 2025, with the Estonian situation creating uncertainty, they still bought back €6.4m of shares and paid a €12.5m dividend. They also voluntarily repaid €10m of debt, which was not required but was the right use of surplus cash.

The Estonian Auto situation is itself a management test. When the vehicle tax was introduced and it became clear that Estonian car transactions were going to collapse, there was a decision to be made about whether to adjust B2C pricing for Estonian car dealers.

BCG’s annual B2C pricing actions were implemented in September and October 2024 for most of its markets. For Estonian Auto dealers specifically, the pricing adjustments were postponed.

They made the right call.

Hitting Estonian dealers with a price increase at the same moment the tax was destroying their market would have been extractive and short-sighted. Instead they held pricing for that one market, kept almost all their dealer base, and positioned themselves for recovery when the comparisons normalise.

That is the Bloopers Test in action. A management team making obvious, decent decisions rather than clever ones.

Share

The one area where I would want more information is on organic growth strategy beyond price increases.

BCG’s revenue growth has been driven primarily by yield expansion (charging existing users more) rather than by growing the user base materially. The number of real estate brokers grew 4% in the half. Auto dealers actually fell 1%. Jobs companies grew 1%. These are not large customer base expansions. The strategy is to extract more value from an already-dominant position rather than to expand the addressable market.

That is a legitimate and often excellent strategy. It is the strategy of a business that knows it has already won its market. But it does mean that the long-term growth trajectory is bounded by how far you can push yields before the customer pushes back. Management has not yet found that wall. Three consecutive years of pricing increases and no material customer loss suggests the wall is further away than bears assume. But it exists, and an honest assessment of management has to acknowledge it.

Overall: yes, I trust these people.

They have done what they said they would do, they have made sensible operational decisions under pressure, and they have not done anything that suggests they are optimising for personal benefit at the expense of shareholders.

The one acquisition

BCG has made one acquisition since listing: Untu.lt, a Lithuanian real estate technology platform, acquired in FY2025 for €1.0m.

One million euros. For context, BCG generates approximately €60m in operating cash annually. This was not a transformative deal. It was a targeted technology acquisition.

What Untu.lt brought was a set of AI-assisted real estate tools;

• The property valuation technology that became Property Price Compass

• The call tracking and lead management system now used by agents on the platform.

BCG acquired a technology capability that deepened their Real Estate moat rather than a revenue stream or a new market.

The price discipline is the notable thing.

€1.0m for a technology acquisition that is now embedded in the workflow of thousands of real estate brokers across the Baltics. That is not a trophy acquisition. That is a specific capability bought at a sensible price and integrated into the core product.

What it signals about BCG’s M&A appetite: they are not empire builders. They are not using their cash generation to buy adjacencies or to pay premium multiples for growth. The one acquisition they made was small, targeted, and directly additive to the moat they already have. That is the right behaviour for a dominant franchise in a small market.

Where the numbers came from; the history since listing

BCG listed in July 2021. Here is what has happened to the business since then:

When BCG listed, revenue for the financial year ending April 2021 was approximately €42m. This financial year, ending April 2026, the annualized run rate puts revenue somewhere around €87m. That is a doubling in five years, which implies a compound annual growth rate of roughly 16%.

The margin history tells an equally interesting story. In financial year 2021, the EBITDA margin was 76%. In 2022 it was closer to 47% , which was a significant step down driven by post-IPO operating cost increases and heavy amortization of acquired intangibles.

But that was a one-off. From 2023 onwards the margin has moved consistently upward: 75.6% in 2023, 76.9% in 2024, 78% in 2025. The business did not find efficiency after a period of inefficiency. It went through a brief post-IPO cost absorption period and then returned to its structural margin level, which has been improving gently ever since.

The capital return history is consistent with the stated policy. Dividends have been paid every year. Buybacks have been ongoing. The share count has been declining. A company that listed with approximately 489 million shares now has approximately 483 million in issue, and that number will continue to fall as buybacks continue.

The key question to ask about the historical picture is whether yield expansion is accelerating or slowing.

The data from H1 2026 suggests it is actually accelerating in some segments. Real Estate C2C per-listed-ad yield grew 27% in the half. Auto C2C per-listed-ad yield grew 29%. These are larger percentage increases than in prior periods.

That tells you one of two things:

1. either BCG has more pricing power than the market appreciates

2. or the base effects from low prior-year prices are temporarily flattering the percentages.

The truth is probably some of both. But the direction is right and the magnitude is not trivial.

Given everything that has happened since the IPO, is BCG better or worse than the prospectus implied? Better. Revenue has grown faster than the prospectus suggested. Margins have recovered to above the listing-period levels.

Capital returns have been disciplined and consistent. The Estonian shock is the first genuine operational headwind since listing, and the business absorbed it without a revenue decline. That is a better track record than most businesses produce in the same period.

Subscribe now

Valuation

Start with what BCG actually earns for its owners, because reported profit and owner earnings are not the same thing for this business.

Net income for H1 2026 was €26.4m. That is the GAAP number. Three adjustments are needed to get to owner earnings.

First, add back the non-cash amortization of acquired intangibles — €3.7m in the half — because this is an accounting artefact of the 2020 acquisition of business client relationships, not a real economic cost. Those intangibles are now fully amortized, so this adjustment disappears going forward.

Second, subtract stock-based compensation of €0.5m. This is real dilution and must be treated as a cost regardless of how it is classified in the income statement.

Third, the maintenance capex is €0.4m against depreciation of €0.4m on non-acquired assets. These roughly cancel out.

The result: owner earnings of approximately €25.9m for the half, or €52m annualised.

Now the assumptions;

Revenue growth: Management has guided for H2 revenue growth above H1’s 7%, and double-digit growth for the full 2027 financial year. The B2C price increases implemented in September and October 2025 will fully annualise into H2 2026 revenue.

Estonian Auto comparables flip positive from January 2026. Real Estate is growing at 20% with a strong market backdrop. A 10% to 12% annualised revenue growth rate for the next three years is conservative relative to management guidance and recent trends.

Margin assumption: BCG has guided that EBITDA margins will remain in the mid-seventies even with investment in data and AI. The structural reason for this is simple: the cost base does not need to grow proportionally with revenue. More listings, higher yields, more broker customers — none of these require meaningful additional headcount or infrastructure. A 75% to 77% EBITDA margin assumption is conservative.

At 10% revenue growth and 76% margins, owner earnings grow from approximately €52m today to approximately €70m by year three. At 12% growth and 77% margins, the number is closer to €78m.

What multiple is warranted?

BCG operates in a market so small that it has essentially pre-empted competition. The moat is deepening through product development. The capital requirements are negligible.

The management team has a clean track record. For a business with these characteristics and this growth profile, a multiple of 22 to 25 times owner earnings is defensible.

At the lower end of the growth assumption and the lower end of the multiple range, the implied value is approximately €1.28bn. At the upper end of both, the implied value reaches €1.95bn.

The base case: €1.28bn, or approximately 220p per share.

The bear case: growth disappoints at 5%, margins compress to 74%, and the multiple contracts to 18 times. Implied value approximately €846m, or around 145p. The current price of 190p sits about 30% above the bear floor.

The bull case: growth runs at 14%, margins hold at 78%, and the multiple expands to 25 times as the debt-free capital return story gets priced in. Implied value approximately €1.95bn, or around 330p.

This is not a business trading at a severe discount to obvious value. It is a good business at a fair price, temporarily de-rated by a policy shock, with a specific catalyst — the Estonian comparison base — that is already reversing.

So, the case for owning it is not that it is obviously cheap. The case is that the de-rating was an overreaction to a temporary problem in a business that has now demonstrated it can absorb that problem without a revenue decline.

The distinction between what BCG earns on its own capital and what an investor earns at today’s price is worth dwelling on. BCG’s return on tangible capital is essentially infinite — the platform requires almost no tangible capital to operate.

But the investor buying at 190p and an enterprise value of approximately €1.05bn is paying for that extraordinary capital efficiency at a 4.9% owner earnings yield. The business earns extraordinary returns. The investor earns fair returns. These are different numbers, and conflating them is one of the most common errors in quality growth investing.

BCG is a wonderful business at a fair price. Not a fair business at a wonderful price.

Share

The balance sheet

BCG ended October 2025 with net cash of €5.1m. Six months earlier it had net debt of €3.6m. The transition happened because BCG generated €30.6m in operating cash in the half, used €12.5m to pay the final dividend, bought back €6.4m of shares, and voluntarily repaid €10m of the bank loan.

The remaining loan balance is €15m, maturing in July 2026. This is trivial against the cash generation rate.

BCG generates the equivalent of the entire loan balance in operating cash every six months. Watch the refinancing terms when the facility matures, not because there is a solvency concern but because the interest rate will likely improve if they choose to maintain any debt capacity.

Management has said the company could be debt-free by the end of the current financial year; April 2026. When that happens, the full approximately €60m of annual operating cash flow becomes available for buybacks and dividends.

At the current market capitalisation of £880m, that is a total cash return yield of around 7%, growing at double digits. That is what the capital return conversation will look like in a few months.

The item on the balance sheet that looks alarming until you understand it is the goodwill. BCG’s balance sheet shows €330m of goodwill out of €386m of total assets. That is 85% of total assets in a single line item. For most businesses that would be a serious red flag.

For BCG it reflects the structure of how the business was assembled. Before the IPO, private equity assembled the portfolio of Baltic classifieds portals through a series of acquisitions. The goodwill represents the premium paid for those businesses above their book value. The assets underlying the goodwill are not tangible — they are the brand relationships, the audience, the network effects, the professional workflows. These are real assets.

They generate €60m of operating cash per year. They have never required impairment. BCG has not paid too much for businesses that subsequently disappointed. The goodwill is the price of buying dominant market positions and those positions have produced exactly what was paid for.

The risks

Let’s go through the risks, because there are things we need to be aware of with Baltic Classifieds.

The first is the free-listing attack.

If a platform with existing Baltic user trust (Facebook Marketplace for example) decided to offer free listings at scale across all three Baltic states, the consumer C2C side of BCG’s business would come under genuine pressure.

Private sellers deciding whether to pay €42 to list their car or €0 to list it on Facebook will choose free. The question is what percentage of BCG’s economics depends on private seller behaviour versus professional subscriber behaviour.

On the Auto segment, B2C dealer subscriptions represent the majority of revenue. Dealers will not stop paying for Autoplusas because Facebook offers free listings — they need the professional tools and they need the inventory depth that comes from a dominant platform.

On Real Estate, professional broker subscriptions and the workflow tools like Property Price Compass represent the core economics. A free-listing generalist platform does not threaten that. The vulnerable segment is generalist C2C, which is 15% of group revenue and where the competitive moat is weakest.

The early warning signal for this risk is BCG’s Auto and Real Estate C2C active ad volumes. If those numbers start declining for multiple consecutive quarters without a named policy explanation, it is the signal that inventory depth — the foundation of the flywheel — is eroding. Track those numbers every half-year.

The second mechanism is regulatory.

BCG raises prices every year on customers who have no practical alternative. That is the definition of a business that attracts antitrust scrutiny.

The Estonian Competition Authority has active enquiries open. BCG’s directors have assessed the likelihood of a material outflow from those enquiries as remote. I give that assessment weight because BCG’s market, while dominant, is also small. The harm from BCG raising annual prices on Lithuanian real estate brokers is genuinely minimal in the context of the national economy. The antitrust case is harder to make when the affected market is modest in size.

But the risk is real and observable. Any formal announcement from the Lithuanian or Estonian competition authorities opening an investigation specifically into BCG’s B2C pricing practices would change the valuation materially. The primary driver of BCG’s revenue growth is yield expansion. A cap on that expansion is a cap on the growth thesis.

And we can’t leave out macro/geopolitical risk.

I rarely include macro in my analysis, but in this case, we must address the elephant in the room.

Lithuania, Latvia, and Estonia sit at the eastern edge of NATO, sharing borders with Russia and Belarus. Estonia and Latvia border Russia directly.

Lithuania borders the Kaliningrad exclave and Belarus. These are not abstract geographic observations. The three Baltic states have lived under Soviet occupation within living memory, they understand the threat more clearly than most of their NATO allies, and they have responded to it seriously. All three now exceed NATO’s 2% of GDP defence spending commitment, with Estonia leading European allies at roughly 3.4%. NATO has substantially increased its forward presence in the region since 2022, with multinational battle groups deployed in each country.

None of this eliminates the risk. A significant escalation in the wider conflict or a direct threat to Baltic territory would affect consumer confidence, transaction volumes, and the attractiveness of Baltic equities to foreign investors in ways that BCG’s moat cannot insulate against. Small open economies with active property and labour markets are sensitive to geopolitical uncertainty in a way that larger, more diversified economies are not.

The Baltics have been pricing in proximity to Russia for thirty years. Their institutions are strong, their NATO integration is deep, and their populations have not needed recent events to remind them what is at stake. BCG’s business has compounded through every period of elevated tension since 1999. That does not make the risk theoretical. It means the risk is understood and priced locally in a way that external investors sometimes overreact to.

Of the risks in this piece, geopolitics is the one I think about least in terms of the day-to-day thesis and most in terms of position sizing. It is not a reason to avoid BCG. It is a reason not to make it the only thing you own.

The verdict

After going through all of this, here is where I land.

Going through the segment data, the management track record, the acquisition history, and the full financial picture has not weakened the case to me since writing the thesis brief. Real Estate is the strongest it has ever been. Auto is temporarily impaired in a way that is already reversing.

Management has done exactly what it said it would do since listing. The balance sheet is about to become debt-free. The capital return story is getting more interesting by the half-year.

If I had to name one thing that would most change my view, it would be evidence that the professional workflow integration is not as sticky as the product narrative suggests.

If broker churn on Aruodas.lt were rising, or if dealer retention on Autoplusas were declining, or if the ARPU growth started requiring unusual commercial concessions to maintain, any of those would suggest the moat is less embedded than it appears. None of those signals are present in the current data.

The verdict

After going through all of this, here is where I land, and it is not where I expected to when I started:

First, the business is excellent. The moat is real and deepening. Management has earned trust. The Estonian shock is temporary and the comparables will flip. None of that has changed in my mind after going through every corner of this company.

What has changed is my view on the price and what I am being paid to own it.

At 4.9% owner earnings yield, BCG is fairly priced for what it is.

Not cheap. The base case gets you to around 220p, which is roughly 15% from current levels.

That is a reasonable return for a dominant franchise in a small market. But reasonable is doing a lot of work in that sentence. Fifteen percent upside, with a 4.9% starting yield, is not the kind of asymmetry that makes me want to concentrate capital.

The bull case requires things to go better than guided. The bear case, where regulation caps pricing or a free-listing platform gets traction, does not look remote enough to dismiss.

And then there is the… map.

I laid out the geopolitical picture honestly in the section above. The Baltics have strong institutions, deep NATO integration, and thirty years of experience pricing in proximity to Russia. I believe all of that.

But when I combine a valuation that requires things to go right with a geography that carries tail risk I cannot model, I find myself looking for a better combination elsewhere.

And not because BCG is a bad investment. Because my opportunity cost is real and the risk-adjusted return here does not clear my bar.

I bought a small starter position in BCG earlier. I have since sold it. Not because the thesis broke (it did not) but because the price never gave me enough margin of safety to sit comfortably with the geography.

A wonderful business at a fair price is still a fair price. I want wonderful businesses at prices that make the geopolitical risk feel adequately compensated.

If BCG trades down toward 145p to 155p, the conversation changes entirely. At that level the owner earnings yield moves above 7%, the bear case is closer to priced in, and the same geopolitical risk starts to feel adequately compensated. That is the price where I would consider to own this properly rather than tentatively.

Until then: thesis intact, position closed, watching carefully.

Sources and references

• Baltic Classifieds Group PLC — Half Year Results for the Six Months Ended 31 October 2025. Published 4 December 2025. Available at balticclassifieds.com.

• Baltic Classifieds Group PLC — Half Year Results Presentation, Six Months Ended 31 October 2025. Published 4 December 2025.

• Baltic Classifieds Group PLC — Annual Report and Accounts 2025. Financial year ended 30 April 2025.

• Maanteeamet (Estonian Road Administration) — Monthly vehicle registration and transaction data, calendar year 2025.

• Similarweb — Time on site data, leadership position vs nearest competitor. Referenced via BCG H1 2026 results presentation.

• Google Analytics — Baltic site visit data, averaged based on Baltic population statistics. Referenced via BCG H1 2026 results presentation.

• Skandinaviska Enskilda Banken (SEB) — Nordic Outlook Update, November 2025. GDP, inflation, unemployment, and wage growth forecasts for Baltic states.

• Eurostat — Real GDP per capita CAGR, 2000–2024.

• Swedbank — Average apartment prices in Vilnius, Riga, and Tallinn, calendar H1 2025.

• State Enterprise Centre of Registers Lithuania, Land Register Latvia, Land Board Estonia — Real estate transaction volumes.

• Euromonitor — E-commerce market growth, Lithuania and Estonia, 2019–2028.

• Börsdata — Financial data export, Baltic Classifieds Group PLC, historical annual and quarterly figures.

Disclaimer
This Deep Dive is an educational breakdown of a public company based on information available in the materials provided (e.g., annual/quarterly reports, investor presentations, earnings transcripts) and my interpretation of that information. It is designed to be a “bolt-on” intelligence layer to your own due diligence — not a replacement for it.

Independence: I do not accept compensation of any kind from the companies discussed. My research is driven solely by my personal search for high-quality compounders.

Skin in the Game: Unless otherwise stated, assume the author may hold a long position in securities mentioned. Any position creates bias — treat this as commentary, not gospel.

Not Financial Advice: Nothing here is investment advice, a recommendation, or a solicitation. I am not a financial advisor. You are responsible for your own decisions. The stars rating is not a buy recommendation, but meant as a guide to understand the quality of the financial statement of the respective companies.

Error & Update Risk: Financial statements change, companies restate, guidance evolves, and I can be wrong. Verify key figures in the primary filings and consider reading the footnotes before deploying capital.

First off, let me tell you how this week’s Deep Dive actually came to be.

I just finished reading Christopher Mayer’s fantastic book, 100 Baggers, for the fourth time. Yup, the fourth. When you find a legitimate blueprint for building generational wealth, you study the architecture until you memorize the floor plan.

After putting the book down this last time, I decided to put theory into practice. I sat down and built a custom stock screener specifically designed to hunt for the traits Mayer talks about: insanely high returns on capital, long runways for reinvestment, and management teams with real skin in the game. I wanted to find the compounding machines of tomorrow before everyone else caught on.

But here is where the story gets really interesting.

As the ticker symbols started spitting out of my screener, I noticed a massive, undeniable coincidence. A huge chunk of the companies lighting up my list were exactly the same businesses sitting right inside legendary British investor Terry Smith’s current portfolio.

If you’ve been reading Atomic Moat for a while, you know I hugely admire Terry Smith. His entire investing philosophy is beautifully, brutally simple: buy good companies, don’t overpay, then do nothing.

And this philosophy aligns perfectly with holding on to potential 100-baggers.

When my 100 Baggers screener started rhyming perfectly with the Fundsmith playbook, I knew immediately that I was digging in the right spot.

And sitting right there, gleaming near the top of that meticulously filtered list, was a clinical research organization called Medpace.

On the initial screen, the numbers looked almost too good to be true. The return on capital was nuclear, the cash generation was pristine, the buybacks were massive, and the founder was still steering the ship. It looked incredibly interesting, despite being a tad on the larger side, market-cap-wise.

So, I did what we always do here: I rolled up my sleeves, ignored the Wall Street noise, and ripped into the actual SEC filings to see if the engine matches the paint job.

Here is what I found.

Subscribe now

If You’re In a Rush

• What they do: They are the outsourced operational engine for clinical trials. When a biotech company invents a new drug, Medpace steps in to recruit the human subjects, run the specialized lab tests, manage the massive datasets, and navigate the regulatory red tape.

• Why it’s ‘hated’: CROs sit straight downstream of biotech funding cycles. If biotech catches a cold, CROs get pneumonia. On top of that, Medpace carries a persistent “founder control + related-party optics” fog that can keep institutional money managers incredibly twitchy, even when the actual P&L results are squeaky clean.

• What fixes it: To shake off the bears, their awards need to stay visibly ahead of their revenue (meaning a book-to-bill > 1). The backlog needs to actually grow (not just convert into cash), and FY2026 needs to land safely inside their guidance without relying on any “working-capital magic” to make the math work.

What is Medpace?

If you are new to Medpace, your first thought may be if they are biotech company. And people are easy to dismiss biotech companies due to their riskier, binary nature.

Medpace invents exactly zero drugs.

They own zero drug patents.

If one of the trials they run cures a major disease, Medpace does not get a single dime of the royalty checks.

They are a pure service business.

Think of them like a general contractor building a custom home. The biotech company is the wealthy architect who drew the blueprints and owns the land. Medpace is the construction crew. They show up with the hard hats, pour the concrete, run the plumbing, and make sure the house passes the strict city inspections (the FDA).

Here is why investors love this specific setup: the biotech company takes 100% of the scientific risk. If a drug completely fails in Phase III testing and turns out to be useless, the biotech company’s stock might collapse to zero.

Medpace generally gets paid for work performed, but it does not get paid for future work if projects are cancelled or delayed.

I like to think of them like they sell the picks and shovels to the gold miners. Whether the miner strikes gold or finds dirt, Medpace gets paid for the shovel. And this is a business model I like.

Imagine you’re a biotech CEO with a molecule that looks brilliant on paper. Now imagine the nightmare of turning that molecule into an FDA-approved drug. You have to coordinate investigators, patients, specialized labs, data, and regulators across the globe, without blowing your budget or ruining the dataset.

That exact mess is the problem Medpace sells relief from.

In their own words, Medpace is a global provider of clinical research-based drug and medical device development services.

They handle full-service Phase I–IV clinical development, plus central lab coordination, project management, regulatory affairs, clinical monitoring, data management/analysis, pharmacovigilance, and post-marketing support. They also run the heavy science: bio-analytical lab services, clinical pharmacology, imaging, and ECG reading support.

The “product” here is not a software license. It’s execution under constraints.

How do they get paid?

They earn fees through customer contracts that are typically fixed-fee or unit-of-service. Most contracts are structured beautifully: they collect an upfront fee at signing, with the balance collected over the contract duration (via a billing schedule or milestones). Revenue is recognized over time as services are performed (they note progression is measured primarily using an input method, cost-to-cost).

August J. Troendle is the Chief Executive Officer and founder. He controls approximately 19.0% of the outstanding shares through direct and beneficial ownership (including shares held by Medpace Investors LLC).

In other words: this is a true owner-operator structure, not a “CEO-for-hire” passing through.

Share

So where’s the moat and the switching cost?

Medpace pitches a disciplined, integrated full-service operating model, complete with coordinated labs, Phase I clinic capabilities, and internal technology.

Their contracts can range from a few months to several years. But customers can terminate, typically with ~30 days’ notice terms, and cancellations happen in the ordinary course of business.

The switching costs here are less about “contractual lock-in” and much more about “operational pain.” Once a trial is underway, swapping out the team managing your sensitive data, monitoring, labs, and regulatory rhythm is like changing pilots mid-flight: it’s technically possible, but it’s expensive, risky, and slow.

Atomic Take: Medpace sells “complexity insurance” (not software) and the moat is integrated execution plus practical switching friction, even if the contracts aren’t ironclad.

Falsifier: Evidence of weakening practical stickiness (a sustained drop in awards/book-to-bill alongside rising cancellations or backlog pressure disclosed over time).

The Setup

Here’s the scoreboard we actually have right now, and the common-sense questions it forces us to ask.

FY2025 revenue was $2,530.2m and net income was a very healthy $451.1m. Net new business awards were $2,646.8m and ending backlog sat at $3,027.2m. Net book-to-bill for FY2025 was 1.05x.

So, what exactly is the market’s underlying argument hiding inside those numbers?

The market worry is incredibly simple, like worrying about your pantry during a snowstorm: FY2025 revenue grew fast, but backlog growth was modest.

If revenue is being fed by backlog conversion rather than fresh backlog replenishment, that growth can keep going… right up until the pantry is empty.

Subscribe now

The bull case is equally simple and grounded: backlog conversion rose because Medpace’s execution is fantastic, not because customer demand is weak.

If awards stay healthy and that conversion rhythm remains steady, Medpace can just keep compounding per-share economics.

The bear case is the classic, old-school CRO bear case: demand in this industry is notoriously cyclical, trial sponsors can terminate projects, and the financial pain usually shows up with a frustrating lag.

Add in that founder control and the related-party headlines I mentioned earlier, and the stock’s multiple can compress even if the actual P&L doesn’t implode.

Atomic Take: This is fundamentally a “duration vs price” debate. The underlying business looks exceptionally high-quality, but that $12.82B market cap is underwriting continued excellence and clean optics from here to eternity. Owner yields look more forgiving, but only if buybacks stay funded.

Falsifier: A full-year net book-to-bill falling below 1.00x (company KPI) would be the clearest, cleanest sign that underlying demand is slipping, even if the accounting on the surface still looks totally fine.

Founder & Control-Room

Dr. August Troendle, MD founded Medpace back in July 1992 and has been the CEO and Chairman ever since. Before he built Medpace, he actually worked at the FDA as a Medical Review Officer, and later as a manager at Sandoz (which is now Novartis) focusing on clinical development.

The man knows the bureaucratic maze from the inside out. He also holds an MD from the University of Maryland and an MBA from Boston University.

From a shareholder’s perspective, the point of bringing this up isn’t just to recite biography trivia. It’s all about incentives and control. The company’s 10-K explicitly flags his heavy influence via ownership and control as a core governance factor.

They also separately disclose related-party items, things like the company using CEO-owned lodging and leases, and certain travel services structures.

A recent SEC Form 4 from November 2025 shows Dr. Troendle reported direct sales of shares on 11/24/2025 and 11/25/2025.

Now, he is still reporting substantial direct and indirect beneficial ownership (including massive holdings through Medpace Investors, LLC), so he hasn’t abandoned ship. But when the founder sells, Wall Street notices.

How the Business Actually Makes Money

Medpace sells execution. Specifically: they sell the ability to run incredibly complex clinical trials with fewer nasty surprises, fewer regulatory delays, and fewer “why is this data missing?” moments that make drug developers age in dog years.

Subscribe now

Mechanically, CRO economics all come down to three linked plumbing pipes.

• Pipe 1: awards. When a trial sponsor officially signs on for work, Medpace records “net new business awards” (a company KPI). That’s the very top of the funnel. The water entering the house.

• Pipe 2: backlog. Those awards then become backlog (company KPI), which is simply the reservoir of contracted work that hasn’t yet been recognized as official revenue.

• Pipe 3: conversion. Finally, that backlog converts into revenue as the actual work is performed, recognized under their specific revenue recognition model.

That chain matters deeply because Medpace is not selling a simple one-time widget off a store shelf. They are selling complex, multi-period project execution where timing shifts, scope changes, and sudden sponsor decisions can dramatically change the shape of their reported growth.

The customer mix tells you exactly what kind of economic cycle you’re exposed to. By 2024 revenue mix (company classification), they were 79% small biopharma, 17% mid-sized biopharma, and just 4% large pharma.

That is a wonderful, highly profitable place to be when biotech funding budgets are healthy and flowing.

And it’s a completely jittery, nerve-wracking place to be when the funding taps dry up.

Now, here is the “don’t get fooled by the top line” part of the story.

In FY2025, Medpace had $1,037.5m of reimbursed out-of-pocket expenses.

Now, that is very real revenue under GAAP accounting, but economically, it behaves a lot more like pass-through spend (like reimbursing an employee for a flight) than it behaves like “we raised our prices and got rich.”

If pass-through spending rises, total revenue rises. That does not automatically mean the underlying service economics of the business improved one bit.

The Audit: Reimbursed out-of-pocket revenue can turn standard growth charts into a highly flattering funhouse mirror. It’s not fake money, but it’s certainly not proof of a widening moat, either. If you want to truly understand whether Medpace is getting better at its job, watch the operating income and the cash conversion, not just the total revenue fireworks.

Finally, there’s a subtle balance-sheet tell that’s incredibly easy to ignore if you’re skimming: advanced billings.

Medpace often receives customer cash well in advance (recorded as contract liabilities/“advanced billings”), which helps brilliantly to reduce credit risk and support their working capital. That’s a fundamentally good, sound operational design, but it also means their cash position can look “extra strong” in periods where advanced billings happen to rise.

Atomic Take: The true moat here is their operational integration and execution reliability, but you have to judge that moat through awards/backlog conversion, operating profit, and cash generation, not through flashy top-line growth.

Falsifier: Ending backlog falling year-over-year (as of Dec 31) while revenue still manages to grow would heavily imply that the cash flywheel is being fed by drawing down the backlog pantry, not by replenishing it.

Share

What Went Wrong

Let’s be clear: nothing “went wrong” in the sense of a broken business. The wheels didn’t fall off.

What went wrong is the strict setup for future expectations: FY2025 delivered a very strong year on paper, but some of the underlying KPIs raise the legitimate question of how much of that success was “conversion intensity” (working faster) versus actual “demand acceleration” (getting more work).

FY2025 revenue was $2,530.2m, up sharply year-over-year. Ending backlog, however, was $3,027.2m, which was only up modestly.

And importantly, the backlog conversion rate rose to 21.8%, up from 18.2%.

The other part of the puzzle is capital allocation.

They aggressively repurchased $912.9m of shares in FY2025, with another $821.7m still authorized for buybacks.

Cash and cash equivalents ended the year at $497.0m.

Big buybacks can be a phenomenal use of capital, but when the stock screens expensive on headline multiples, they can also read as ‘we’re paying up just to retire shares,’ which makes the market hypersensitive to any hint of a cycle wobble.

Then you’ve got the governance cloud. It’s the kind of thing that doesn’t just disappear just because the income statement behaves itself for a few quarters.

The CEO/founder controls about 19.0% of outstanding shares.

Rebound Catalysts

Alright, this isn’t some distressed turnaround checklist where we’re praying for a miracle. Medpace isn’t broken.

This is a proof checklist. We just need them to prove to Wall Street that the engine is still humming normally.

Catalyst 1: book-to-bill climbs. Think of book-to-bill like a bathtub. Awards are the water from the faucet; revenue is the water going down the drain. FY2025 net book-to-bill was 1.05x (company KPI), meaning the water is still rising, but barely.

A sustained improvement here would be the cleanest, simplest signal that new awards are outpacing revenue in a durable way. I want to see that tub filling up.

Catalyst 2: backlog growth re-accelerates. Ending backlog was $3,027.2m (as of Dec 31, 2025). That’s our reservoir of future earnings.

If this backlog starts growing faster than the low-single-digit pace implied by FY2025, the whole “we have a massive future runway” narrative improves overnight. Wall Street loves a long runway.

Catalyst 3: 2026 conversion visibility holds. The company laid their cards on the table: they disclosed an expectation that $1,890.0m–$1,910.0m of backlog would convert to revenue in 2026.

That is a beautiful chunk of locked-in visibility. If that conversion stays steady while new awards remain strong, the dreaded “biotech cycle risk” starts looking incredibly manageable.

Catalyst 4: guidance lands inside the box. FY2026 guidance calls for revenue $2,755.0m–$2,855.0m and net income $487.0m–$511.0m.

Here is a secret about high-expectation stocks: the market doesn’t always need a massive, hero-level beat. It just needs boring, predictable reliability. Hit the numbers. Don’t cause a fuss.

Subscribe now

Catalyst 5: cash stays boring too. FY2025 operating cash flow was a massive $713.2m. If cash generation remains this strong without relying on unusually favorable working-capital magic (which we’ll discuss below), that’s an absolute credibility amplifier. Cash in the bank is a fact; everything else is an accounting opinion.

Atomic Take: The so-called “rebound” here is simply awards and backlog proving to the market that FY2025 wasn’t just a one-year backlog conversion party.

Falsifier: Two consecutive quarters of net book-to-bill falling below 1.00x (company KPI) would be an early-warning air raid siren that the next shoe is dropping. If the tub is draining faster than it fills, you eventually run out of water.

Financial Quality Rubric

Profitability (4/5)

FY2025 income from operations was $534.9m on revenue $2,530.2m. Run the math, and that implies an operating margin around 21%. Let me be clear: that is premium services economics.

You don’t get a 21% operating margin in a highly competitive industry unless you are exceptionally good at what you do.

Growth quality (4/5)

FY2025 delivered very strong growth. But remember, the key quality question in the CRO business is not simply “did it grow?” — it’s “was that growth actively replenished by new awards, or was it just pulled forward by burning through the backlog?” Right now, it’s a mix, which is why we deduct a point.

Cash conversion (4/5)

FY2025 operating cash flow was $713.2m versus a net income of $451.1m. That is spectacularly strong conversion. However, some of that boost is tied to working-capital timing. We’ll separate the genuine fuel from the temporary turbo down below.

Balance sheet resilience (4/5)

There is absolutely no long-term debt outstanding (as of Dec 31, 2025) and they carry meaningful cash. The liabilities they do have skew toward operational obligations rather than dangerous financial leverage. This is the kind of balance sheet that lets you sleep soundly at night during a recession.

Governance & alignment (3/5)

Founder ownership is usually a massive positive for alignment. I want the captain heavily invested in the ship. But control combined with related-party disclosures can, and often will, keep a persistent valuation discount on the equity. It just makes institutions nervous.

Capital allocation discipline (3/5)

Share buybacks were huge in FY2025. Buying back your own stock can be incredibly shareholder-friendly, but it also violently increases the importance of not overpaying for it. If you buy high, you’re destroying value, not creating it.

Atomic Take: These financial statements read exactly like a premium operator should read. The only deductions are the durability uncertainty (the biotech cycle) and the optics (the governance cloud).

Falsifier: Operating margin compressing materially away from that ~21% zone (derived) while book-to-bill hovers near a stagnant 1.0x would strongly suggest real competitive or cost pressure is building, not just harmless cycle noise.

Share

The Statements

Balance Sheet (The Geiger Test)

As of Dec 31, 2025, cash and cash equivalents were $497.0m, total assets were $1,975.5m, and total liabilities were $1,516.4m.

A major, vital part of their current liabilities is advanced billings (also called contract liabilities) of $854.4m (as of Dec 31, 2025). That’s customer cash received before the work is recognized as revenue. In plain English: their biotech sponsors are pre-funding a meaningful slice of the Medpace engine. Warren Buffett loves insurance float; I love operational float. It’s free money to run the business.

Accrued expenses were $408.4m (as of Dec 31, 2025). On the asset side, goodwill sat at $662.4m. I don’t mind goodwill as long as the underlying earnings power of those past acquisitions stays durable and real.

The clean, common-sense takeaway: this is absolutely not a balance sheet built on dangerous financial leverage. It’s built on operational float. That is incredibly valuable, but it’s also directly linked to the health of client demand and project cadence.

Atomic Take: This balance sheet is operationally pre-funded rather than financially levered. It’s a massive strength that only becomes a risk if biotech sponsor behavior suddenly shifts and they stop paying up front.

Falsifier: Advanced billings falling sharply while accounts receivable/unbilled rises (both as of Dec 31) would be a huge red flag indicating weakening prepayment dynamics and potentially much higher collection risk.

Cash Flow (The Turbine)

Start with the adult reconciliation.

FY2025 net income was $451.1m. FY2025 operating cash flow was $713.2m. That is a massive spread. Why did cash flow dwarf net income so heavily?

A key contributor was simply working capital shifting in their favor: advanced billings increased by $143.8m, and accrued expenses increased by $97.1m. This was partially offset by an increase in accounts receivable and unbilled of $106.2m. When you collect cash before you pay bills, your cash flow looks like a superhero.

Now, let’s look at “capex-like” spending. Property and equipment expenditures were just $31.1m. This is a wonderfully asset-light business. Using a conservative proxy, a sustainable free cash estimate can be Operating Cash Flow minus those PPE purchases: roughly $681.9m.

They also present a non-GAAP free cash flow figure of exactly $681.9m, which lines up beautifully with our simple proxy.

And then, we have the big cash outflow: share repurchases of $917.4m. That’s the entire turbine output (and then some) getting shoveled straight into per-share math.

Atomic Take: The cash turbine is undeniably strong, but part of the extra torque this year comes from working-capital timing — and buybacks are aggressively consuming most of the cash generated.

Falsifier: A year where operating cash flow suddenly falls below net income without a clear, logical timing explanation would signal either a nasty working-capital reversal or structurally weaker cash economics.

Income Statement (The Reactor)

FY2025 revenue was $2,530.2m. Income from operations was $534.9m. Net income was $451.1m. Diluted EPS came in at a beefy $15.28.

But the revenue composition is what prevents lazy investors from making bad conclusions.

Direct service costs (excluding D&A) and reimbursed out-of-pocket expenses are the large line items here. Crucially, those reimbursed out-of-pocket expenses were a massive $1,037.5m.

If you want to think clearly about this business, you absolutely must separate their actual “service margin engine” from the “pass-through volume” of reimbursements.

Therapeutic mix shows a heavy concentration in areas where biotech spending tends to be highly meaningful. FY2025 oncology revenue was $747.6m and metabolic was $745.0m, with other therapeutic areas contributing the remainder. They are fishing in the ponds where the biggest fish live.

Atomic Take: The reactor runs incredibly hot on operating profit and cash — but that top-line growth needs careful interpretation because massive pass-through revenue can severely inflate the optics of growth.

Falsifier: EPS declining year-over-year while revenue somehow still rises would point directly to mix deterioration or hidden cost creep hiding behind the smokescreen of pass-through growth.

Valuation (The Terry Smith Lens)

Let’s get down to brass tacks. You can love a business to the moon and back, but if you pay the wrong price, it will still break your heart.

At the time of writing, the share price is $451.75. They generated FY2025 diluted EPS of $15.28.

I am using the exact shares outstanding of 28,381,283 (as of Feb 6, 2026, shares).

Multiply the shares by the price, and that implies a total equity value of about $12.82B (derived: price × shares).

That is the sticker price for the entire farm today. Now we have to figure out if the soil is rich enough to justify it.

Quality Gate

Before we even talk about whether the stock is expensive, we have to ask if the business is actually any good. Investors like Terry Smith don’t buy junk, and neither do we.

FY2025 income from operations (EBIT) was a massive $534.9m on total revenue of $2,530.2m.

If you run the math, that gives you an ~21.1% EBIT margin.

FY2025 operating cash flow was $713.2m. Meanwhile, the money they actually had to spend to keep the lights on and the labs running—property & equipment expenditures—was a tiny $31.1m.

Atomic Take: Quality absolutely clears the gate here. This is a high-margin operator with serious, undeniable cash generation. They aren’t just making accounting profits; they are making cold, hard cash.

Falsifier: A full-year period where operating cash flow drops below net income without a very clear, temporary timing explanation (like a weird working capital or tax shift).

Owner Yields (The “what do owners get?” question)

Earnings yield:

If we look at standard earnings, the FY2025 earnings yield = EPS $15.28 / $451.75 = 3.38%.

Flip that upside down, and that’s a ~29.56x trailing P/E (derived). If you only look at that one number, you might choke on your coffee and call it too expensive.

FCF yield:

But earnings are an opinion; cash is a fact.

FCF ≈ operating cash flow of $713.2m minus PPE purchases of $31.1m = $682.1m.

If we take that FCF yield ≈ $682.1m / market cap $12.82B, we get 5.32%.

So, if you’re looking strictly at the “earnings multiple,” you’ll instantly call it expensive.

But if you’re looking at the actual “cash yield” they generate, it already starts to look significantly less premium and a lot more reasonable.

Atomic Take: On a pure owner-earnings basis, the stock looks meaningfully more attractive than the P/E headline suggests, entirely because their cash conversion is incredibly strong.

Falsifier: That cash proxy yield falling materially while the earnings multiple stays stubbornly elevated.

Capital Return Yield

This is where the story gets aggressive.

FY2025 share repurchases were a staggering $912.9m (USD) for 2,961,924 shares, with $821.7m still authorized for future buybacks (as of Dec 31, 2025, USD).

If you map that out, the buyback yield ≈ $912.9m / $12.82B = 7.12%.

But here’s the crucial reality check, the thing most investors miss: those buybacks actually exceeded the FCF proxy by about $230.8m. That “extra” return wasn’t magic. It was funded by actively reducing their cash on hand and shifting other balance sheet movements.

This matters deeply for how we frame the valuation:

• If buybacks continue anywhere near this FY2025 intensity, the per-share claim on earnings and cash can rise incredibly quickly (the share count retired is ~10% of year-end shares, proxy-derived). That is a monster tailwind.

• However, if buybacks are forced to revert to “FCF-funded only” because the savings account runs low, the buyback yield could be lower. But even then, it would still be potentially very meaningful given that strong ~5.3% FCF yield.

Atomic Take: If you treat your “owner return” as FCF yield plus sustainable net buybacks, the whole setup can look fairly decent in terms of value, even with a high level P/E on the surface. But only if the buyback machine remains credibly funded by recurring cash flow, not by simply drawing down the balance sheet until it hurts.

Falsifier: Repurchases slow down materially while the book-to-bill/backlog weakens (a nasty two-hit combo: less per-share support from buybacks, and less operating momentum from the business).

Insider Incentives

Wall Street loves to talk about “alignment,” but usually, it’s just a CEO with a few stock options. Not here.

The CEO/founder’s ownership is, as I stated earlier, officially disclosed at approximately 19.0% of outstanding shares (as of Dec 31, 2025).

When you own nearly a fifth of the company, it creates an unusually direct, burning incentive to heavily favor per-share compounding (aggressive buybacks, ruthless spending discipline, and fierce margin defense).

Yes, it also increases the control and optics risk we talked about earlier — but strictly on the topic of incentives: this is absolutely not a “hired-gun CEO paid in vibes” situation. He eats his own cooking.

Atomic Take: The entire incentive stack is brutally aligned with per-share outcomes, which makes this massive buyback program much more than just a corporate PR stunt.

Falsifier: Any SEC disclosure indicating a meaningful change or reduction in founder ownership/control, or a sudden, unexplained shift away from buybacks as the primary capital return mechanism.

Why Now?

At $451.75, here is exactly what you’re getting:

• A 3.38% GAAP earnings yield.

• A healthy 5.32% FCF yield.

• A massive 7.12% buyback yield

So the verdict is this: If Medpace is expensive or not entirely depends on which of those yields you believe is actually durable, and how safely you can underwrite the sustainability of those buybacks.

Atomic Take: If your core base case assumes that (a) their cash generation stays right near those strong FY2025 levels and (b) buybacks remain meaningfully funded without blowing up the balance sheet, it’s entirely reasonable to call this stock not expensive on an owner-return basis, even if standard stock screeners scream “expensive”.

Falsifier: Net book-to-bill drops < 1.00x for a full year (company KPI) and buybacks suddenly grind to a halt due to cash constraints — that lethal combination would permanently break the “attractive owner return” argument.

Risks (The Meltdown List)

Let’s be brutally honest. If our entire valuation case is built on the three-legged stool of “owner yield + aggressive buybacks + founder alignment,” then our risk checklist needs to attack that exact chain.

We aren’t going to waste our time with generic CRO industry hand-wringing. We need to know what breaks our specific machine.

Biotech funding cycle (demand shows up first in awards)

• FACT: FY2025 net new business awards were $2,646.8m and FY2025 net book-to-bill was 1.05x.

• HYPOTHESIS: If the biotech sponsor funding environment suddenly gets tight, it’s going to show up in the awards first. The backlog and the actual revenue will follow, but with a frustrating lag.

• MONITOR: Keep a hawk’s eye on the net book-to-bill and net new business awards each quarter (company KPI).

Cancellations and the “backlog isn’t money” reality

• FACT: The 10-K is refreshingly explicit here: awards and backlog can be impacted by scope, timing, and cancellations, and “the majority of our customers can terminate our contracts without cause upon 30 days’ notice.”

• HYPOTHESIS: A sudden cancellation wave makes our beautiful owner-yield stack look terrifyingly temporary because the revenue runway shrinks overnight.

• MONITOR: Watch the ending backlog and their specifically disclosed 12-month conversion expectation ($1,890.0m–$1,910.0m expected to convert to net revenue in 2026, as of Dec 31, 2025, USD).

Working-capital “turbo” reverses (cash can be timing theater)

• FACT: FY2025 operating cash flow was a massive $713.2m. But that was helped significantly by working-capital movements including advanced billings adding +$143.8m and accrued expenses adding +$97.1m, partly offset by accounts receivable/unbilled dropping by -$106.2m.

• HYPOTHESIS: If the growth in advanced billings slows down or reverses, their operating cash flow (CFO) can step down violently, even if the operating profit (EBIT) still looks perfectly fine.

• MONITOR: Track advanced billings, which sat at $854.4m, and accounts receivable/unbilled at $402.1m, and see where they move directionally year-over-year.

Human capital / culture risk (owner-yield sustainability)

• FACT 1: The company says success depends on attracting/retaining qualified staff and highlights internal promotion pipelines.

• FACT 2: External employee-review data shows low overall ratings and low CEO approval.

• HYPOTHESIS: If the culture drives persistent churn, Medpace may need to “buy” retention via higher comp/benefits or accept quality slippage — either way, margins and execution durability become less certain.

• MONITOR: Glassdoor rating/CEO approval trend, hiring, any evidence in filings that retention is becoming harder than management describes.
  Falsifier: A sustained deterioration in execution KPIs (book-to-bill/backlog quality) alongside evidence of talent strain (harder hiring/retention language in filings).

Atomic take: (note: I am 100% anecdotical here). From a strict investment standpoint, I believe an unpopular CEO isn’t always a deal-breaker. In fact, some of the best 100-baggers were run by “difficult” founders. And I think these are the reasons why:

• The “Shared Misery” Effect: Management at Medpace is as hard on themselves as they are on staff. They are not living high on the hog while the workers suffer; they are obsessed with winning.

• The Training Ground: Despite the hate, Medpace is known as the best place to learn the industry. It’s like the “Goldman Sachs” of clinical research. It’s miserable while you’re there, but your resume is golden afterward.

• Capital Allocation vs. Popularity: August Troendle probably doesn’t care about being liked. But it is obvious that he cares about ROE (Return on Equity) and share buybacks. He treats the company like a private partnership where the goal is maximum efficiency.

Falsifier: If employee sentiment stays poor yet the company continues to show consistent KPI performance with no retention stress appearing in filings, the “culture will break the machine” concern weakens materially.

Buyback sustainability (the valuation case lives or dies here)

• FACT: FY2025 repurchases were a staggering $912.9m (USD) for 2,961,924 shares, with $821.7m still remaining authorized (as of Dec 31, 2025, USD).

• FACT: In that same year, operating cash flow was $713.2m and property & equipment expenditures were $31.356m (FY2025, USD, US GAAP) — meaning those repurchases exceeded a conservative cash proxy (CFO − PPE) by roughly ~$231m (derived).

• HYPOTHESIS: If repurchases slow down materially (or swing from being a massive “accelerator” to completely “off”), our entire “not premium on owner return” framing weakens incredibly fast.

• MONITOR: Compare the repurchase pace versus CFO and the actual cash balance ($497.0m cash & cash equivalents as of Dec 31, 2025, USD).

Founder control

• FACT: August controls approximately 19.0% of outstanding shares via direct and beneficial ownership and this can exert significant influence over corporate actions requiring shareholder approval.

• FACT: Related-party transactions exist (e.g., a CEO-owned hotel is used for lodging; the HQ lease is with an entity owned by the CEO; and there are disclosed relationships with entities where executives have board roles/equity).

• HYPOTHESIS: Even if the actual business operations stay pristine, bad optics can act like a ceiling, capping the multiple you’re “allowed” to have in the market. Which matters deeply when Wall Street is debating whether it is expensive or not.

• MONITOR: Watch for any changes in the risk-factor disclosure language and any expansion in the scale or frequency of these related-party items.

Sneaky risk: “owner yield” depends on definitions

• FACT: Management actively defines Free Cash Flow as operating cash flow less capex and the principal portion of payments related to campus leases classified as deemed landlord liabilities (a non-GAAP definition).

• HYPOTHESIS: If those internal definitions start to shift (or if that lease component suddenly becomes much more material), our “FCF yield” comparisons get incredibly noisy, and the entire story gets much easier to misread.

• MONITOR: The consistency of the Free Cash Flow definition across future presentations and SEC filings.

Atomic Take: The biggest risks here aren’t “a bad quarter” — they’re strictly (1) awards/backlog weakening, and (2) buybacks losing sustainability, because that’s what actually makes the owner-yield stack look attractive.

Falsifier: A full-year net book-to-bill dropping below 1.00x (company KPI) combined with a visible deceleration in repurchases versus FY2025’s $912.9m (USD) pace.

The Atomic Verdict

Status: High Conviction ★★★★☆

(IMPORTANT: Star system is not a buy recommendation, only a judgement for the quality of the business based on my parameters. DYOR)

Here’s exactly why the “expensive” label gets significantly shakier once you stop blindly worshiping the P/E ratio and start doing the Terry Smith-style math:

The Owner-Yield Stack (FY2025 basis):

• Earnings yield: 3.38%

• FCF proxy yield: 5.32%

• Buyback “yield”: 7.12%

Watch that last number closely. In FY2025, Medpace spent roughly ~$231m more on share buybacks than the actual free cash the business generated (CFO minus PPE). They funded the difference simply by drawing down their existing bank account.

If you are buying this stock because you love that massive buyback yield, you have to decide exactly how long they can keep outspending their own cash flow before the corporate ATM runs dry.

Founder incentives matter immensely here because the capital allocation lever is so huge.

The CEO/founder controls ~19.0% of outstanding shares and can heavily influence corporate actions. You can dislike that control structure if you want — but the incentive to care deeply about per-share outcomes is absolutely not ambiguous.

Upgrade triggers (measurable)

• Net book-to-bill trends up from 1.05x (FY2025, company KPI) and backlog growth re-accelerates (ending backlog, as of year-end).

• Repurchases remain substantial but become definitively “cash-covered” (meaning CFO − PPE consistently covers the buybacks without a persistent cash drawdown).

• Evidence that trial cancellations remain strictly “normal course” (no visible hit to awards/backlog in their disclosures).

Downgrade triggers (measurable)

• Net book-to-bill drops below 1.00x for a sustained period (company KPI).

• Ending backlog declines year-over-year (as of Dec 31), or the 12-month conversion expectation steps down materially.

• Repurchase pace slows sharply despite the remaining authorization of $821.7m (as of Dec 31, 2025), suggesting either a real cash constraint or a management valuation/regret pivot.

Medpace sits right at the fascinating intersection of a Terry Smith quality engine and a Chris Mayer 100-bagger. It is a beautiful, highly profitable compounding machine hiding in the unglamorous plumbing of the biotech industry.

If Dr. Troendle can keep funding those massive buybacks with real cash flow, this execution engine has plenty of runway to justify the premium price tag.

Now, I want to hear your take:

Are you willing to pay up for this level of execution, or does the biotech funding cycle keep you safely on the sidelines?

Drop your verdict in the comments below. Let’s debate.

References

• Medpace Holdings, Inc. — 2025 Investor Presentation (PDF: “Medpace 2025 Investor Presentation.pdf”).

• Medpace Holdings, Inc. — Q4/FY2025 Earnings Presentation (PDF: “MEDP_4Q25_Earnings_Presentation.pdf”).

• Medpace Holdings, Inc. — Form 10-K (FY ended December 31, 2025) (PDF: “0001668397-26-000006.pdf”).

• Financials / model workbook (XLSX: “MEDP-Medpace Holdings Inc (1).xlsx”).

Company / regulator sources referenced for narrative context (founder / insider activity)

• Medpace corporate website — Leadership bio page for Dr. August J. Troendle (founder/CEO background).

• SEC EDGAR — Form 4 insider transaction filing(s) from November 2025 referenced in the founder/insider activity discussion.

Disclaimer
This Deep Dive is an educational breakdown of a public company based on information available in the materials provided (e.g., annual/quarterly reports, investor presentations, earnings transcripts) and my interpretation of that information. It is designed to be a “bolt-on” intelligence layer to your own due diligence — not a replacement for it.

Independence: I do not accept compensation of any kind from the companies discussed. My research is driven solely by my personal search for high-quality compounders.

Skin in the Game: Unless otherwise stated, assume the author may hold a long position in securities mentioned. Any position creates bias — treat this as commentary, not gospel.

Not Financial Advice: Nothing here is investment advice, a recommendation, or a solicitation. I am not a financial advisor. You are responsible for your own decisions. The stars rating is not a buy recommendation, but meant as a guide to understand the quality of the financial statement of the respective companies.

Error & Update Risk: Financial statements change, companies restate, guidance evolves, and I can be wrong. Verify key figures in the primary filings and consider reading the footnotes before deploying capital.

New to The Atomic Moat? This analysis of Salesforce ($CRM) is a prime example of how we dissect high-quality compounders. If you want these deep dives sent to your inbox, join 1,000+ other investors below.

Join for Free

Folks, the market loves to panic first and do the math later. Right now, Wall Street is having a full-blown existential crisis over enterprise software.

The reigning narrative? AI is going to eat the SaaS application layer alive. The fear is that autonomous AI agents will simply “DIY” complex workflows. Why pay for expensive per-seat licenses when a smart chatbot can do the job for free?

Right now, the market is throwing the baby out with the bathwater. We’re watching highly durable, cash-minting businesses get priced like melting ice cubes simply because they happen to sell software.

But indiscriminate panic is exactly where the best opportunities hide. And that led me straight to the granddaddy of the industry: Salesforce.

I wanted to look under the hood and strip away the noise. Is the Salesforce fortress actually cracking under the weight of AI disruption, or is Wall Street just hallucinating a structural decline for a company that is still printing billions in cash?

Let’s dig in and find out.

Subscribe now

If You’re In a Rush

Let’s break this down to the studs before we get into the weeds:

• What they do: Salesforce sells enterprise “systems of record” for customer work—sales, service, data, and automation. They make their money mostly as subscription & support revenue.

• Why it’s hated: The market is currently flirting with the terrifying idea that AI agents will simply “DIY” the application layer. Wall Street thinks SaaS is going to get commoditized, which means multiples compress even if the underlying fundamentals actually improve.

• What fixes it: Proving that “agentic” demand is additive—not cannibalizing. We need to see that bookings, backlog, and cash conversion stay rock-solid while the AI stack shifts.

• Atomic Position: No position, but on my watchlist.

The Setup

The stock is sitting at $189.72 per share today (Feb 14th 2026). That is a long, painful drop from the $358 top it hit back in December 2024.

On the inside, the scoreboard looks a heck of a lot sturdier than the price action suggests. In Q3 FY26, Salesforce reported $10.259B revenue and a 21.3% GAAP operating margin, alongside a very healthy 35.5% non-GAAP operating margin.

On the outside, however, sentiment is being governed by a very specific, very loud fear. The fear is that generative AI is “injurious to the SaaS-based application layer,” and that enterprise customers will simply DIY their way around legacy vendors.

Variant Perception

Let’s look at the three sides of this narrative coin:

• Market: “AI agents kill app-layer value capture; incumbents become expensive UI wrappers.”

• Bull: “Agents make systems-of-record more valuable; Salesforce monetizes trust + data + workflow + governance.” (By the way, this is the company’s “agentic enterprise” framing).

• Bear: “AI attach is theater; growth stays single digits; buybacks become the only growth story left.”

This deep dive is about whether Salesforce can convert the AI narrative from a radioactive leak (driving multiple compression) into a controlled reactor (driving forward demand + monetization + cash).

Atomic Take: Salesforce looks operationally healthier than its stock chart implies. But I’ll be blunt: it’s still trading under the massive “SaaS gets disintermediated” cloud until forward indicators force Wall Street to re-rate it.

Falsifier: Q4 FY26 cRPO growth misses the company’s ~15% Y/Y guide (Q4 FY26, %, company guide) — if they miss that, it would validate the bear thesis that “demand isn’t inflecting.”

How the Business Actually Makes Money

Salesforce is, at its core, a massive subscription engine with a services sidecar bolted onto it. Subscription and support revenue is $9.726B, versus professional services and other bringing in just $0.533B.

Mechanically, their compounding loop is beautifully straightforward. I love businesses like this:

1. Land a workflow that becomes the default answer to “how the company does customer work.”

2. Expand across products (and across departments) until the customer has hardwired their processes and permissions deep into your platform. It’s like pouring roots into concrete.

3. Renew, because ripping out the system-of-record is the enterprise equivalent of open-heart surgery performed by a committee. Nobody wants to do it.

Quick note: Salesforce uses a fiscal year, so ‘Q3 FY26’ is the quarter ended Oct 31, 2025. Think of it like the Premier League: the 2025–26 season includes matches played in both 2025 and 2026. Same idea here, where ‘FY26’ includes quarters played in 2025. Companies do this because a fiscal year that lines up with their business cycle makes budgeting and performance comparisons cleaner. And for Salesforce, it keeps the big end-of-year sales/renewal season from being split awkwardly across two reporting years.

In Q3 FY26, Salesforce’s revenue disaggregation is presented under the new “Agentforce” naming umbrella (and the company explicitly states the rename did not change revenue allocation).

That’s incredibly useful.

It tells you this is, at least partly, packaging and positioning. And make no mistake, packaging matters deeply when the market is busy repricing narratives.

Let’s look at the major subscription buckets (Q3 FY26, USD, GAAP):

• Agentforce Service: $2.495B

• Agentforce Sales: $2.297B

• Agentforce 360 Platform, Slack and Other: $2.180B

• Agentforce Integration and Agentforce Analytics: $1.393B

• Agentforce Marketing and Agentforce Commerce: $1.361B

Now, let’s talk about the “AI option.” Management isn’t pitching “hey look, we added a chatbot.” They’re pitching “we are the control plane for deployed agents inside the enterprise.”

The measurable proof points they’ve chosen to emphasize are adoption and ARR-style monetization, including Agentforce and Data 360 ARR reaching nearly $1.4B, and Agentforce ARR surpassing half a billion.

Atomic Take: Salesforce doesn’t need to “invent a new category”; it just needs to remain the system-of-record and become the system-of-agents that runs on top of it.

Falsifier: Agentforce paid deals stop growing quarter-over-quarter (paid deals were >9,500 and up 50% QoQ in Q3 FY26, company KPI) — if that stalls, it would imply the AI attach motion isn’t sticking.

Share

What Went Wrong

First, growth slowed. Multiples got severely compressed. Then, the market found a convenient story that rationalizes paying less for the stock forever.

What it shows is a shifting growth regime: subscription & support constant-currency growth steps down and sits at 9% through Q3 FY26.

That’s the basic setup for a re-rate. Once growth moves from being a “teen compounder” to “high-single-digit durability,” the market stops paying for endless possibility and starts demanding cold, hard proof.

Then comes the “Narrative Scare”: the belief that AI agents make the application layer less valuable, and that customers will DIY around vendors. Again, this isn’t speculative panic from the cheap seats. It is literally presented on the call as the investor expectation.

So what does Marc Benioff (and management) think of it?

Their stance is basically this: investors are pricing a world where AI agents DIY the application layer and shrink the value of SaaS, but management says customer reality is moving the other way.

They describe a “mismatch” between what the market expects (DIY replacement) and what they’re seeing in the field: companies tried building their own models/toolkits, then ran into the messy reality that enterprise deployment is the hard part.

In Benioff’s framing, the value isn’t in tinkering. It’s in shipping deployed “customer agents” that live inside real workflows.

That matters because it implies a likely path forward; not “magic growth next quarter,” but a migration from experimentation to production:

• Customers move from pilots to deployed agents in live workflows.

• Deployment forces governance: permissions, auditability, workflow integration, and real data context.

• The system-of-record becomes the control plane (because that’s where the workflows + context already live).

• Monetization shifts toward attach and expansion (agents/data/products layered onto the core), not just seat growth.

Atomic Take: Salesforce’s bet is that AI doesn’t delete the system-of-record, but it increases its importance, because enterprises will pay for agents that are governed, contextual, and reliably executable.

Subscribe now

Rebound Catalysts

Wall Street doesn’t need Salesforce to be a flawless diamond right now. It just needs the boogeyman story (the “SaaS gets disintermediated by AI” campfire tale) to stop being believable.

I’ve always said, you don’t need a perfect company to make money; you just need a mispriced one that proves the loudest skeptics wrong.

Here are the testable catalysts we need to watch. The beauty is, these are tied to the exact KPIs Salesforce is already disclosing. No guesswork required:

1. Backlog credibility stays strong

Think of the backlog like a farmer’s silo. If the silo is full of grain, you don’t worry about starving next winter.

Right now, Salesforce’s cRPO (current Remaining Performance Obligation) is sitting at a massive $29.4B, and total RPO is an even bigger $59.5B.

Even better, management is guiding Q4 FY26 cRPO growth at ~15% Y/Y. As long as that contracted backlog keeps swelling, the “DIY AI” threat remains a ghost story.

2. Agentforce monetization keeps moving line of sight

It’s one thing to have a slick AI presentation; it’s another to get enterprises to open their wallets for it. But the numbers are moving. Agentforce and Data 360 ARR reached nearly $1.4B, up an impressive 114% Y/Y.

Drill down further: pure Agentforce ARR surpassed half a billion, up a staggering 330% Y/Y. We are also seeing paid deals up 50% QoQ.

That is the sound of the cash register ringing, folks.

3. The DIY fear flips into “deployment pain” demand

Management describes a secular shift toward the “agentic enterprise.” Here is the plain English translation: building AI models in a lab is fun, but deploying them securely across 10,000 employees is a nightmare.

The “last mile” is where enterprise IT projects go to die. Ultimately, big companies prefer buying from vendors who can actually ship functional, secure products. When DIY pain sets in, they will call Salesforce.

4. The reacceleration clock stays intact

During the call, Robin Washington reiterated they’re on track to reaccelerate revenue in 12 to 18 months. I love this. That’s a claim with a hard deadline. Deadlines are fantastic for investors because they keep management honest and their feet held to the fire.

Atomic Take: The “way forward” for this stock isn’t about riding AI vibes — it’s about watching the backlog + paid adoption + reacceleration-by-deadline, with heavy cash returns acting as the stabilizer while the narrative fight plays out in the market.

Falsifier: Agentforce/Data 360 ARR is high but cRPO growth actually decelerates — that would be a red flag, implying they are generating temporary excitement without securing durable, long-term contracts.

Share

Financial Quality Rubric (1–5)

I like to look at the underlying plumbing of a business before I buy the house. Let’s grade the structural integrity here:

• Revenue durability: 4/5

  Make no mistake, this is still a highly sticky subscription business, not a low-margin services business.

• Margin discipline: 4/5

  They posted a 21.3% GAAP operating margin and an eye-popping 35.5% non-GAAP operating margin. The discipline is holding.

• Cash conversion: 4/5

  This is a cash-minting machine. Operating cash flow is incredibly robust versus the net income of $2.086B.

• Balance sheet resilience: 4/5

  Cash & cash equivalents were $8.978B balanced against noncurrent debt of $8.438B. They have the fortress they need.

• Capital allocation: 4/5

  The repurchases and dividends here are not symbolic, tiny gestures meant to appease activists; they are large and ongoing (we will get to the cash flow below).

• Narrative risk (AI + culture/politics): 2.5/5

  The AI disintermediation fear is loud enough that it literally had to be discussed on the earnings call.

  
  Separately, Wired reports employee backlash tied to Marc Benioff’s ICE comments.

  Never ignore culture. Internal politics can quickly become an execution drag during major platform transitions.

Now, let me remind you that non-GAAP margins can sometimes look like a superhero cape masking a dad bod.

GAAP is where the actual laws of financial physics live, and it includes items (like restructuring costs) that can linger a lot longer than management’s sunny PowerPoint deck suggests.

Atomic Take: Financially, Salesforce looks exactly like a mature compounder with a highly measurable AI call option attached to it. The entire debate is whether this AI option translates into durable, long-term economics or if it’s just a beautifully produced demo loop.

Falsifier: GAAP operating margin drops while revenue growth stays stuck in the high-single digits — that would suggest their newfound “discipline” was just a temporary phase, not a structural foundation.

Alright, let’s have a look at the financial statements.

The Balance Sheet (The Geiger Test)

Let’s look under the hood. As of Oct 31 2025, Salesforce held a massive $8.978B in cash & cash equivalents. Add to that marketable securities of $2.345B. They have plenty of dry powder.

But the “enterprise software reality” is also here: they are carrying goodwill of $52.4B. Now, this is perfectly normal for a company with a long, aggressive acquisition history (MuleSoft, Slack, Tableau, etc.). But it is exactly why forward demand signals matter so much. When growth stops, that mountain of goodwill stops looking like a history of smart buys and starts looking like a massive future impairment question.

On the liability side, the recurring-revenue engine shows its absolute beauty in unearned revenue: $14.996B. That is money sitting in their pocket for services yet to be rendered.

Atomic Take: The balance sheet is highly liquid and not debt-stressed. The key sensitivity here isn’t bankruptcy; it’s growth durability. Why? Because goodwill-heavy balance sheets absolutely hate stagnation.

Falsifier: A sustained decline in total RPO (as of quarter-end, USD, company metric) — that would be the canary in the coal mine, turning “normal goodwill” into “headline impairment risk.”

Subscribe now

Cash Flow (The Turbine)

Cash is the lifeblood of any business, and here’s the plain-language bridge of how the Salesforce turbine is spinning.

GAAP net income was $2.086B. But the actual cash coming in the door—Operating cash flow—was $2.316B.

You have to maintain the machine, so capital expenditures were $0.139B.

If we take OCF and back out those capex costs, the simple cash proxy (or free cash flow) is ~$2.177B.

Now, let’s look at the capital return reality check. Share repurchases were an astonishing $3.801B and dividends were $395M.

Did you catch that?

The money they are sending back to shareholders exceeds the simple cash proxy for the quarter. That means these returns are not being funded purely by that quarter’s operating cash after capex.

It tells you the tempo of their capital return program is incredibly aggressive.

Atomic Take: Cash generation is undeniably strong — but the capital return pace is even stronger. As an investor, you need to watch the turbine (the operating cash flow), and not just get distracted by the fireworks (the massive buybacks).

Falsifier: Operating cash flow growth misses the company’s ~13–14% Y/Y guide (FY26, %, company guide) — if that engine sputters, it would eventually force management to rethink their aggressive capital return tempo.

Income Statement (The Reactor)

Let’s look at the core reactor for Q3 FY26 (GAAP):

Revenue was $10.259B (USD). The gross margin is shown at an elite 78%. Operating income pumped out $2.188B, and the bottom line net income was $2.086B.

When you strip out the noise, non-GAAP operating margin was 35.5%, driven largely by the company’s adjustments.

Atomic Take: The core reactor of this business is highly stable. The entire debate on Wall Street right now is whether the new fuel (AI agents) actually increases the reactor’s output, or if it just adds a glossy new control-room label to the exact same machine.

Falsifier: Non-GAAP remains perfectly smooth while GAAP profitability silently deteriorates — that’s the moment when “adjustments” start doing too much heavy lifting and narrative work.

Valuation and Quality

Before I even look at a stock chart, a company has to earn the right to be in my portfolio by passing a basic quality check. Salesforce clears the “quality business” bar with room to spare.

Look at the profitability and cash conversion: they are sitting on a massive 78% gross margin and a 21.3% GAAP operating margin. When you have margins that thick, you have the ultimate shock absorber for whatever the economy throws at you.

Owner Yield

Alright, let’s talk about what we are actually paying for this machine.

• Share price: $189.72.

• Shares used in computing diluted EPS: 952M (Q3 FY26, shares).

Do the simple math, and that implies a rough market cap proxy of ~$180.6B (USD, derived from price × diluted shares).

Now, let’s look at the cash proxy anchored in the actual reported lines, not the adjusted fantasies. For the first nine months of the year (9M FY26), operating cash flow was $9.532B and capex was a measly $0.453B.

Subtract the capex from the operating cash, and your 9M FCF proxy is ~$9.079B (USD, derived).

If we annualize that 9-month figure as a straight run-rate (and I’ll be the first to tell you, this is a blunt tool, so treat it as such), you get ~$12.1B/year.

Put that against our market cap proxy, and it implies a run-rate FCF yield of ~6.7%.

Now, a run-rate math is not the gospel truth. It’s a Geiger counter. It tells you the “implied burden” the current stock price puts on the business; it doesn’t tell you the exact temperature of the reactor.

The reason is that the market is currently pricing the “SaaS scare” as a permanent, structural decline.

Management, on the other hand, is screaming from the rooftops that this is just a temporary mismatch, and they are putting a hard timeline on reacceleration: 12 to 18 months.

Because of the steep drop to today’s price of $189.72, the burden of proof has entirely shifted. You no longer have to believe in “hypergrowth to the moon” to make money here. You just have to believe in “durability + credible reacceleration.”

Atomic Take: At this level, the debate isn’t “is Salesforce cheap?” — it’s “is the market over-penalizing a highly durable cash generator just because the AI narrative temporarily spooked the multiple?”

Falsifier: If the 12–18 month reacceleration timeline slips meaningfully, and there is no forward-demand strength (like cRPO) to compensate, then you are stuck in a value trap: a cash machine with a permanently discounted story.

Subscribe now

Risks (The Meltdown List)

I never buy a stock without looking for the trap doors. Here are the very real risks we are monitoring:

1. AI disintermediation risk (the core “SaaS scare”)

• FACT: The fear that AI is “injurious to the SaaS application layer” is explicitly and officially an issue.

• HYPOTHESIS: Value capture shifts upward to the AI models and orchestration layers, leaving legacy apps as thin, easily replaceable wrappers.

• MONITOR: We watch Agentforce paid-deal momentum vs guidance. If they can’t sell the AI, they become the wrapper.

2. Backlog conversion risk

• FACT: RPO includes unbilled amounts that are not yet recorded on the balance sheet. It’s contracted, but it’s not cash in the bank yet.

• HYPOTHESIS: The backlog stays high on paper, but it actually converts to cash much slower because enterprise AI deployments elongate and get bogged down.

• MONITOR: We need to watch the cRPO-to-revenue conversion trend and management’s commentary on deal cycles.

3. Capital return sustainability risk

• FACT: Repurchases were a massive $3.801B and dividends were $395M (Q3 FY26, USD, GAAP cash flow).

• HYPOTHESIS: If the core cash generation softens, the tempo of these massive returns will slow down—and investor sentiment will take a brutal second hit.

• MONITOR: Keep a hawk’s eye on FY26 operating cash flow growth vs guidance.

4. M&A / integration risk

• FACT: The Informatica acquisition is referenced as completed in the presentation deck.

• HYPOTHESIS: Digesting massive acquisitions is hard. Integration could distract management from execution right in the middle of a critical AI platform shift.

• MONITOR: Watch for any slippage in margin discipline and their forward guidance posture.

5. Culture / politics / brand risk (the sneaky risk)

• FACT: Wired reports employee backlash tied to Marc Benioff’s ICE-related comments, including calls for denouncement and restrictions on use.

• HYPOTHESIS: Talent bleeding and customer optics degrading at the exact worst possible moment (during a major platform transition) can cripple a software company.

• MONITOR: Watch for repeated internal friction becoming external customer friction (but only react when explicitly acknowledged in primary sources, not just rumor mills).

Atomic Take: The biggest risk here is not “Salesforce can’t do AI.” The biggest risk is “the market simply decides Salesforce can’t capture AI economics,” and keeps the multiple severely compressed until the KPIs force a reset.

Falsifier: Agentforce/Data 360 ARR stays strong on the surface, but cRPO stalls out — generating PR excitement without contracted demand is exactly how narratives die.

The Atomic Verdict

★★★★☆ (4/5)

(IMPORTANT: Star system is not a buy recommendation, only a judgement for the quality of the business based on my parameters. DYOR)

The cash engine is humming.

They posted $2.188B in operating income and $2.086B in net income (USD, GAAP), backed up by operating cash flow of $2.316B (USD, GAAP). Let me be clear: this is not a liquidity panic or a margin collapse story. The physics of the business are sound.

Why the stock can still act allergic anyway: The market is currently trading a thesis about the future of SaaS economics—and that fear is directly captured in the transcript.

Management’s rebuttal is essentially: DIY has peaked, real enterprise deployment is hard, and customers are moving from science-fair experimentation to production agents. Wall Street will only believe this when the backlog and monetization make it too expensive not to.

Upgrade Triggers (Measurable things that make me more bullish):

• cRPO growth stays around the guided ~15% Y/Y level while subscription & support constant-currency growth improves from its ~9% rut.

• Agentforce paid deals continue meaningful QoQ growth from the >9,500 paid baseline.

• Hard evidence that the “12–18 months” reacceleration clock is actually showing up in reported growth rates, not just in executive talking points.

Downgrade Triggers (Measurable things that make me run for the hills):

• cRPO growth misses guidance and trends down from the $29.4B base.

• Agentforce/Data 360 ARR growth decelerates sharply from that impressive 114% Y/Y level (Q3 FY26, company-defined).

• GAAP operating margin compresses meaningfully from the 21%+ level while growth stays stuck in the high-single digits.

Atomic Take: Salesforce is a highly disciplined cash engine fighting a brutal narrative war. And in the stock market, narrative wars end when the next two quarters’ KPIs make one side look completely unserious.

Falsifier: If forward demand (cRPO) fails to support the reacceleration timeline, this stock stays Overhang-Heavy regardless of how good the cash returns look.

Note from me:

On Wednesday, February 25, 2026, Salesforce reports Q4 and full-year FY2026 results after the close. Headlines will do what headlines do, so the “cheat sheet” is the underlying scoreboard: cRPO (forward demand), Agentforce paid deals (AI monetization traction), and whether subscription & support constant-currency growth and GAAP operating margin are holding up.

Reference List (Sources Used)

1. Salesforce, Inc. “Q3 FY26 Quarterly Investor Deck”. Used for: KPI tiles, subscription & support disaggregation, constant-currency growth tables, regional growth, capital return framing, and non-GAAP expense profile (slides 5, 8, 9, 15, 16, 17, 18, 19).

2. Salesforce, Inc. “Q3 FY26 Earnings Press Release”. Used for: GAAP income statement items, margins, cash flow figures (CFO, capex, FCF), balance sheet line items (cash, unearned revenue), and RPO/cRPO disclosures.

3. Salesforce, Inc. “Q3 FY26 Earnings Conference Call Transcript” (Salesforce Investor Relations / Q4 CDN-hosted transcript; accessed Feb 14, 2026). Used for: the “SaaS scare” framing (investor question), management’s rebuttal (DIY vs deployed agents), and the 12–18 month revenue reacceleration timeline language.

4. Salesforce Investor Relations site (investor.salesforce.com; accessed Feb 14, 2026). Used for: document hosting/verification context.

5. Wired (web article; accessed Feb 14, 2026). “Letter Salesforce Employees Sent After Marc Benioff’s ICE Comments.” Used for: culture/politics headline risk context.

6. Nasdaq (web page; accessed Feb 14, 2026). “CRM Insider Activity.”

Disclaimer
This Deep Dive is an educational breakdown of a public company based on information available in the materials provided (e.g., annual/quarterly reports, investor presentations, earnings transcripts) and my interpretation of that information. It is designed to be a “bolt-on” intelligence layer to your own due diligence — not a replacement for it.

Independence: I do not accept compensation of any kind from the companies discussed. My research is driven solely by my personal search for high-quality compounders.

Skin in the Game: Unless otherwise stated, assume the author may hold a long position in securities mentioned. Any position creates bias — treat this as commentary, not gospel.

Not Financial Advice: Nothing here is investment advice, a recommendation, or a solicitation. I am not a financial advisor. You are responsible for your own decisions. The stars rating is not a buy recommendation, but meant as a guide to understand the quality of the financial statement of the respective companies.

Error & Update Risk: Financial statements change, companies restate, guidance evolves, and I can be wrong. Verify key figures in the primary filings and consider reading the footnotes before deploying capital.

⚠️ MICROCAP ALERT: This is a MSEK 408 flyweight. Expect thin liquidity and wild swings. A single big trade can move the needle here, so don't bet the farm—size your position for the turbulence. This is not a buy recommendation; do your own research (DYOR).

New to The Atomic Moat? This analysis of AVTECH ($AVT B) is a prime example of how we dissect high-quality compounders. If you want these deep dives sent to your inbox, join 1,000+ other investors below.

Join for Free

Everyone seems to be looking for the latest flashy AI startup or the electric flying car or space company nowadays. But if you’ve followed my writing for a while, you know I have a soft spot for something a bit different.

I like the “picks and shovels”. The companies that don’t necessarily build the plane, but make the plane 10% more profitable just by existing.

That brings us to AVTECH.

I own this because it represents a specific kind of growth machine that is becoming increasingly rare in the microcap universe. Here’s why I’m holding this one for the long haul:

1. The “Invisible” Toll Booth

AVTECH has spent years embedding itself into the cockpit workflow. Once an airline like Wizz Air realizes they can save a fortune on fuel and hit their arrival times more consistently using this software, they don’t just “unplug” it.

It’s high-margin, mission-critical, and increasingly recurring. I love businesses where the customer would have a headache if they tried to leave.

2. A Fortress Built on Common Sense

A lot of small caps are “burning the furniture to keep the house warm”—they have massive debt and negative cash flow. AVTECH is the opposite.

They have a 91.8% equity ratio and zero long-term debt. This is a company that can survive a storm, which gives them the luxury of reinvesting when their competitors are just trying to keep the lights on.

3. The “Land and Expand” Runway

They are already on 2,200 aircraft, but here’s the kicker: most of those planes aren’t using the full suite of tools yet.

We aren’t just betting on them finding new customers (though they are); we’re betting on them selling more to the friends they already have. That’s the cheapest kind of growth there is.

4. Discipline That’s Actually Discernible

It’s easy for a CEO to talk about growth. It’s much harder to grow sales by 38% year-over-year while remaining profitable and paying out a dividend. That tells me management isn’t just chasing “growth at any cost”; they’re chasing profitable compounding.

The Bottom Line: AVTECH is still a “tiny” company in a massive global industry. They have the balance sheet of a conservative grandfather and the growth engine of a hungry startup. That’s a combination I’m willing to wait on.

Let’s dig in.

Subscribe now

If You’re In a Rush

• The Gig: AVTECH builds digital tools that help airplanes talk to air traffic control. They optimize for the stuff that keeps airline CEOs awake at night: fuel costs, emissions, punctuality, and safety.

• The “Ugh” Factor: It’s a microcap. That means low volume and price swings. Plus, selling to airlines takes forever, and currency swings (FX noise) can make a good quarter look messy on paper.

• The Fix: As long as that Annual Recurring Revenue (ARR) keeps climbing like a plane on takeoff, and they don’t spend like drunken sailors to get it, the story stays intact.

• Atomic Position: I hold shares in AVTECH at the time of writing. Please have this in mind when reading, as I can be biased.

The Setup: Numbers That Make You Grin

AVTECH’s year-end report is the kind of document that makes a spreadsheet geek smile and everyone else look for a calculator. It’s a microcap in a complicated industry, but the numbers are starting to look very good.

Let’s look at the hard evidence from the Q4 2025 print:

• Record Growth: Q4 2025 net sales hit MSEK 12.9. Management says that’s the highest quarterly revenue in their history. And get this: they did it even with a weak dollar working against them.

• The Annual Scorecard: For the full year 2025, sales climbed to MSEK 47.4 (up from MSEK 34.3 in 2024). Net earnings followed suit, jumping to MSEK 17.0 from MSEK 12.6.

Now, here is the “Microcap Paradox.” Wall Street usually thinks a company this size is either too small to matter or too small to trust. Both can be true on a Monday and totally wrong by Thursday. That’s why we ignore the “vibes” and stick to the facts.

The Fact Base:

• ARR: MSEK 51.2 (as of end of Q4 2025).

• Rule of 40: 67.6% (Growth + Margin). In the software world, anything over 40% is an “A.” 67.6%? That’s “Valedictorian” territory.

This isn’t a “turnaround” story. It’s a “scaling” story. They aren’t fixing a broken plane; they’re trying to see how high this one can fly without the engines overheating.

The Atomic Take: The debate isn’t whether planes will keep flying. The debate is whether AVTECH’s adoption flywheel stays intact while they start spending more on staff and development in 2026.

The “Show Me” Metric (Falsifier): If that MSEK 51.2 ARR drops for two quarters in a row, the thesis is grounded. Period.

How The Business Actually Make Money

AVTECH says they do “digital air traffic management.” That sounds like jargon designed to put you to sleep.

Let’s translate: they sell software that helps pilots and ground crews make smarter decisions in real-time.

Here is what we know for sure:

1. It’s a Subscription Model (The Holy Grail): They track ARR by taking the last month of the quarter and multiplying by 12. It’s a proxy, sure, but it shows they are moving away from one-off projects and toward “renting” their genius.

2. They Are Already On the Planes: Management says about 2,200 aircraft use their services. The beauty here isn’t just finding new airlines; it’s selling more “add-ons” to the planes they already have.

3. The “Wizz Air” Proof Point: You want to see how they scale? Look at Wizz Air. They went from testing products on 20 planes to rolling them out to the whole group of 200 planes this past year. That is how you win in this business; start small, prove it works, and then take over the whole fleet.

The Strategy: Management is getting ready to launch new “on-time” and “dispatch” tools this spring. If those land, the “upsell” story gets a massive boost.

The Atomic Take: This is a “land and expand” play. They’ve already landed. Now we watch the expansion.

The “Show Me” Metric (Falsifier): If FY2026 sales don’t beat MSEK 47.4 despite management talking about a “growing pipeline,” we’ll know the talk is cheap.

Share

What Went Wrong (The “No-Panic” Audit)

If you came here expecting a dramatic story of a company cratering into the earth, I’m going to have to disappoint you. This is the more irritating kind of story; a strong year where one quarter reminds you that microcaps don’t do “clean” narratives.

They’re messy, like a teenager’s bedroom.

Q4 2025 delivered record revenue, yet the bottom line didn’t do a victory lap. Why?

• The Currency Bug: Management flags an unfavorable USD/SEK impact. When the exchange rates wiggle, your paper gains can shrink.

• The Earnings Dip: Net earnings were MSEK 4.0 in Q4 2025, compared to MSEK 4.2 a year ago.

• The Investment Bill: They spent more on people (personnel costs) and ate a ~MSEK 1.1 one-off expense for an R&D tax repayment. They’re appealing it in 2026, but for now, the cash is out the door.

• Lumpy Cash: Q4 cash flow was positive but felt a bit thin because some customers were late with their checks.

This is a fork in the road. Management says they invested intentionally in sales and product development. In my book, spending money to make more money later isn’t a “miss”; it’s a strategy. But you have to make sure that “investment” doesn’t just become a permanent “expense.”

The Atomic Take: Nothing broke. But Q4 is a reminder that when a company is accelerating and reinvesting at the same time, the road is going to have some potholes.

The Rebound Catalysts: A Checklist for 2026

Management hasn’t just given us excuses; they’ve given us a checklist. Now we get to sit back and watch them try to tick the boxes.

1. New Blood in Sales: Philip Nordfeldt took over as Chief Commercial Officer (CCO) on January 19, 2026. His one job? Turn that tech into more ARR.

2. The Need for Speed: They’ve promised faster deployment. In aviation, “fast” is relative, but if they can get software onto planes quicker, the cash flows faster.

3. The “Full Meal” Strategy: Remember, many of those 2,200 aircraft are only using some of AVTECH’s tools. Selling a current customer a second or third product is way cheaper than finding a brand-new airline.

4. Spring Training: New functions for “on-time” and “dispatch” are hitting the market this spring. This is a near-term “put up or shut up” moment for the product team.

5. The Wizz Air Blueprint: Expanding from 20 to 200 planes with Wizz Air proved they can handle a fleet-wide rollout. Now, can they do it again with someone else?

The Atomic Take: These catalysts aren’t about the economy or the price of jet fuel. They are about execution. Either they do it, or they don’t. It’s that simple.

The Financial Quality Rubric

I like to grade companies on a scale of 1 to 5. Here’s how AVTECH stacks up:

• Growth Quality: 4/5 — Annual sales jumped from MSEK 34.3 to MSEK 47.4. That’s not a fluke; it’s momentum.

• Profitability Durability: 4/5 — They’re sporting a 34.7% operating margin for the year. That’s a massive cushion.

• Cash Conversion: 3/5 — Net earnings were MSEK 17.0, but operating cash flow was MSEK 12.1. It’s positive, but there’s a gap we need to watch.

• Balance Sheet: 5/5 — They have zero long-term debt and MSEK 33.9 in cash. This is a fortress. They aren’t going broke anytime soon.

• Reinvestment Discipline: 3/5 — They capitalize their R&D (turning expenses into “assets” on the balance sheet). It’s legal, but it can make profits look prettier than they are. We’ll keep an eye on it.

Share

Capital Efficiency: Getting More “Bang” for Your Buck

To me, AVTECH’s best hidden feature is that it doesn’t need a lot of “stuff” to make money. This isn’t just a feeling; it’s baked into the math. When you look at how little capital and how few people are required to keep this reactor humming, the efficiency is actually quite startling.

The “Per Head” Math:

• Sales Productivity: With MSEK 47.4 in sales and only 17 employees, they’re pulling in roughly ~MSEK 2.8 per person.

• Profit Productivity: Net earnings hit MSEK 17.0, which means each employee is essentially generating ~MSEK 1.0 in pure profit.

• The Cash Reality: Operating cash flow was MSEK 12.1, or roughly ~MSEK 0.7 per employee.

The Asset-Light Engine: Now, look at what they actually “own.” As of December 31, 2025, they had MSEK 63.7 in total assets. But here’s the kicker: MSEK 33.9 of that is just cold, hard cash sitting in the bank.

This means the business isn’t trapped in heavy machinery or expensive real estate. In Peter Lynch terms, it’s an “asset-light” dream. Their sales-to-assets ratio is roughly 0.74x. It tells us one thing loud and clear: this is not a hungry machine that needs constant infusions of cash to stay alive.

The Audit: The only place this efficiency can look a little “polished” is in the MSEK 14.1 of balanced development expenses. Because they capitalize some of their work (turning salary costs into an asset on the balance sheet), it can make the efficiency look slightly better than a pure cash-expense model.

The Atomic Take: The model is incredibly lean. The only catch? That capitalized development “asset” has to keep paying rent. If it doesn’t lead to more ARR and more cash, then the efficiency is just an accounting trick. So far, the gauges say it’s real.

The “Show Me” Metric (Falsifier): If operating cash flow starts to drop significantly below MSEK 12.1 while capitalized development stays high, we’ll know the efficiency is turning into “accounting-backed optimism” rather than real-world results.

Subscribe now

The Statements (The Geiger Counter Test)

The Balance Sheet: A Bunker

AVTECH is sitting on MSEK 63.7 in total assets and almost no liabilities. Their equity ratio is 91.8%. In plain English: they own almost everything outright.

The Risk: The only “isotope” to watch is the MSEK 14.1 in “balanced development expenses.” That’s money they’ve spent developing software that they expect to pay off later. If the software doesn’t sell, that “asset” vanishes.

The Cash Flow: The Turbine

Operating cash flow for the year was MSEK 12.1. What’s impressive? They’re paying a dividend of SEK 0.10 per share (~MSEK 5.6).

A microcap paying a dividend while growing sales at this rate is like finding a teenager who actually saves their allowance. It shows a level of discipline you don’t often see in the small-cap world.

The Income Statement: The Reactor

Sales are up, but personnel costs rose from MSEK 10.7 to MSEK 15.4 for the year.

The Audit: Management is betting on people. They are fueling the reactor. In 2026, we need to see that extra fuel turn into more “output” (sales), not just more “heat” (expenses).

The Atomic Take: The financials are rock solid, but the “truth serum” will be whether the cash flow keeps pace with the earnings report in the next six months.

The “Show Me” Metric (Falsifier): If operating cash flow (CFO) stays below MSEK 12.1 while they claim they are “thriving,” something is leaking in the pipes.

Share

Valuation: When the Market Starts Charging Rent

This is the part of the story where the market stops patting you on the back for a “nice year” and starts asking for the rent.

On February 6, 2026, the tape showed AVTECH at a share price of SEK 7.22, giving it a market cap of SEK 407.78m.

That gives:

• P/E Ratio: 23.99x.

• Earnings Yield: ~4.17%

At a ~24x P/E, the market isn’t treating this like a dying cigar butt. It’s pricing AVTECH like a real, recurring software business.

Based on my metrics, the price today is fair, but it’s not “dirt cheap.”

To justify this multiple, the cash reality eventually has to give the earnings reality a big, warm hug. If the earnings stay high but the cash stays “lumpy” for too long, the market will lose patience.

The Atomic Take: At a SEK 408m market cap, the durability test is simple: do the cash-based metrics improve, or are we just living on a pretty narrative?

Subscribe now

The Meltdown List (Risks to Watch)

We DO need to have the risks in mind when we analyze, because every investment has a “self-destruct” button.

Here’s where AVTECH’s is hidden:

1. The “One-Trick Pony” Risk: A huge chunk of revenue still comes from one core product. They’re diversifying, but for now, they are heavily dependent on that single engine.

2. The Waiting Room: Airlines take a long time to buy things. Management admits decision processes are getting longer. A “slow” sales cycle can kill a microcap’s momentum.

3. Currency Gremlins: They hit record revenue in Q4, but the USD/SEK exchange rate still took a bite out of them. If the dollar stays weak, the uphill climb gets steeper.

4. The “Check is in the Mail” Problem: Delayed customer payments dragged down Q4 cash flow. If “delayed” becomes “defaulted,” we have a problem.

5. The Paper Asset: They have MSEK 14.1 in “balanced development expenses.” That’s money spent that hasn’t been “expensed” yet. We need to see that turn into actual sales.

The Atomic Take: The biggest danger isn’t a spectacular crash, but a “slow leak” where the KPIs look great, but the bank account doesn’t grow.

The Atomic Verdict: Trust the Gauges, Not the Vibe

★★★★☆ (4/5)

(IMPORTANT: Star system is not a buy recommendation, only a judgement for the quality of the business based on my parameters. DYOR)

I own shares of AVETCH at the time of writing this. And I intend to hold them for a very long time, unless the fundamentals crack.

And to me, the fundamentals are not showing any cracks to fear as of now:

• Sales Momentum: MSEK 47.4 (FY2025) vs MSEK 34.3 (FY2024). That’s real growth.

• The Bottom Line: MSEK 17.0 in net earnings. The reactor is producing power.

• The War Chest: MSEK 33.9 in cash and a 91.8% equity ratio. This company is built like a tank.

• The Speedometer: MSEK 51.2 ARR. This is the number that dictates the share price.

What would make me buy more (The Upgrades):

• ARR keeps climbing above that MSEK 51.2 baseline through mid-2026.

• Operating cash flow stays healthy and consistent compared to the MSEK 12.1 full-year figure.

• Margins stay resilient (targeting that 34.7% FY average) even as they hire more people.

What makes me walk away (The Downgrades):

• ARR stalls or drops for two quarters in a row.

• Cash flow (CFO) falls way behind reported profits. If the profit is “real,” the cash should follow.

• Hiring keeps going up, but sales stay flat. That means the “investment” was actually just “bloat.”

The Atomic Take: This looks like a compounding machine in the making; as long as the ARR staircase keeps going up and the cash conversion doesn’t start to rot.

What do you think? Does AVTECH’s 91% equity ratio make you feel safe, or does the “capitalized development” make you nervous? Let’s talk about it in the comments.

References

1. Company disclosures and investor materials

2. AVTECH Sweden AB (publ). “Year-End Report January–December 2025.” (Full-year report / financial statements). Dated 2026.

3. AVTECH Sweden AB (publ). “Q4 2025” (Quarterly results presentation/report). Dated 2026.

4. AVTECH Sweden AB (publ). “AVT B – Avtech” (Financial model / compiled financials workbook). Excel file.

5. Market data and third-party screeners (7 Feb 2026)

6. StockAnalysis. “AVT.B (Avtech Sweden AB) — Market Cap / Valuation snapshot” (price, market cap, P/E).

7. TipRanks. “AVTECH Sweden — Statistics” (screen-level cash flow, capex, derived free cash flow).

8. TipRanks. “AVTECH Sweden — Financial Ratios” (screen-level FCF yield and ratio series).

9. Simply Wall St. “AVTECH Sweden — Valuation”.

10. Yahoo Finance. “AVT-B.ST — Quote”

Disclaimer

This Deep Dive is an educational breakdown of a public company based on information available in the materials provided (e.g., annual/quarterly reports, investor presentations, earnings transcripts) and my interpretation of that information. It is designed to be a “bolt-on” intelligence layer to your own due diligence — not a replacement for it.

Independence: I do not accept compensation of any kind from the companies discussed. My research is driven solely by my personal search for high-quality compounders.

Skin in the Game: Unless otherwise stated, assume the author may hold a long position in securities mentioned. Any position creates bias — treat this as commentary, not gospel.

Not Financial Advice: Nothing here is investment advice, a recommendation, or a solicitation. I am not a financial advisor. You are responsible for your own decisions. The stars rating is not a buy recommendation, but meant as a guide to understand the quality of the financial statement of the respective companies.

Error & Update Risk: Financial statements change, companies restate, guidance evolves, and I can be wrong. Verify key figures in the primary filings and consider reading the footnotes before deploying capital.

New to The Atomic Moat? This analysis of Sezzle ($SEZL) is a prime example of how we dissect high-quality compounders. If you want these deep dives sent to your inbox, join 1,000+ other investors below.

Join for Free

Every once in a while, you stumble upon a situation that would make Peter Lynch sit up in his chair and reach for a yellow legal pad. Recently I did a FINS analysis of this company, and I got intrigued to say the least.

The business is growing like a weed, but it’s sitting in a sector the market currently treats like a skunk at a garden party. It’s fallen from the “glory days” highs, the “experts” have moved on to the next shiny object, and suddenly, you look at the price tag and realize you might be looking at a steak dinner for the price of a side of fries.

Sezzle it’s the rare fintech story that can look like a perfectly tuned, tidy little reactor one quarter... and a live-action demonstration of “risk appetite” the next.

I’ve seen this movie before. It’s the kind of business that makes you lean in closer to the screen, wondering if they’ve found the secret sauce or if they’re just playing with the dials.

Let’s find out which one it is.

If You’re In a Rush

• What they actually do: Sezzle is in the business of saying “yes” at the cash register. They offer Buy Now, Pay Later (BNPL) credit, plus subscription goodies like Sezzle Premium and Sezzle Anywhere. They make money when you buy, when you subscribe, and—here’s the rub—from “other sources” (like fees).

• Challenges: The profit recipe is a mix of credit, fees, and subscriptions. That’s a trio that usually gets regulators’ ears perking up. Plus, management just admitted they’re widening the net to find more customers. In plain English: they’re turning up the “growth” dial, and that means the “credit risk” dial is turning right along with it.

• The “Fix”: For this to be a long-term winner, they need to keep their take rate at 11.2% (Q3 2025) while making sure the money they lose on bad loans stays in the neighborhood of 2.5%–2.75% of total sales (2025 management target).

  Disclaimer: I have got high conviction here, and have initiated a large position, as you can see in my recent portfolio update. Have this in mind when you read, as I can be biased.

What is Sezzle, exactly?

Think of Sezzle as a “financial buffer” for your shopping.

You know when you find a great pair of shoes for $100, but you don’t want to drop the full hundred right now because you have other bills? Normally, you’d either have to wait until payday (and hope the shoes don’t sell out) or put them on a credit card and deal with high interest later.

Sezzle lets you take the shoes home today for just $25.

How it works in real life:

• Today: You pay $25 (25%) at the register.

• In 2 weeks: They automatically charge you $25.

• In 4 weeks: Another $25.

• In 6 weeks: The final $25, and you’re done!

The best part for you, the customer? As long as you pay on time, it’s totally interest-free.

Sezzle make their money by charging the store a fee for the privilege of having you as a customer, so you aren’t stuck paying extra just to wait a few weeks. It’s basically a modern way to “layaway” items, except you get the product immediately instead of waiting until the final payment.

Subscribe now

The Setup: Dials, Knobs, and Big Numbers

The “why now” for Sezzle is pretty straightforward. Their Q3 report looked like a clean, well-oiled machine. They showed us scale, they showed us margins, and they showed us cash.

But (and there’s always a “but”) they got there by cranking up two knobs that investors usually love to fear: marketing and underwriting.

It’s like a restaurant that starts seeing huge lines because they started spending a fortune on flyers and decided to let everyone start a tab. It works great for the crowds, but you’ve got to make sure they eventually pay the bill.

The baseline numbers aren’t exactly shy:

• Total Revenue: $116.8 million (Q3 2025).

• Total Sales (GMV): $1,047.3 million (Q3 2025).

That’s right, the company just had its first $1 billion sales quarter. That’s a milestone in anyone’s book.

The “Engagement” side of the house moved, too. As of September 30, 2025, they had 2,971,000 Active Consumers and 784,000 Monthly Online Deep-divers (MODS). Inside that group, they’ve got 0.6 million Active Subscribers and 0.2 million On-Demand users.

How the Business Actually Puts Bread on the Table

If I had to describe Sezzle over a cup of coffee, I’d say it’s a “money-making stack” built on people coming back again and again.

They don’t just take one bite of the apple. They take several small nibbles across three different buckets. Here’s how the $116.8 million in Q3 2025 revenue actually broke down:

1. The “Per-Usage” Bucket (Transaction Income): $59.5 million. This is the classic stuff—fees from merchants and the interchange fees when you swipe.

2. The “Membership” Bucket (Subscription Revenue): $24.3 million. This is the recurring “club” fee for Sezzle Premium. It’s steady and predictable—the kind of revenue I love.

3. The “Extras” Bucket (Income from Other Sources): $32.9 million. This is the spicy stuff: late fees and affiliate marketing.

Hey, wait a minute! Look at these two sub-lines very closely:

• Consumer Fees: $33.5 million.

• Late Payment Fees: $21.0 million.

This is why Sezzle doesn’t care as much if merchant fees get squeezed by competition. They aren’t just relying on the store to pay them; they’re layering on subscriptions and fees.

Management likes to point to their “Take Rate” of 11.2%. That’s basically how much they keep out of every dollar that flows through the system. After paying the direct costs of the transactions, they’re left with $63.3 million—or about 54.2% of their total revenue.

Share

Is There a Moat Here?

I don’t think Sezzle has a “fortress” around the merchants. Instead, they’re betting on habit. They’ve moved their marketing budget toward getting people to subscribe—reaching 568,000 subscribers by the end of the quarter.

The idea is simple: if Sezzle is the first thing you think of when you’re about to buy a pair of shoes, they win. They get to monetize you over and over without having to “buy” your attention again.

• The Fact: Sezzle makes money from everywhere; merchants, subscribers, and late fees.

• The Theory: As long as people keep coming back (high usage) and staying subscribed, the business stays healthy even if the BNPL market gets crowded.

• The Warning: Don’t confuse a “mix shift” with magic. If the revenue is growing because fees are going up rather than more people buying things, that’s a different story.

The “Atomic Take”: Sezzle’s engine is humming, and the diversification is impressive. But it’s only a “Moat” if this growth doesn’t lead to a massive credit hangover down the road. If those “Active Subscribers” keep going up but the revenue doesn’t follow, we’ll know the story has hit a snag.

The “What Went Wrong”

Now, this is the part where most fintech companies start pointing at the “macro economy” or “the Fed” like they’re reporting on the weather.

Not Sezzle. They did something much more refreshing—and a bit more nerve-wracking: they told us they changed their own behavior.

Look, two lines in the ledger tell the whole story here:

First, Marketing went through the roof: $8.775 million in Q3 2025, compared to just $2.726 million a year ago.

Second, their Provision for Credit Losses (the money they set aside for people who won’t pay them back) jumped to $32.177 million, up from $15.402 million last year.

Management is leaning into a classic growth playbook: pay more to get ‘em in the door and widen the funnel so more people qualify. They’re using Sezzle Premium as the main hook, and On-Demand as the backup for folks who aren’t ready for a subscription.

But there’s an uncomfortable truth here: levers work in both directions. If you turn up the heat right as you enter the “seasonal credit-heavy period” (Q4 is historically a bit of a mess for charge-offs), you’re running your reactor hot just as the cooling water starts to warm up.

The Atomic Take: The real risk isn’t some hidden line on the balance sheet; it’s the fact that they’ve got the Underwriting and Marketing dials cranked to “11” at the same time. If credit losses (as a % of sales) keep rising while their profit per sale stays flat, the story changes fast.

The Rebound Catalysts: No Magic, Just Math

We don’t need “AI buzzwords” to see the path forward here. Sezzle has given us a clear roadmap.

1. Credit Normalizes: This is the big one. Management says credit provisions hit 3.1% of GMV in Q3, but they expect that to trend down toward 2.5%–2.75% for the full year 2025. If they hit those numbers while sales stay strong, the profit machine gets a massive boost without needing a miracle.

2. The “Subscription-First” Pivot: They’ve got 0.6 million Active Subscribers now. If they can keep these folks around, it creates a “sticky” base that’s much cheaper to maintain than finding new customers every month.

3. Marketing Payback: Management claims they get their money back on marketing spend in six months or less. That’s a bold claim for an $8.775 million spend. We need to see if this is a “sugar rush” or real, durable growth.

4. Growth Room: They recently bumped their credit facility from $150 million to $225 million. As of Sept 30, they had $32.0 million in unused capacity. They have the “gas” in the tank to keep growing, provided the engine doesn’t knock.

5. Side Projects Stay Small: They spent about $1.3 million on “strategic projects” this quarter. The goal is to keep these from becoming a distraction or a money pit.

Subscribe now

The Financial Quality Rubric (Scorecard: 1-5)

• Profitability Quality (4/5): $26.7 million in Net Income (GAAP). This isn’t “funny money” or weird adjustments. This is real profit.

• Cash Conversion (4/5): They brought in $55.6 million in operating cash over nine months, compared to $90.4 million in net income. Don’t panic—the gap is mostly because they’re using cash to fund new loans ($85.2 million went into notes receivable).

• Reinvestment Intensity (5/5): This business is lean. They only spent $0.6 million on property and equipment. If the credit stays good, this thing is a cash-flow fountain.

• Balance Sheet Simplicity (4/5): It’s mostly cash ($134.7 million) and receivables ($184.1 million). It’s easy to see what’s under the hood.

• Risk Discipline (3/5): They intentionally lowered the bar to grow. That’s either a genius move to grab market share or the start of a slippery slope. We’re watching this closely.

• Capital Allocation (4/5): They bought back $34.6 million worth of shares this year. I love seeing a company return cash to owners, as long as they don’t leave themselves short on liquidity.

The Audit: The biggest trick here is the “weak” cash flow. It’s not weak; it’s just being used to grow the loan book. If you think the loans are good, the cash flow is fine. If you think the loans are shaky, the cash flow is a red flag.

The Atomic Take: Sezzle is a high-quality machine right now, but you have to treat credit and growth as “first-class” risks. If operating cash drops while profits look good, it means the quality of those profits is starting to rot.

The Statements: Reading the Geiger Counter

Folks, if you want to know if a nuclear reactor is safe, you don’t just look at how much power it’s putting out; you check the radiation levels.

In a lending business, the Balance Sheet is your Geiger counter.

As of Sept 30, 2025, the “Big Five” numbers look like this:

1. Cash (the oxygen): $134.7 million.

2. Notes Receivable (the loans): $184.1 million.

3. Allowance for Credit Losses (the safety net): $33.7 million.

4. Line of Credit (the debt): $117.3 million.

5. Equity (the owner’s stake): $155.3 million.

One sneaky thing to watch is something called DSIP (Delayed Settlement). It’s basically merchants saying, “Keep our money for a bit, pay us interest, and we’ll use it as a float.”

Sezzle has $43.0 million sitting in that bucket. It’s not “toxic,” but it’s not a normal payable either. It’s helpful funding, but in a crisis, merchant behavior can change fast.

The Atomic Take: The balance sheet is simple enough to stress-test on a napkin. But watch that unused borrowing capacity ($32.0 million). If that starts to shrink while loans grow, the car is running out of gas right as it’s trying to speed up.

Cash Flow: The Turbine

In most businesses, “Cash is King.” In Sezzle’s world, Cash is a Turbine. They made $90.4 million in Net Income over nine months, but “only” saw $55.6 million in Operating Cash.

Why the gap? Because they used $85.2 million to fund new loans.

The real “Free Cash” (Cash from Ops minus the tiny $2.1 million they spent on equipment and software) is about $53.5 million.

The Audit: The question isn’t “Where did the cash go?” It’s “Are these new loans high-quality?” If the loans are good, the cash flow is a coiled spring. If the loans are bad, the turbine is about to wobble.

The Income Statement: The Reactor Room

The reactor was humming in Q3:

• Revenue: $116.8 million.

• Operating Income: $35.6 million.

• Net Income: $26.7 million.

• Adjusted EBITDA Margin: A fat 33.9%.

But here’s the control rod: the Provision for Credit Losses was $32.177 million. Because they changed their underwriting to find more customers, they had to set aside more “just in case” money.

Unlike a software company that can just cut its travel budget to save money, Sezzle lives and dies by the “spread.” They have to make sure the 11.2% Take Rate stays comfortably above the cost of the loans going bad.

Share

Here is the “Capital Efficiency” section, rewritten to match the Simple Truths style and tone.

The Radiation Test (Capital Efficiency)

If you want to understand why Sezzle looks “cheap” on a screen but feels dangerous in your gut, you have to look at how hot the reactor is running. The business is generating a massive amount of profit relative to the tiny amount of capital actually sitting in the bank.

That is great... until it isn’t.

• The Metric: Return on Equity (ROE).

• The Number: ~74% (First 9 Months of 2025).

• The Translation: For every $1 of equity the owners have in the company, Sezzle generated $0.74 of profit in just nine months. If you annualized that, it’s nearly a 100% return.

• The Reality Check: Net Income was $90.4M while the average equity was only ~$121.6M.

The “Audit” (Why This is Scary): The returns are sky-high not just because the business is good, but because the equity base is small.

Think of it like a sports car that has been stripped of its roll cage and bumpers to save weight. It goes incredibly fast (high efficiency), but if it hits a single bump (a spike in bad loans), it doesn’t just dent; it shatters.

The Atomic Take: Don’t admire the glow without checking the containment vessel.

• The Good: They are printing money with very little capital required (ROCE is ~52%).

• The Bad: This only works if the “credit dial” stays disciplined.

• The Falsifier: Watch the Return on Capital (ROCE). If it drops below 30% in the next two quarters, the “capital-light profit machine” is broken, and the stock deserves to stay cheap.

Valuation

Alright, let’s address the elephant in the room. In July 2025, this stock hit $182. Today, as of writing this deep dive, it’s trading at $69.

That is a ~62.1% drawdown. Yikes.

That is not a “dip.” Not even the “dip of the dip”. That is a full-blown haircut. When a stock drops like that, you don’t just ask, “Is it cheap now?” You have to ask the harder question: “What did the market believe at $182 that it stopped believing at $69?”

The Math of the Fall

Let’s assume the share count hasn’t changed much (Sezzle told us they had 34,153,102 shares outstanding as of Nov 3, 2025). At $69, that gives us a market cap of roughly $2.4 billion.

Now, let’s translate that price drop into what actually changed: the valuation multiple.

Scenario A: The $69 Reality (Today)

If we look at the business right now, here is what you are paying for:

1. The Earnings Lens (GAAP):

• Net Income (9 months): $90.4 million.

• Run-Rate (Annualized): About $120.5 million (just taking the 9 months and adding a third).

• The Price Tag: At a $2.4 billion market cap, you’re paying a P/E of ~19.9x.

• The Yield: That’s an earnings yield of about 5.0%.

If you prefer per-share math:

• Run-Rate EPS: ~$3.53.

• P/E: Still ~19.6x.

2. The Cash Flow Lens (The “Truth Serum”): We can build a disciplined proxy for Free Cash Flow (FCF), not to take a victory lap, but to see how much cash actually hits the till.

• Cash from Operations: $55.6 million (9M 2025).

• Capex (Property + Intangibles): A tiny $2.1 million ($0.6m + $1.5m).

• FCF Proxy: $53.5 million.

• Run-Rate FCF: ~$71.3 million.

• The Price Tag: A P/FCF of ~33.6x.

• The Yield: An FCF yield of about 3.0%.

Scenario B: The Ghost of $182 (What Were We Thinking?)

Now, let’s hop in the time machine back to July. If the share count was roughly the same, that $182 price tag implied a market cap of about $6.33 billion.

Run those same earnings through that price, and here is what the market was paying:

• Implied P/E: ~52.5x.

• Implied Earnings Yield: ~1.9%.

• Implied P/FCF: ~88.7x.

• Implied FCF Yield: ~1.1%.

The Story in One Sentence: The market stopped treating Sezzle like a “Premium Tech Compounder” (where you pay 50x earnings for growth) and started treating it like a “Credit-Sensitive Cash Converter” (where you demand a discount because lending is risky).

So, Is It Cheap?

The market is basically saying: “I’ll buy it at 19x earnings, but I need you to prove you aren’t going to blow up.”

It all comes down to a simple spread trade:

1. The Take Rate: 11.2% (Q3 2025). This is the money coming in.

2. The Credit Losses: 3.1% (Q3 actual), trending toward 2.5%–2.75% (2025 target). This is the money leaking out.

If the Take Rate holds and the Credit Losses drop to that 2.5% range, then 19.9x earnings could be a bargain, in my view.

But if credit stays sticky or the take rate drops, the valuation compression wasn’t just a bad mood—it was math.

Atomic Take: The move from $182 to $69 was a massive “Multiple Reset.” We went from a ~2% earnings yield to a ~5% earnings yield. The market is now demanding proof that the profit margin is real and not just a temporary luck of the draw.

Falsifier: If the provision for credit losses doesn’t trend down to 2.5%–2.75% while the take rate stays at 11.2%, then the “Cheap P/E” narrative falls apart. The earnings aren’t durable, and the stock isn’t cheap—it’s just risky.

Subscribe now

The “Meltdown” List (Risks)

1. The Seasonal Spike: Q4 is “Danger Zone” time for credit. They’re widening the net right as the shark-infested waters get choppy.

2. The “Fee” Optics: They made $33.5 million from consumer fees and $21.0 million from late fees. Regulators love to look at those numbers. It’s a headline risk.

3. The “Optional” Parasites: They’re spending $1.3 million on “strategic projects” like exploring a bank charter (ILC) and dealing with Shopify legal matters. It’s small now, but distractions can grow.

The Atomic Verdict: High Conviction

★★★★☆ (4/5)

(Star system is not a buy recommendation, only a judgement for the quality of the business based on my parameters)

Sezzle has earned the “Real Business” label. They make actual GAAP profit, and the balance sheet isn’t a house of cards. But it’s “Overhang-Heavy.” The next act depends on discipline, not just momentum.

Watch for the Upgrade: If credit costs drop toward 2.5% and subscribers keep growing past 0.6 million, I’m pounding myself on the chest.

Watch for the Downgrade: If cash from operations starts to look puny compared to net income, or if the “borrowing gas tank” hits empty, we’ve got a problem.

If you are pure numbers guy, and dont want anything else but the data,you can read the FINS analysis of Sezzle here.

Are you a believer in the “Subscription BNPL” model, or do you think the reliance on consumer fees is a regulatory ticking time bomb? Let’s hash it out in the comments below!

Sources

• Sezzle Inc. — Form 10-Q (PDF): “10-Q_-Sezzle_INC-_11-06-2025.pdf”

• Sezzle Inc. — Q3 2025 Earnings Call Transcript (PDF): “Q3_2025_Earnings_Call_Transcript.pdf”

• Sezzle Inc. — Investor Presentation (3Q25) (PDF): “Sezzle_Investor_Presentation_3Q25.pdf”

• Sezzle Inc. — IR Factsheet (3Q25) (PDF): “3Q25_IR_Factsheet_vFINAL.pdf”

• Atomic Moat Research — FINS Financial Audit (PDF): “Sezzle ($SEZL) Stock Analysis_ FINS Financial Audit.pdf”

• Sezzle Inc. — Financial Model / Dataset (XLSX): “SEZL-Sezzle Inc.xlsx”

Disclaimer

This Deep Dive is an educational breakdown of a public company based on information available in the materials provided (e.g., annual/quarterly reports, investor presentations, earnings transcripts) and my interpretation of that information. It is designed to be a “bolt-on” intelligence layer to your own due diligence — not a replacement for it.

Independence: I do not accept compensation of any kind from the companies discussed. My research is driven solely by my personal search for high-quality compounders.

Skin in the Game: Unless otherwise stated, assume the author may hold a long position in securities mentioned. Any position creates bias — treat this as commentary, not gospel.

Not Financial Advice: Nothing here is investment advice, a recommendation, or a solicitation. I am not a financial advisor. You are responsible for your own decisions. The stars rating is not a buy recommendation, but meant as a guide to understand the quality of the financial statement of the respective companies.

Error & Update Risk: Financial statements change, companies restate, guidance evolves, and I can be wrong. Verify key figures in the primary filings and consider reading the footnotes before deploying capital.

New to The Atomic Moat? This analysis of Adobe ($ADBE) is a prime example of how we dissect high-quality compounders. If you want these deep dives sent to your inbox, join 1,000+ other investors below.

Join for Free

I have spent more Friday nights than I care to admit staring at a Premiere Pro timeline, fighting with keyframes and cursing at render times. I know the muscle memory of Photoshop shortcuts better than I know my own phone number. To me, these aren’t just “apps.” They are the language I speak. You don’t just “switch” languages because a new AI tool came out on a Tuesday.

But I’ve also seen the other side of the beast.

In a previous life, I worked at a European bank. We used Adobe Analytics (the old Omniture engine), not Google Analytics. Why? Because we had to.

In banking, you operate under a strict license. We couldn’t just pipe sensitive customer data to Google’s public cloud and hope for the best with GDPR. We needed a tool that was audit-proof, compliant, and guaranteed data sovereignty. Adobe was the only adult in the room.

So when I hear the Wall Street narrative that “Adobe is dead” or “AI is going to kill the moat,” I don’t just look at the spreadsheet. I look at my own scar tissue.

I see a Creative Cloud that is sticky by design, and an Experience Cloud that is entrenched by compliance.

The market thinks Adobe is a fragile painting. I know it’s a concrete bunker.

Let’s dig in.

If You’re In a Rush

• What they actually do: Adobe sells subscriptions. They are the toll booth for the creative economy, documents, and enterprise experiences. We are talking about a massive $22.904B subscription revenue operation. That is not a startup; that is a sovereign state of software.

• Why it’s hated: Wall Street is treating generative AI like a pricing grenade, and they think Adobe is running around the room waving the pin. The company recently disclosed that “Total new AI-influenced ARR now exceeds one-third of our overall business”. To an optimist, that’s adaptation. To a pessimist, that’s a sign they are running to stand still.

• What fixes it: You fix the narrative by showing the money. AI monetization has to show up in the boring, beautiful recurring engine. Management has given us the scoreboard to watch: 10.2% YoY Total Adobe ending ARR growth. If they hit that, the story changes.

Atomic Take: Here is the plain truth: If the ARR engine holds up, the “AI hate” stops being about economics and starts looking like just bad sentiment.

Falsifier: If Total Adobe ending ARR growth prints below 10.2% YoY, the bears were right.

Subscribe now

The Setup

Folks, the market’s argument right now isn’t “is Adobe a good business today?”

Anyone with eyes can see it is.

The argument is, “is Adobe’s business model about to get re-priced by AI?”

Look at the hard evidence:

• They delivered $23.769B total revenue and brought home $7.130B net income. That is real profit.

• The cash machine is humming. They generated $10.03B operating cash flow and bought back about 30.8M shares. That’s shareholder yield, plain and simple.

• They exited the year with $25.20B Total Adobe ARR and $19.20B Digital Media ARR

• The backlog is thick. We are looking at $22.52B RPO

Management isn’t hiding under a rock, either. They are guiding for FY2026 total revenue of $25.90B–$26.10B and GAAP diluted EPS of $17.90–$18.10. They are explicitly telling us to watch the customer-group subscription revenue plus total company ending ARR growth. They want us to look at the recurring revenue.

The stock is being priced as if “creation becomes free” happens faster than Adobe can figure out how to package AI into their paid tiers.

But we need to watch three things like a hawk:

1. Total Adobe ending ARR growth vs 10.2% YoY (FY2026 target).

2. RPO growth off that massive $22.52B base.

3. The GAAP vs non-GAAP bridge. When the narrative gets shaky and the adjustments get heavy, that’s when investors get jumpy.

Atomic Take: The company is still throwing off recurring, high-margin economics. The bear case requires you to be a fortune teller (forward-looking), because the historian (backward-looking) sees a great business.

Falsifier: If RPO stops growing year-over-year (exiting-quarter YoY %, company-reported), the engine is stalling.

How the Business Actually Makes Money

Let’s start with the boring truth, because that’s usually where the money is. Adobe is a subscription company that masquerades as a cool creative culture icon.

On the P&L, subscription is the whole movie.

I mean it. $22.904B subscription revenue out of $23.769B total revenue. If you’re looking for license fees or one-time sales, you’re in the wrong decade.

Segment-wise, the center of gravity is still Digital Media. That’s the Photoshop and Premiere behemoth. It pulled in $17.65B segment revenue and sits on $19.20B ending ARR.

Then you have Digital Experience (the enterprise marketing side) at $5.86B segment revenue with $5.41B subscription revenue.

Now, here is what has changed (quietly, but it’s important): Adobe is reframing who it sells to. They now highlight two customer groups:

1. Business Professionals & Consumers subscription revenue: $6.50B.

2. Creative & Marketing Professionals subscription revenue: $16.30B.

Why does this matter? Because the AI panic attacks the low end first.

If generative tools make “good enough content” cheaper, the risk isn’t that the Hollywood editor cancels Premiere Pro tomorrow. The risk is that the next million casual creators never pay what they used to pay.

Adobe’s counter is obvious in the materials: put AI assistants inside Acrobat, Reader, and Express for the broad audience, while Creative Cloud and Firefly deepen the power workflows for the pros.

That kind of switching cost doesn’t trend on X (formerly Twitter), but it tends to show up in the ARR and RPO lines.

Management also discloses that “over 75% of Digital Media net new ARR” was driven by subscription growth + cross-sell/up-sell.

That is organic demand, folks.

Atomic Take: Adobe’s moat is workflow gravity. Subscriptions are the toll booth, and AI is just the new express lane they’re trying to charge for.

Falsifier: If Digital Media ending ARR growth falls materially below 11.5% YoY for multiple quarters, the gravity is weakening.

Share

What Went Wrong

Nothing “broke” in the financial statements. Read them again. They’re fine.

What broke was the market’s willingness to pay a premium multiple for durability when that durability is getting rhetorically challenged by AI every single day.

Adobe is explicitly integrating AI across products and measuring “AI-influenced” ARR. They state that “Total new AI-influenced ARR now exceeds one-third of our overall business”.

That sentence right there? That is the hate accelerant. It invites two competing interpretations:

• Interpretation A (The Bull): “AI is a new paid layer on top of a sticky subscription base. It’s an upsell.”

• Interpretation B (The Bear): “AI reduces differentiation, so Adobe is forced to bundle AI just to stand still. They are running on a treadmill.”

The materials show Adobe positioning itself as a platform that blends its own models with partner models across apps. This is where the “AI hate” splits into things we can debunk and things we can’t.

Hate Narrative #1: “AI makes creation free, so Adobe’s pricing power dies.”

FACT: Adobe is describing monetization mechanics (Generative Credits) and a path to higher-value offerings or add-ons as usage rises (prepared remarks).

HYPOTHESIS: If Generative Credits become a meaningful “value usage” meter, monetization becomes measurable. It stops being about “trust the demo” and starts being about the meter.

MONITOR: Management’s disclosed usage/monetization commentary. Does it translate into sustained ending ARR growth (total + Digital Media)? Note: This isn’t fully debunkable yet. AI can compress differentiation at the low end. The only honest way to fight that is to show the paid tiers still compounding.

Hate Narrative #2: “AI will blow up costs and crush margins.”

FACT: Adobe guides to non-GAAP operating margin assumptions around ~45% for FY2026 targets.

HYPOTHESIS: If AI compute costs were quietly eating the business alive, we’d expect pressure in operating profitability. We aren’t seeing it in the guide.

MONITOR: GAAP operating margin trajectory and operating cash flow vs net income (GAAP). Note: We don’t get a clean “AI compute” line item. This stays a watch item, not a victory lap.

Hate Narrative #3: “Adobe is behind in agentic workflows and new surfaces.”

FACT: The company points to integrations and “conversational platforms” surfaces (prepared remarks + deck). They disclose early traction such as “over 25,000 businesses” purchasing Express or Studio for the first time in Q4

Atomic Take: The hate is mostly a forward multiple debate disguised as a product debate. The only referee that counts is ARR.

Falsifier: If customer-group subscription revenue growth decelerates meaningfully while “AI-influenced” messaging intensifies, you have a problem. Specifically, watch Business Professionals & Consumers and Creative & Marketing Professionals subscription revenue (company-reported).

Subscribe now

Rebound Catalysts

This is one of the nicer setups in the land of the mega-caps. Why? Because management isn’t giving us vague poetry; they are giving us a scoreboard that we can actually check.

Here is the checklist. If these four things happen, the stock works.

Catalyst 1: Hit the “Magic Number” (ARR Growth)

Management targets 10.2% YoY Total Adobe ending ARR growth. This is the whole ballgame. If they hit that number while continuing to scream about “AI-influenced” offerings, the simplest bear framing—“AI breaks the subscription model”—gets very hard to defend.

Catalyst 2: Watch the Crystal Ball (RPO)

RPO exited FY2025 at $22.52B, growing 13% YoY. RPO (Remaining Performance Obligations) isn’t “revenue already earned,” but it is the backlog. It’s the proxy that tends to crack before revenue does if demand is actually drying up. Right now, the crystal ball looks clear.

Catalyst 3: AI Needs to Be Boring (Document Adoption)

We all look at the cool AI images. But the money is in the boring stuff.

Customer reception for Acrobat Studio is real: “nearly 50% of Acrobat commercial ETLA’s renewed in Q4 already upgrading.

Over 25,000 businesses purchased Express or Studio for the first time in Q4 alone. If AI monetization is real, it should show up in business-pro workflows first. Why? Because businesses pay for productivity. Consumers pay for vibes. We want productivity.

Catalyst 4: The Cash Shield

FY2025 operating cash flow was $10.03B. The market can stay irrational longer than you can stay solvent, but Adobe has $10B a year to buy time. That is a very expensive shield.

Atomic Take: Adobe doesn’t need to “win AI” in a philosophical sense. It needs to turn AI into a measurable subscription expansion story, quarter after quarter.

Falsifier: If FY2026 ending ARR growth misses 10.2% YoY (FY2026 actual vs target, %), even if they hit their EPS targets, run. An engine mismatch is the tell.

Financial Quality Rubric

Let’s grade the report card.

Revenue Durability (4/5)

Subscription concentration is extreme in the best possible way—$22.904B subscription revenue. They disclose growth across customer groups, so we can see where the money comes from.

Profitability Under Reinvestment (4/5)

They are spending money to make money, but they are still making a lot of money. GAAP operating income was $8.706B.

If you like the adjusted numbers, non-GAAP operating income was $10.986B.

Cash Generation Quality (5/5)

This is elite. $10.03B operating cash flow versus $7.130B GAAP net income.

Cash is doing the heavy lifting here.

Balance Sheet Resilience (4/5)

They can weather a storm. Cash & equivalents were $5.431B and short-term investments were $1.164B.

Compare that to $6.210B long-term debt (as of Nov. 28, 2025) It’s balanced.

Capital Allocation Discipline (4/5)

They are eating their own cooking. ~30.8M shares repurchased with $5.90B remaining on authorization.

Disclosure/Metric Integrity (3/5)

Here is the smudge on the report card. Management notes it will “periodically update” offerings included in Total AI-Influenced ARR. That is “definition drift” risk.

The Audit: Non-GAAP is useful, but it’s also where “AI optimism” and “adjustment optimism” hold hands. Adobe’s reconciliation shows stock-based and deferred comp is a material add-back inside operating income and EPS bridges. That isn’t a sin—everyone in tech does it—but don’t treat it as imaginary money. It dilutes you.

Atomic Take: The financial quality is real. The biggest fragility is narrative credibility around AI monetization definitions.

Falsifier: If GAAP operating cash flow falls below GAAP net income for a full fiscal year (FY basis, USD, GAAP), the cash conversion is breaking.

The Statements

Let’s open the hood and look at the engine block.

Balance Sheet (The Geiger Test)

Here is the adult version of “is this balance sheet safe?” We check for liquidity, leverage, and hidden obligations.

• Liquidity: Cash and cash equivalents were $5.431B plus short-term investments of $1.164B.

• Leverage: Long-term debt was $6.210B. Current debt was $0.

• The Future Money: Deferred revenue (current) was $6.905B and deferred revenue (long-term) was $0.125B.

This isn’t a “distress” setup. It’s a “durable subscription model with manageable leverage” setup. The real Geiger-test question isn’t solvency; it’s whether the subscription backlog signals (deferred revenue + RPO) keep pointing up.

Atomic Take: The balance sheet looks like a company built to absorb narrative volatility without being forced into financial contortions.

Falsifier: If Deferred revenue declines materially year-over-year while revenue still grows (FY basis, USD), you have a pull-forward or billing quality risk.

Cash Flow (The Turbine)

If the stock is hated, the turbine is the rebuttal: show me the cash.

• Operating cash flow was $10.03B.

• GAAP net income was $7.130B

That spread is meaningful. It means the business is generating more cash than accounting profits.

We have a clean look at Q4:

Q4 net cash provided by operating activities was $3.160B.

What drove it? Changes in deferred revenue of $0.496B and changes in other operating assets and liabilities of $0.244B.

Capex Hygiene: They run light. Purchases of property and equipment were $(34)M.

Capital Return: They are returning cash aggressively. Repurchases of common stock were $(2.474)B.

The Audit: We can keep watching whether operating cash stays strong as AI features scale, because if AI were a massive cost sink, cash is where it would leak first.

Atomic Take: The turbine is still ripping. Just don’t confuse “cash today” with “cash forever” without watching timing and reinvestment.

Falsifier: If Operating cash flow drops materially below $10.03B while repurchases remain large, they are stressing the coverage.

Subscribe now

Income Statement (The Reactor)

If AI were already commoditizing Adobe, you’d expect to see it as margin decay. FY2025 doesn’t read like that.

• Top Line: Total revenue was $23.769B

• Gross Profit: Gross profit was $21.218B

• The Bottom Line: Operating income was $8.706B

• The Future: R&D expense was $4.294B

The reactor story is simple: high gross profit, substantial reinvestment, and still strong operating income. That’s the profile of a company funding its own transition.

Crucial Note on Trends: Don’t miss the FY2024 one-off buried in plain sight. Acquisition termination fee was $1.000B (FY2024, USD, GAAP operating expense line).

That was the Figma breakup fee. It distorts the Year-over-Year comparison, so keep that in mind when looking at growth rates.

Atomic Take: The income statement says “premium software economics still intact,” not “AI is eating our lunch.”

Falsifier: If Revenue grows while GAAP operating income declines year-over-year for a full fiscal year (FY basis, USD, GAAP), margin leakage is becoming visible.

Valuation

We are going to pin the stock to the wall.

The Price Tag:

• ADBE last trade price: $296.12 (Jan 17, 2026, USD).

• Market Cap: $123.96B (as of Jan 16, 2026, USD).

• Enterprise Value: $124.01B (as of Jan 16, 2026, USD).

The Engine (What you get for that price):

• GAAP Diluted EPS: $16.70.

• GAAP Operating Cash Flows: $10.03B.

• The Guide: FY2026 GAAP EPS guidance of $17.90 to $18.10.

• The Growth Target: 10.2% YoY Total Adobe ending ARR growth (FY2026 target).

Now, let’s run the numbers. This is what “Hated Quality” actually looks like:

• Trailing P/E (GAAP): ~17.7x (Price $296.12 ÷ GAAP EPS $16.70)

• Forward P/E (GAAP): ~16.5x (Price $296.12 ÷ Forward Midpoint $18.00)

• Operating Cash Flow Yield: ~8.1% ($10.03B OCF ÷ $123.96B Market Cap)

• EV / OCF: ~12.4x ($124.01B EV ÷ $10.03B OCF)

Do you see what I see?

That is not a “tech bubble” valuation. That is a utility company valuation. The market is offering you a high-margin, subscription-heavy cash engine for a mid-teens multiple and a high-single-digit (8.1%) cash flow yield.

The skepticism isn’t subtle. The narrative is screaming that AI-native tools are going to destroy seat-based pricing. But the numbers are whispering that the business is on sale.

Atomic Take: At ~17.7x trailing GAAP earnings and an ~8.1% operating cash flow yield, Adobe is being valued like a durable cash machine with a PR problem; not like a broken business.

Falsifier: If FY2026 ending ARR growth misses 10.2% YoY, then the “durability” argument dies. At these multiples, durability is the only case.

Risks (The Meltdown List)

The valuation is forgiving, but it could be forgiving for a reason. Here is why the penalty box exists.

Risk #1: The Seat-Based Pricing Trap

FACT: The market thinks AI pushes software toward usage-based pricing (pay for what you generate) and kills the “seat” model (pay per employee).

MONITOR: Watch Total Adobe ending ARR growth and customer group subscription revenue trends. If seat growth stalls, the model is shifting under our feet.

Risk #2: The “Hidden Tax” of Compute

FACT: Adobe generated $10.03B operating cash flows while investing in AI.

HYPOTHESIS: “AI everywhere” increases the cost-to-serve. If I pay $20/month but generate $30 worth of AI images, Adobe loses money on me.

MONITOR: OCF stability vs FY2025’s $10.03B baseline. If margins dip while revenue grows, we have a leak.

Risk #3: The Moving Goalposts

FACT: Adobe explicitly states it will “periodically update” what’s included in Total AI-Influenced ARR.

MONITOR: Any redefinition that breaks comparability. If they change the definition, assume they missed the number.

Risk #4: The Low-End Squeeze

FACT: The narrative is full of fears about AI-native platforms.

HYPOTHESIS: Even if Adobe keeps the Hollywood pros, the “good enough” AI tools steal the casuals. This caps ARPU (Average Revenue Per User).

MONITOR: Compare Business Professionals & Consumers subscription revenue (FY2025: $6.50B) growth vs Creative & Marketing Professionals (FY2025: $16.30B). If the consumer side flatlines, the funnel is broken.

Risk #5: The Value Trap (”Cheap stays Cheap”)

FACT: The stock is at $296, ~$124B market cap. The market is already underwriting “slower, riskier duration.”

MONITOR: RPO ($22.52B exiting FY2025). This is your forward-looking radar. If the backlog shrinks, the “value” is a trap.

Atomic Take: The risk isn’t bankruptcy. The risk is that AI changes the value capture model faster than Adobe can figure out how to charge for it. The multiple won’t re-rate until ARR proves the capture is intact.

Falsifier: If Total Adobe ending ARR growth stays at/above 10.2% YoY while OCF remains at/above $10.03B, the “AI kills economics” fear is proven wrong.

The Atomic Verdict

Status: ★★★★☆ (3.9/5)

This isn’t a debate about quality anymore. The statements ended that debate.

• GAAP EPS: $16.70.

• Operating Cash Flows: $10.03B.

• Total Adobe Ending ARR: $25.20B.

The new element is the price. The market has stopped paying “love multiples.”

• Stock Price: $296.12 (Jan 17, 2026, USD).

• Trailing P/E: ~17.7x.

• OCF Yield: ~8.1%.

So, the verdict tightens into a single underwriting question:

Can Adobe keep compounding ARR in a world where the market thinks AI makes creation free?

The Scorecard

Upgrade Triggers (The Green Lights):

1. FY2026 Total Adobe ending ARR growth meets/exceeds 10.2% YoY.

2. RPO ($22.52B exiting FY2025) continues to grow without sharp deceleration.

3. Operating cash flow stays near or above $10.03B. If cash stays high while AI scales, it means they are monetizing, not subsidizing.

Downgrade Triggers (The Red Lights):

1. FY2026 ending ARR growth falls below 10.2% YoY.

2. GAAP EPS meets guidance but RPO softens. That smells like financial engineering.

3. OCF meaningfully declines from $10.03B. If the cash drops while the AI hype rises, the economics are leaking.

Atomic Confidence Builder:

I want to see two to three quarters where Adobe hits the FY2026 “scoreboard” (ARR + RPO) and the bear narrative starts losing oxygen—not because the pundits get nicer, but because the KPIs refuse to cooperate with them.

Atomic Take: The stock is being treated like a credibility case, not a quality case. At today’s multiples, recovering that credibility is the lever that rips the stock higher.

Falsifier: Total Adobe ending ARR growth misses 10.2% YoY (FY2026 actual vs target, %, company metric).

Comment Below: We’re looking at an 8.1% cash flow yield on a software monopoly. Are you buying the “AI Death” narrative, or are you buying the cash? Let me know.

Reference notes

• Adobe Reports Record Q4 and FY2025 Revenue (press release / earnings release; includes FY2025 GAAP EPS $16.70, FY2025 operating cash flow $10.03B, Total Adobe ending ARR $25.20B, FY2026 targets incl. Total revenue $25.90B–$26.10B and Total Adobe ending ARR growth 10.2% YoY) (Dec. 10, 2025).

• Adobe Q4 FY2025 Earnings Script & Slides / Prepared Remarks deck (includes “Total new AI-influenced ARR now exceeds one-third of our overall business,” RPO $22.52B, and FY2025 highlights) (Dec. 10, 2025).

• Adobe AI strategy / Investor presentation (Adobe AI Platform stack; partner model ecosystem; AI Assistants / Firefly / GenStudio / AEP + Apps / AEM + Agentic Web framing) (2025).

• Adobe Q4 FY2025 earnings call transcript (CEO/CFO prepared remarks; AI model strategy; Creative Cloud / Acrobat / Express positioning) (Dec. 10, 2025).

Market data (web)

• Adobe (ADBE) last trade price used for valuation math: $296.12, Jan. 18, 2026.

• Yahoo Finance “Key Statistics” page for Adobe valuation measures (Market Cap, Enterprise Value, P/E, etc.) (Jan. 18, 2026).

• CompaniesMarketCap page for Adobe market capitalization (January 2026 snapshot) (Jan. 18, 2026).

AI “hate” narrative / investor commentary (web, recent)

• MarketWatch coverage of software stocks sliding amid AI fears (includes Adobe move and framing of investor concerns) (published Jan. 2026; Jan. 18, 2026).

• Investopedia coverage summarizing analyst concerns about AI disruption and seat-based vs usage-based pricing dynamics (published Jan. 2026; Jan. 18, 2026).

Primary-source duplicates (web versions of the same company documents)

• Adobe-hosted Q4 FY2025 earnings call transcript PDF (Adobe Investor Relations) (Dec. 10, 2025; Jan. 18, 2026).

• Adobe-hosted Q4 FY2025 “Earnings Script & Slides” PDF (Adobe Investor Relations) (Dec. 10, 2025; Jan. 18, 2026).

Disclaimer

This Deep Dive is an educational breakdown of a public company based on information available in the materials provided (e.g., annual/quarterly reports, investor presentations, earnings transcripts) and my interpretation of that information. It is designed to be a “bolt-on” intelligence layer to your own due diligence — not a replacement for it.

Independence: I do not accept compensation of any kind from the companies discussed. My research is driven solely by my personal search for high-quality compounders.

Skin in the Game: Unless otherwise stated, assume the author may hold a long position in securities mentioned. Any position creates bias — treat this as commentary, not gospel.

Not Financial Advice: Nothing here is investment advice, a recommendation, or a solicitation. I am not a financial advisor. You are responsible for your own decisions. The stars rating is not a buy recommendation, but meant as a guide to understand the quality of the financial statement of the respective companies.

Error & Update Risk: Financial statements change, companies restate, guidance evolves, and I can be wrong. Verify key figures in the primary filings and consider reading the footnotes before deploying capital.

Note: I built this analysis using the latest Q3 2025 numbers, but I leaned heavily on the FY2024 Annual Report as well, because that is where the company explicitly details the risks we need to have in mind (cyber threats, regulatory friction, and customer concentration). So I used it to ensure we aren't flying blind.

IF YOU’RE IN A RUSH

The Elevator Pitch:

Evolution is the production studio for online casinos. They hire the card dealers, run the video cameras, and stream the live footage to gambling websites. They provide the live dealers and the random number generator (RNG) games. They sell this to gaming operators, mostly on a commission model.

When the operator makes a profit on an Evolution table, Evolution takes a cut. They also charge fees for setting up fancy, dedicated environments.

Why the Stock is in the Penalty Box:

The latest scorecard was soft. In Q3 2025, net revenues came in at €507.1m (IFRS), down 2.4% YoY. Management is being brutally honest about the mess: “Asia remains volatile,” and they are dealing with real-world operational disruption and cyber attacks.

The Fix:

Three things need to happen:

1. Asia needs to stabilize.

2. The product machine needs to keep firing (management outlook is 110+ new games in 2025).

3. The costs of regulatory plumbing need to stop shaving the clean edges off their margins.

Atomic Position: Overhang-Heavy. The economics are still elite, but the operational and regulatory fog is still in the room. We need the windshield to clear up.

THE SETUP

Folks, let’s be clear: Evolution’s 2024 and 2025 was not a collapse. It was a warning label stuck on a very profitable machine.

And make no mistake, this machine prints cash.

In FY2024, total operating revenues were €2,214.1m (IFRS) and operating profit (EBIT) hit €1,419.6m. That is a 64.1% EBIT margin.

In the world of physical businesses, that number is usually a typo. Here, it’s reality.

The business is also steadily becoming “more regulated-market shaped.” Regulated markets made up 40% of net revenues in FY2024, which is stable compared to 2023.

Stop & Think: Why does this matter? Regulation is a double-edged sword. On one side, it’s a moat—you need licenses, compliance, and trust to play. On the other side, it’s a tax. It demands ring-fencing, audits, and friction.

Management knows the narrative is getting messy, so they are trying to out-ship the bad news. They are calling 2024–2025 “product leap years.”

They launched 109 new games in 2024, and the outlook calls for 110+ new games in 2025.

Here is the argument on Wall Street:

• The Market View: “Sure, margins are huge. But growth is wobbling and the world is tightening the screws on gambling. The party is getting raided.”

• The Bull View: “This is the best content factory and distribution stack in online casino history. The disruptions are temporary. The capital returns will keep compounding value per share.”

• The Bear View: “Regulators, cyber attacks, and operator concentration mean Evolution is being forced into a lower-power mode… permanently.”

This deep dive is about answering one single question: Are the recent disruptions just turbulence around a compounding machine, or is the machine being forced into a structurally slower, more expensive way of life?

Atomic Take: Evolution can be both “still exceptional” and “not currently clean enough for High Conviction.” Our job is to separate temporary noise from permanent friction.

Falsifier: If net revenues stay negative YoY for multiple quarters while management maintains the 66–68% margin framework, the “temporary turbulence” story stops fitting the data.

HOW THE BUSINESS ACTUALLY MAKES MONEY

Let’s strip away the corporate speak. Evolution sells online casino infrastructure and content to operators. The operator owns the customer (the gambler). Evolution supplies the engine.

The revenue model isn’t complicated, which is exactly why it’s so powerful.

1. The Commission (The Main Event):

Most revenue comes from commission, calculated as a percentage of the operator’s profit generated by Evolution’s platform. This is perfect alignment. If Evolution helps an operator earn more money, Evolution earns more money. It scales beautifully.

2. The “Bespoke” Fees:

Evolution also earns fees for dedicated tables and environments. Think of this as selling “premium real estate” inside the casino. An operator wants branded tables, VIP setups, or dedicated dealers? They pay for the privilege.

3. The Secret Sauce: The Operating System

Now, here is the part investors routinely underweight. Evolution isn’t just a library of video games. It’s an operating system plus a broadcast-quality production machine.

Evolution’s “One Stop Shop” (OSS) is a single integration layer and unified back office across its brands. It is explicitly positioned to make integration faster and simpler. It’s the base for distributing more products to more operators.

And reliability? It’s not a “nice to have.” In live casino, reliability is the permission slip to be the default supplier.

• System availability was 99.96% in 2024, excluding scheduled maintenance.

That is incredible. That is “TV-studio” reliability.

The distribution breadth is real, too. Evolution had about 800 customers at the end of 2024. While concentration exists (we’ll get to that later), this platform isn’t a one-client wonder.

But make no mistake, this business is still mostly about the Live Dealer.

• In Q3 2025, Live net revenues were €431.7m.

• RNG (slots/animated) net revenues were €75.5m (IFRS).

The Flywheel: Build content at scale → Distribute it through one integration stack (OSS) → Keep uptime high → Earn commission that scales with the operator’s success.

Atomic Take: Evolution’s real product is “operator-grade casino infrastructure.” The content is the bait; the integration and uptime are the handcuffs.

Falsifier: If availability (2024: 99.96%) degrades meaningfully while competition intensifies, the “infrastructure moat” claim weakens fast.

WHAT WENT WRONG

So, if the machine is so good, why is the stock giving people heartburn?

2024 wasn’t a meltdown. It was operational reality asserting itself. Management describes two categories of pain, and neither is “players stopped playing.”

1. Physical Disruption (The Real World):

The CEO describes disruptions at the Georgia studio caused by external sabotage. This isn’t just code breaking; this is physical interference.

2. The Cyber War:

They faced cyber-attacks targeting Asian operations, including sophisticated attacks against video streaming.

Management says this contributed to slower growth than in prior years. More importantly, they warned that the measures taken to fix this are expected to affect growth for “some quarters ahead.”

Stop & Think: This matters because Evolution is operationally heavy. Live casino isn’t pure software. It’s studios, dealers, cameras, streaming, monitoring, and compliance. When the video feed gets attacked, you interrupt the “always on” expectation that operators depend on.

3. The Regulatory Plumbing:

This is less dramatic than sabotage, but arguably more important. Evolution describes implementing technical measures to ring-fence locally regulated markets. This ensures games are available only through locally licensed B2C operators where local licenses exist.

They noted that these adjustments affected margins in the short term. “Regulated markets” being 40% of net revenues doesn’t tell you how hard it is to operate inside those markets. It tells you where the money is showing up. The friction shows up in costs, process, and occasionally headlines.

The Scorecard:

Q3 2025 put numbers on the bad vibes.

• Net revenues: €507.1m (IFRS), down 2.4% YoY.

• Adjusted EBITDA margin: 66.4% (company-defined non-IFRS).

So the “break” isn’t “the model stopped working.” It’s that the model got stress-tested in the two places you least want surprises: operations (disruption/cyber) and compliance plumbing (ring-fencing).

Atomic Take: Evolution’s bear case isn’t “bad games.” It’s “a world-class machine forced to run with sand in the gears.”

Falsifier: If Asia is described as “volatile” quarter after quarter while net revenues remain negative YoY, the “one-off disruption” framing stops being credible.

REBOUND CATALYSTS

Catalysts must be testable, or they are just motivational posters hanging in the breakroom. We don’t invest in “hope”; we invest in “evidence.”

Here is the checklist to see if this machine is getting back on the highway:

1. The Asia Repair Job

Management still frames Asia as having huge long-term potential. But they are also admitting the region has been hit by cyber-attacks and remains “volatile.”

• HYPOTHESIS: If the cyber mitigation works and the operational routing holds up, Asia stops being a drag and becomes neutral-to-positive again.

• MONITOR: Watch the quarterly commentary like a hawk. Does the phrase “Asia remains volatile” disappear from the outlook? Do they stop talking about streaming disruptions? Silence is golden here.

2. The Relentless Product Conveyor Belt

Evolution is trying to build its way out of trouble. They are calling 2024–2025 “product leap years.” They launched 109 new games in 2024, and the outlook said 110+ new games in 2025.

• HYPOTHESIS: High release cadence plus that “One Stop Shop” (OSS) distribution keeps their share of the operator’s wallet rising, even if the total market slows down.

• MONITOR: The Live vs. RNG split. In Q3 2025, it was €431.7m Live vs. €75.5m RNG. We need to see if RNG momentum is real, or just marketing copy pasted into the earnings slides.

3. The Margin Moat

The outlook maintains a margin estimate of 66–68%. That is the line in the sand.

• HYPOTHESIS: Evolution can keep margins in that elite range while spending enough to fix the plumbing and defend the moat.

• MONITOR: Whether ring-fencing and compliance measures continue being cited as margin headwinds. If the costs stay high, the moat is leaking.

4. The Shareholder Paycheck

The Q3 2025 material repeats the promise: a minimum of 50% of net profit and 100% of excess cash goes back to you. They also flagged a repurchase intention of €500m for 2025.

• In Q3 2025 alone, repurchases were €187.0m.

• Year-to-date 2025 by Q3, they bought back €406.5m.

Atomic Take: The rebound setup is straightforward: Stabilize Asia, keep shipping new games, and let the capital return engine do its thing. But each piece has a measurable scoreboard.

Falsifier: If the margin estimate (66–68%) holds, but operating cash flow after investments (Q3 2025: €342.1m) starts to shrink materially, something is leaking in the engine room.

Accounting profit is nice; cash is truth.

FINANCIAL QUALITY RUBRIC (1–5)

Let’s grade the exam paper.

Margins / Scalability: 5/5

FY2024 EBIT margin was 64.1% (IFRS) and adjusted EBITDA margin was 68.4%. Folks, these are software monopoly numbers.

The economics remain exceptional. It costs them very little to add the next million dollars of revenue.

Cash Generation: 4/5

In Q3 2025, operating cash flow after investments was €342.1m.

Cash conversion was 83%.

That is a cash cannon. However, don’t ignore the fine print: working capital contributed €35.2m to that number. That’s a temporary tailwind, not a permanent feature.

Balance Sheet Resilience: 4/5

As of 30 Sep 2025, cash and cash equivalents stood at €656.4m and the bond portfolio was €103.2m. Total equity was €3,819.6m (IFRS).

This is not a distressed asset. This is a fortress. They aren’t answering angry calls from the bank.

Customer Concentration Risk: 2/5

The top customer was 13% of net revenues in 2024. The top five were 46%. This is the “Whale Risk.”

It’s not fatal, but it is absolutely a “one operator can bruise a quarter” setup. If a big client sneezes, Evolution catches a cold.

Operational Robustness: 3/5

System availability in 2024 was 99.96% (excluding scheduled maintenance). That is world-class.

So why the low score?

Because the battlefield is hostile. The year included sabotage and cyber disruption. The system is excellent; the environment is dangerous.

Regulatory/Tax Durability: 3/5

The government is coming for its cut.

The effective tax rate increased to 13.5% in FY2024 (up from 6.8% in 2023). The annual report expects Pillar II changes to increase the effective tax rate from 1 Jan 2024.

The model is so profitable it can absorb taxes easily, but the uncertainty taxes the stock’s multiple.

The Audit Note: FY2024 includes “other operating revenues” of €151.1m tied to a reduced earnout liability. The Adjusted EBITDA removes this. This is a reminder: “Headline profitability” can include accounting reversals. Always read the reconciliation.

Atomic Take: The fundamentals still glow. The only reason it’s not a perfect scorecard is the customer concentration plus the real-world operational combat.

Falsifier: If adjusted EBITDA margin (FY2024: 68.4%) compresses sharply without a clear mix or regulatory explanation, the “scalable infrastructure” claim is damaged.

THE STATEMENTS Q3 2025

BALANCE SHEET (THE GEIGER TEST)

As of 30 Sep 2025 (IFRS), Evolution reported:

• Total Assets: €5,173.6m

• Total Equity: €3,819.6m

• Non-Current Liabilities: €150.9m

• Current Liabilities: €1,203.1m

The big question isn’t “can they survive a storm?” They have a bunker. The question is “what do they do with their strength?” Capital allocation is now the main event.

The folks in the C-suite point to €759.6m of cash + bond portfolio as of 30 Sep 2025. This pile of money is clearly tied to shareholder distribution (dividends + buybacks).

Stop & Think: Buybacks are (often) a truth serum. They tell you what management thinks the business is worth.

As of 31 Dec 2024, Evolution held 5,270,961 own shares bought at an average of SEK 1,050.81. And in 2024, they repurchased shares at average prices of:

• SEK 1,278.08 (Jan–Mar)

• SEK 1,010.78 (Jul–Oct)

As I write this in January 2026, the share price (in Swedish krone) of Evolution is SEK 592.

That gap is massive. It’s not a valuation conclusion by itself, but it is a factual reminder: Sentiment can change violently even when the business remains profitable. Management thought it was a steal at 1,000+. The market now hates it at 600.

Someone is wrong.

Atomic Take: Balance sheet strength is a feature, but the real “asset” here is optionality. They can defend the moat, reinvest in the business, and still return capital to you.

Falsifier: If cash + bond portfolio strength (as of 30 Sep 2025: €759.6m) trends down while capital returns remain high, without matching operating cash generation, then capital allocation becomes a risk, not a perk. You can’t pay dividends from a depleting savings account forever.

CASH FLOW (THE TURBINE)

In Q3 2025, the cash picture is clean in the way investors love: big cash generation, modest investment needs.

• Total Investments: €29.8m into intangible and tangible assets.

• Operating Cash Flow (after investments): €342.1m.

• Cash Conversion: 83%.

This is the essence of Evolution. It can generate enormous cash while simultaneously being “high risk” in a different dimension. Financial leverage isn’t the boogeyman here. Regulation, reputation, and hackers are the boogeymen.

The Audit: Q3 2025 cash performance is strong, but a piece is explicitly working-capital timing (€35.2m contribution). If that reverses, cash can look worse next quarter even if the underlying business didn’t change.

Atomic Take: The turbine is still spinning, but read cash generation with the working-capital footnote stapled to your forehead.

Falsifier: If operating cash flow after investments falls materially from €342.1m (Q3 2025) while profit remains €252.3m (Q3 2025), it suggests cash leakage (capex, taxes, or working-capital drag) that changes the “cash machine” thesis.

INCOME STATEMENT (THE REACTOR)

Q3 2025 is the market’s most recent “temperature check,” and it’s running a little cold.

• Net Revenues: €507.1m (IFRS), down 2.4% YoY and 3.3% QoQ.

• Operating Profit: €296.6m.

• Profit for the Period: €252.3m.

• EPS: €1.25 (IFRS).

• Adjusted EBITDA: €336.9m with a margin of 66.4%.

Zoom out a bit. FY2024 still demonstrates the business’s earnings power.

• Profit for the Year: €1,244.0m (IFRS).

• Effective Tax Rate: 13.5%.

So, the reactor is not “off.” It is not melting down. It’s just not accelerating right now—and the market pays a premium for acceleration, not maintenance.

Atomic Take: Evolution’s earnings power looks intact, but Q3 2025 tells you the trajectory is currently pointed down.

Falsifier: If net revenues re-accelerate to sustained positive YoY growth while margins stay around the 66–68% framework, the “growth is broken” narrative is falsified.

VALUATION

Let’s look at the price tag.

Current Price ($EVVTY): $64.69 per share. Market Cap: $12.9bn.

1. Earnings Yield (The Accounting Reality)

Trailing twelve-month “Profit to Equity holders” is €1,132.4m.

Against a market cap of ~$12.9bn, that is an earnings yield of ~10.6%.

2. Free Cash Flow Yield (The Cash Reality)

Trailing twelve-month “Free Cash Flow” is €1,141.7m.

Using the same market cap, that gives us an FCF yield of ~10.7%.

Stop & Think: Usually, accounting profit and free cash flow drift apart. Here, they are nearly identical. That is a sign of high-quality earnings. The cash is real.

3. The Sanity Check (Per Share)

• TTM EPS €5.64 ≈ $6.80. EPS / Price = ~10.5% earnings yield.

• TTM FCF per share ≈ €5.68 ≈ $6.85. FCF / Price = ~10.6% FCF yield.

The market isn’t arguing that Evolution is “low quality.” The market isn’t stupid. It sees the same margins we do.

What the market is doing is pricing in a persistent “Overhang Discount.” It is terrified of three things:

1. Operational disruption (sabotage/cyber).

2. Regulatory/compliance friction (the “sin tax”).

3. Trajectory risk (growth slowing down).

If those risks fade, a ~10–11% yield starts to look like a massive mispricing of durability. If those risks persist, the yield is a rational “stress premium.” You are getting paid extra to handle the headaches.

Atomic Take: At ~10.6% earnings yield and ~10.7% FCF yield (TTM), the stock is priced like a high-quality cash machine... but one with a non-trivial probability of permanent friction.

Falsifier: If TTM net profit and/or TTM FCF (through future quarters) materially decline from the €1.13bn / €1.14bn run-rate while management keeps presenting the issues as “temporary,” the “overhang discount” stops being a discount and becomes a warning label.

RISKS (THE MELTDOWN LIST)

I don’t like to sugarcoat risks. If you own a stock, you need to know what kills it.

1. The “Whale” Risk (Customer Concentration): The top customer was 13% of net revenues in 2024. The top five were 46%. That is not just “risk.” That is “one customer can ruin your quarter.” If a big operator switches suppliers, the stock takes a hit.

2. The War Zone (Operational Disruption): This isn’t theoretical. Management describes sabotage in Georgia and cyber-attacks targeting Asian operations. This impacted the second half of 2024 and is expected to slow growth for “some quarters ahead.” You can’t run a casino if the cameras are down.

3. The Permission Slip (Regulatory Risk): Evolution discusses licenses, compliance, and ring-fencing regulated markets. Failure here is existential. If you lose the license, you don’t have a business.

4. The Sin Bin (AML/Fraud): The annual report highlights exposure to sanctions, fines, and reputational damage. This is the “sin stock” tax. If they get caught in a money-laundering scandal, the multiple collapses.

5. The Taxman Cometh: Evolution flags Pillar II impacts and broader tax risks (including VAT). Uncertainty here is just another weight on the stock price.

6. The Sneaky Risk: People: In 2024, the Group averaged 20,973 employees. More than 85% of them are tied to operational gaming activity running 24/7. This is a labor-heavy tech company. Any staffing disruption—strikes, labor shortages, illness—becomes a throughput issue, not just a cost line.

Atomic Take: Evolution’s biggest risks are “license to operate” risks. You can’t spreadsheet your way out of reputational damage.

Falsifier: If regulated-market growth continues while margins stabilize (meaning the ring-fencing costs stop rising), the risk stack compresses meaningfully.

THE ATOMIC VERDICT

Status: Overhang-Heavy

Stars: ★★★★☆ (Business Quality) ★★★☆☆ (Current Clarity)

The case for “Still Exceptional” is not subtle. FY2024 showed extreme profitability: €2,214.1m total operating revenues and €1,419.6m EBIT (IFRS).

Even Q3 2025 showed a 66.4% adjusted EBITDA margin.

The case for “Still Messy” is also not subtle. Management describes disruption, sabotage, and cyber issues. Q3 2025 net revenues were €507.1m, down 2.4% YoY.

What Increases Confidence (Measurable):

• Net revenues return to sustained positive YoY growth from the Q3 2025 base.

• Asia stops being flagged as “volatile” in the outlook.

• Operating cash remains strong without needing a working-capital tailwind (Q3 2025 WC contribution was €35.2m).

Upgrade Triggers:

• Sustained positive YoY net revenues.

• Operating cash flow after investments stays robust even as working capital becomes neutral.

• Margin estimate 66–68% is reiterated without escalating caveats.

Downgrade Triggers:

• Continued cyber language persists and growth stays negative.

• Customer concentration worsens beyond the FY2024 baseline (13% top customer, 46% top five).

• Effective tax rate rises further beyond 13.5% without offsetting growth.

Atomic Take: Evolution still looks like a compounding-quality business—but the overhangs are active enough that the signal isn’t clean.

FINAL WORD (FOR NOW)

Evolution is a Ferrari driving through a hailstorm. The engine is perfect (60%+ margins). The driver is skilled (founder-led culture). But the hail (cyber attacks, regulators, Asian volatility) is denting the bodywork.

If the storm passes, you have a Ferrari bought at a discount. If the storm is permanent climate change, you have expensive repairs forever.

Stop & Think: Do you believe the “Asian volatility” is a bad quarter, or a new reality? Your answer to that question determines your trade.

Is the “Sin Stock” discount worth it? Would you buy at SEK 600 knowing management bought at SEK 1,200? Let me know below.

References

• Evolution AB (publ) — Interim Report January–September 2025 (presentation deck), dated 23 Oct 2025.

• Evolution AB (publ) — Annual Report 2024.

• Evolution AB (publ) — Press release: “Acquisitions of own shares in Evolution AB (publ)”, dated 15 Dec 2025.

• Evolution AB (publ) — Share repurchase / buyback overview document (PDF)

• Evolution AB (publ) — Financials workbook (Excel)

Disclaimer

This Deep Dive is an educational breakdown of a public company based on information available in the materials provided (e.g., annual/quarterly reports, investor presentations, earnings transcripts) and my interpretation of that information. It is designed to be a “bolt-on” intelligence layer to your own due diligence — not a replacement for it.

Independence: I do not accept compensation of any kind from the companies discussed. My research is driven solely by my personal search for high-quality compounders.

Skin in the Game: Unless otherwise stated, assume the author may hold a long position in securities mentioned. Any position creates bias — treat this as commentary, not gospel.

Not Financial Advice: Nothing here is investment advice, a recommendation, or a solicitation. I am not a financial advisor. You are responsible for your own decisions. The stars rating is not a buy recommendation, but meant as a guide to understand the quality of the financial statement of the respective companies.

Error & Update Risk: Financial statements change, companies restate, guidance evolves, and I can be wrong. Verify key figures in the primary filings and consider reading the footnotes before deploying capital.

New to The Atomic Moat? This analysis of Admiral ($ADM) is a prime example of how we dissect high-quality compounders. If you want new deep dives sent to your inbox every Sunday, join 900+ other investors below.

Join for Free

Folks, I have a confession. I love boring companies, but sometimes I forget that I do.

I spent last week deep in the archives, reading about Shelby Cullom Davis for a content piece (that you can now find on Atomic Moat under “The Money Mind”).

For those who don’t know, Davis was the guy who turned $50,000 into nearly $900 million. He didn’t do it by chasing the “hot dot-coms” of the 1950s or the “Nifty Fifty” darlings. He did it by buying insurance stocks when they were unloved, misunderstood, and frankly; boring as dirt.

Then, I looked up from my desk and saw what’s happening in the market right now.

It feels like a frantic party where everyone is trying to squeeze through the same door marked 'AI & Space.' You have companies trading at 50x, 70x, even 100x earnings just because they uttered the magic words 'Generative AI' or 'Total Addressable Market.'

That got me thinking: What if the real money isn’t in the stars, but on the pavement?

I wanted to find the antidote to the hype. I wanted to see if there was a “Davis-style” compounder hiding in the absolute most boring sector I could find.

That led me to Admiral.

While everyone else is trying to figure out which semiconductor chip will rule the world in 2035, Admiral is doing something beautifully simple: figuring out how to insure a Vauxhall Corsa without losing money.

This isn’t about being a contrarian for the sake of it. It’s about asking a simple question: Is there solid, double-digit growth available in a sector that most people fall asleep analyzing?

Shelby Cullom Davis proved that you don’t need a rocket ship to get to the moon; sometimes you just need a disciplined insurer and a lot of patience.

Let’s dive in.

If You’re In a Rush (The 30-Second Elevator Pitch)

• What they actually do: They sell UK insurance (mostly cars), dabble in “Beyond Motor” lines (like house insurance), and run a lending arm called Admiral Money that’s starting to actually move the needle.

• Challenges: Insurance results are a “timing factory.” Between the insurance cycle, discounting rates, reserve development, and profit commissions, analysts spend all day arguing about what is “real profit” and what is just “accounting weather.”

• What fixes it: Market price declines are finally slowing down, Admiral is staying disciplined (not chasing bad business), and the growth in non-motor and lending isn’t watering down the quality of the main business.

• Atomic Position: Currently on my watchlist at the time of writing. I see a credible path to High Conviction here, but only if the underwriting and capital return engine stays boring. Boring is good.

Share

The Setup: A scorecard that demands respect

Admiral’s H1 2025 is a strong print. It’s the kind of report that makes people forget just how cyclical and nasty the motor insurance game can be when it rains.

Let’s look at the hard numbers. No fluff, just the scorecard:

Group profit before tax was £521m .

Group combined ratio was 77.7%. Editor’s Note: For those new to insurance, anything under 100% is profit. 77.7% is printing money.

Group customers were 11.4m.

Solvency ratio was 194%. That’s a fortress balance sheet.

Interim dividend was 115.0p.

The Variant Perception: Stripping away the Theatre

Here is where the argument lives.

• The Market Take: “It’s just an insurer. Enjoy the good half-year. But don’t you dare try to annualize these earnings like it’s a SaaS company. The rain is coming.”

• The Bull Take: “You’re missing the point. This is a disciplined underwriting machine with unusually explicit capital return behaviour. They give the money back.”

• The Bear Take: “The cycle is rolling over. Claims inflation and market price declines are going to eat the margin. All those ‘extras’—like assumption benefits and timing differences—are just flattering the headline number.”

This deep dive is about one thing: Is Admiral’s economic engine structural enough to stay attractive when pricing gets soft? Anyone can make money in a hard market. I want to know if they can swim when the tide goes out.

Atomic Take: Admiral looks like a genuine operator. But remember, the insurance cycle is still the radiation source—so you underwrite the dials, not the vibes.

Falsifier: If the Group combined ratio moves back above 79.8% (H1 2024, %, as presented) and stays there for two consecutive half-years, the thesis is in trouble.

How the Business Actually Makes Money

Admiral’s core trick is simple in concept and incredibly hard to pull off in the real world:

1. Price the risk better than the guy across the street.

2. Handle claims better than the guy across the street.

3. Keep costs tighter than a drum.

4. Use risk transfer (reinsurance) to turn that good underwriting into cash we can spend, without living on the edge of bankruptcy.

Don’t get distracted by the revenue story. Revenue is vanity; profit is sanity. Let’s look at where the cash actually comes from. On a continuing basis, segment profit before tax in H1 2025 was:

• UK Insurance: £584m.

  • UK Motor: £559m. <-- This is the engine.

  • Household: £25m.

• Admiral Money: £16m, importantly including £10m from the sale of back book loans.

• Other Group items (including Pioneer): ( £43m ).

Subscribe now

Translation: UK Motor is still the nuclear reactor powering this city. Everything else is either an amplifier, a hedge, or an expensive experiment that you’re hoping matures into something “strategic.”

(Household is currently the most convincing of the diversifiers because it’s already profitable in the period presented. That’s a good sign.)

Stop & Think: Why do insurers try to diversify into “Household” or “Lending”? Because cars are dangerous. But notice that after all these years, £559m still comes from cars and only £25m from houses. It’s hard to beat your core competency.

The Holy Trinity: Underwriting Quality

Underwriting quality is the central KPI because it forces honesty. You can’t fake these numbers for long. Admiral reports:

• Group loss ratio: 57.4%

• Group expense ratio: 20.3%

• Group combined ratio: 77.7%

Why do I obsess over these ratios? Because they drive the entire narrative.

Underwriting profit funds the dividends. It supports the solvency. It buys time for “Beyond Motor” to scale up without the group turning into a balance-sheet gymnastics routine.

The Admiral Money Trap: Admiral Money is the optionality, and the optics trap. This business can absolutely be a value creator (there’s real net interest income and portfolio scale there), but look closely at H1 2025: the profit includes £10m from selling back book loans. The loan book is now £1.28bn on-balance sheet.

If you want a single “don’t get irradiated” rule for investing in insurers who lend money: Separate recurring economics from one-time portfolio actions. Then, watch the credit outcomes like a hawk as the book grows.

Don’t be Naïve

The easiest way to misread Admiral is to treat its reported profitability as a single clean line. It isn’t. Even when everything is “legit” and above board, insurance earnings are full of timing levers:

• Profit commissions

• Reserve development

• Discount unwinds

• Assumption impacts

The right question is not “is this pure?” It’s never pure. The question is: “Is the underlying underwriting behaviour strong enough that the timing levers don’t become a crutch?”

Atomic Take: Admiral’s moat is operating discipline (pricing + claims + cost) plus capital plumbing—but you only get paid if that discipline survives softer pricing.

Falsifier: Watch out if the UK Motor combined ratio deteriorates meaningfully from 71.2% (H1 2025, %, as presented) and stays worse for two consecutive half-years.

What Went Wrong (The “Oops” Files)

Admiral’s “went wrong” story isn’t a scandal. It’s the normal arc of the insurance story: Pricing peaks, competition comes back, claims costs creep up, and suddenly every operational weakness you didn’t know you had becomes a slide deck in the boardroom.

The market data management highlights are clear: Prices are still down year-on-year, but the rate of decline is slowing.

• In Q2 2025, ABI average paid premium was £562 (GBP), down 10% YoY and down 3% versus Q1.

• Confused new business quote was £755 (GBP), down 14% YoY but only down 1% versus Q1.

That “slowing decline” is the cycle pivot point Admiral cares about. Not because lower prices are “good,” but because a softer rate environment tests whether you can grow without buying bad business.

The Cleanup on Aisle 4: Then you have the operational clean-up: the Total Loss review. It’s completed, actions are expected to start in H2, and the estimated total cost is c.£50m (GBP).

Does this matter? Yes. It tells you something about management’s control posture. Either it’s a bounded fix that cleans up a process, or it’s the first crack in the claims discipline story. The number itself is less important than whether it stays contained.

Europe: Europe is another “fix it” theatre. The Europe combined ratio improved to 98.3% (H1 2025, %, as presented) but profitability is still around breakeven to slightly negative (H1 2025, GBP, as presented).

That’s not a meltdown. But it is a reminder that “international diversification” often means “learning expensive lessons in public.”

Atomic Take: Nothing here screams “broken model,” but the cycle + operational clean-ups are exactly how good insurers get humbled.

Falsifier: If Total Loss remediation costs drift materially above c.£50m (GBP, as presented) or expand into a broader claims-control problem, run for the hills.

The Dashboard, The Report Card, and The Plumbing

So, Admiral is a cash-generating machine in a boring industry.

Now, let’s talk about what keeps the machine running—and what could break it.

Most companies try to dazzle you with “synergies” and “AI-driven flywheels.” Admiral doesn’t do that. They have a dashboard of simple, mechanical dials. Our job is to stare at them until our eyes water.

Rebound Catalysts: No Moonshots, Just Mechanics

Admiral’s catalysts are refreshingly measurable. There is no magic here. Just a handful of inputs that will either stay stable or betray you.

Catalyst 1: The Pricing Plateau

The market pricing is stabilizing. This is crucial. We don’t need prices to skyrocket; we just need them to stop plummeting so we can grow without hurting ourselves.

• The Fact: Q2 2025 market trackers show ongoing declines, but the bleeding is slowing. The ABI index sits at £562 (GBP), and Confused is at £755 (GBP).

• The Monitor: Watch this like a hawk. If the “slowing” stops and the freefall continues, or if Admiral tries to grab market share while prices are dropping, run.

Catalyst 2: The Golden Goose (UK Motor)

Can the core business stay elite when the easy money is gone?

• The Fact: UK Motor profit before tax was £559m (H1 2025, GBP) with a Combined Ratio of 71.2% (H1 2025, %, as presented).

• The Hypothesis: Even as the “rate tailwinds” (easy price hikes) fade, Admiral wins because they are simply better at the boring stuff: answering phones, processing claims, and pricing risk.

• The Monitor: Watch the split between the loss ratio and the expense ratio. If either creeps up, the edge is dulling.

Catalyst 3: The “Beyond Motor” Dream

Can they sell something other than car insurance without losing their shirt?

• The Fact: Household profit before tax was £25m (H1 2025, GBP) and the Household combined ratio was 83.9% (H1 2025, %, as presented).

• The Hypothesis: They are taking the “Motor Playbook” (obsessive discipline) and exporting it to home insurance. It looks like it’s working.

• The Monitor: Consistency. One good half-year is luck; three is a trend.

Catalyst 4: Admiral Money (The Side Hustle)

This needs to be a recurring earnings stream, not a one-off transaction factory.

• The Fact: Admiral Money profit before tax was £16m (H1 2025, GBP). BUT—and this is a big but—this includes £10m from the sale of back book loans. On-balance sheet loans are now £1.28bn (as of 30 Jun 2025, GBP).

• The Monitor: Strip out the one-offs. I want to see credit outcomes (are people paying them back?) and recurring profit.

Catalyst 5: The “Free Lunch” (Investment Income)

Finally, interest rates are helping us.

• The Fact: Investment income was £99m. The underlying investment income yield is 3.9%, and they are reinvesting at a rate of ~4.5% with a bond duration of ~3.0 years.

• The Monitor: This is a nice tailwind. As long as they don’t get cute and buy risky junk bonds, this is free money.

Atomic Take: The catalyst stack is clean. If the insurance cycle stays merely “less good” rather than “disastrous,” this model works.

Falsifier: If the Group combined ratio rises meaningfully above 77.7% while the customer count fails to grow from 11.4m, the growth story is dead.

Subscribe now

The Financial Quality Report Card

I like to grade companies like I’m a strict schoolteacher. Let’s see how Admiral scores on the “Sleep Well at Night” rubric.

1. Underwriting Discipline: 5/5 (Elite)

A 77.7% group combined ratio (H1 2025, as presented) is world-class. It’s the Steph Curry of insurance stats. The only question is durability, not competence.

2. Expense Control: 4/5 (Improving)

The expense ratio improved to 20.3% from 22.8% (H1 2025 vs H1 2024, %). That is real operating leverage. If they can keep costs down while pricing softens, that’s a massive competitive advantage.

3. Capital Strength: 4/5 (Solid)

Solvency is 194% (H1 2025). That is a fortress. But remember, this is a dial, not a static number. When you pay out huge dividends, you are draining the moat. Watch it closely.

4. Cash-to-Dividend Coherence: 3/5 (Murky)

This isn’t Admiral’s fault; it’s the industry. Insurance cash flow is a timing instrument. You get paid today for a crash that might happen next year. Dividend debates in this sector are where investors lose their hair.

5. Diversification Quality: 3/5 (Promising but Unproven)

Household looks legitimate (profitable, strong combined ratio). Europe is improving but still hovering around breakeven. Admiral Money has promise, but that H1 profit included a notable one-time boost. I need to see more “boring” profit here.

6. Earnings Transparency: 3/5 (Standard)

Insurance is inherently a timing business. Admiral discloses a lot—more than most—but your job as an investor is to separate “cycle luck” from “structural genius.”

Stop & Think: Admiral reports an ROE of 57% (H1 2025, %, as presented). That number is fantastic. But be careful. In insurance, a super-high ROE is often a warning sign that the cycle is peaking. It’s an outcome of the moment, not a permanent law of nature.

Atomic Take: The quality is real, but the market is unforgiving. It will treat any crack in underwriting as radioactive.

Falsifier: If Solvency trends down materially from 194% (H1 2025, %) while underwriting ratios also deteriorate, the safety net is gone.

The Balance Sheet: The Geiger Test

Let’s look at the plumbing. We start with what can kill you: Liabilities.

• Total Assets: £8,632.8m (as of 30 Jun 2025, GBP).

• Total Equity: £1,435.8m (as of 30 Jun 2025, GBP).

• Insurance Contract Liabilities: £5,261.2m (as of 30 Jun 2025, GBP).

• Subordinated/Other Liabilities: £1,509.7m (as of 30 Jun 2025, GBP).

• Cash and Investments: £5,427.8m (as of 30 Jun 2025, GBP), including cash of £319.2m (as of 30 Jun 2025, GBP).

The Translation: These numbers describe a specific shape: A massive investment base sitting alongside massive potential bills (liabilities), with a layer of Equity sandwiching them together. This is normal for an insurer.

The risk isn’t the structure. The risk is the alignment. Are the assets (cash/investments) safe enough and liquid enough to pay the liabilities when they come due?

Hypothesis: Admiral’s balance sheet is a “containment vessel.” It supports a high distribution model (big dividends) because the underwriting engine throws off enough profit to refill the tank, and the investment base is managed conservatively.

Monitor: Watch Solvency versus payout behaviour. If they start changing the investment mix to chase higher yields (buying junk bonds or weird derivatives), it means the underwriting engine is sputtering.

Falsifier: If Solvency falls materially below 194% (H1 2025 baseline) alongside any sign that the dividend framework is becoming “rigid” (paying out more than they should just to keep the stock price up), get out.

Cash Flow: The Turbine

Insurance cash flow is where investors invent ghosts. They see huge numbers moving in and out and panic.

FACT: Net cash flow from operating activities was £240.0m (H1 2025, GBP).

FACT: Cash flow from operating activities before movements in investments was £559.2m (H1 2025, GBP).

FACT: Purchases of property, equipment and software were ( £32.6m ) (H1 2025, GBP).

FACT: Equity dividends paid were ( £366.5m ) (H1 2025, GBP).

The Translation: Here is the clean way to read this without getting emotionally injured.

First, ignore the headline “Net cash flow” number. Why? Because insurers have massive investment portfolios that churn constantly. If they buy a bond, it looks like cash leaving the building. It isn’t. It’s just moving from the left pocket (cash) to the right pocket (investments).

Look at the “Operating cash before movements in investments” line: £559.2m. That is the real cash engine. Subtract the “capex” (buying computers and desks): £32.6m.

The Napkin Math:

• Real Cash In: ~£526.6m (£559.2m - £32.6m)

• Dividends Paid Out: £366.5m

The turbine is working. The cash coming in covers the dividend with room to spare.

Atomic Take: The turbine works. The investor’s job is to keep the turbine connected to solvency reality, not to a simplistic “cash cover” meme.

Falsifier: If Cash flow (before investment moves) drops materially from £559.2m (H1 2025 baseline) while dividends paid stay high, the tank is leaking.

Income Statement: The Reactor

The P&L is where Admiral looks most obviously attractive—and where the “seasoning” matters most.

• Underwriting Purity: Group combined ratio was 77.7% (H1 2025, %, as presented). This is the steak.

• The Seasoning: Management notes an estimated additional Ogden benefit of c.£15–20m included in profit. Also, Admiral Money profit includes £10m from the sale of back book loans.

The Audit: These “seasoning” items don’t make the result fake. But they do mean you should be careful about annualising the headline number. If you strip them out, the business is still profitable, just slightly less “record-breaking.”

Atomic Take: The reactor is running hot and clean right now—but you don’t call it “normalised” until it survives a softer pricing environment without the extra spices.

Share

What about the valuation?

Now we’ve kicked the tires (the business model) and checked the engine (the catalysts). Now we need to look at the fuel gauge (cash flow) and decide if the asking price is a bargain or a rip-off.

Most investors get lost here. They drown in spreadsheets. We are going to keep it simple. We are going to value this business like we’re buying a local laundromat, not a spaceship.

We have two key inputs and one dangerous temptation.

• Price: ~£30.3 per share (16 Jan 2026).

• EPS: 132.5p (H1 2025, pence)

The Temptation: The lazy move is to double the H1 EPS and declare victory.

• Run-rate EPS ≈ 132.5p × 2 = 265.0p.

• Implied PE Multiple: ~3,030p / 265.0p ≈ ~11.4x.

The Sentiment Meter: At 11.4x earnings, the market is not paying a premium for perfection. It’s pricing in skepticism. It’s saying, “We know this is peak earnings, and we don’t believe it will last.”

This is where the opportunity lives. If the market is pricing in a crash, and you only get a “soft landing,” you make money.

Atomic Take: At ~11.4x annualised H1 EPS, the market is pricing durability, not heroics. The bet here is on underwriting persistence, not financial engineering.

Risks: The Meltdown List

What kills us? Let’s be specific.

1. Cycle Risk: Market pricing is still declining (Q2 2025 ABI £562). If this accelerates, margins get crushed.

2. Inflation: Management shows indexed market total burn cost rising. If inflation beats pricing, game over.

3. The “Fix-It” Bill: The Total Loss review cost is c.£50m. If this number grows, or if they find another closet full of skeletons, the stock tanks.

4. The Sneaky Risk (Dividend Rigidity): The policy is to pay 65% of post-tax profits. The market loves this. But if profits drop and they keep paying out to keep the stock up, they destroy the balance sheet.

5. Admiral Money: On-balance sheet loans are £1.28bn. If the economy tanks, these loans could go bad fast.

Atomic Take: The failure mode isn’t a single explosion; it’s a slow squeeze—pricing down, claims up, and dividends becoming a trap.

The Atomic Verdict

Status: ★★★★☆ (3.9/5)

Earned by underwriting quality, discounted for cycle exposure.

Status: Watchlist (Defensive-ish, with explicit cycle sensitivity).

The Pro Case: It rests on two measurable pillars:

1. Elite Underwriting: A 71.2% combined ratio in UK Motor is not “fine.” It is world-class.

2. Capital Return: They have a fortress balance sheet (194% Solvency) and they actually give the money back to you (115.0p interim dividend).

The Skeptical Case: The market is softening. The “easy money” period of hard pricing is over. Now we find out if Admiral is a structural genius or just a surfer who caught a good wave.

The Cheat Sheet: When to Buy, When to Sell

(Important; the following is not financial advice but only meant as guidelines for what I think is important to focus on IF you decide to research the stock further)

UPGRADE TRIGGERS (Green Light)

• Group combined ratio stays at or below 77.7% while customers grow above 11.4m.

• Household continues profitable scaling with combined ratio staying around 83.9%.

• Admiral Money grows beyond £1.28bn with recurring profit (no more one-off sales).

DOWNGRADE TRIGGERS (Red Light)

• UK Motor combined ratio deteriorates meaningfully above 71.2% and stays there.

• Remediation costs exceed c.£50m or we hear the phrase “another review.”

• Solvency trends down materially from 194% while they keep paying aggressive dividends.

Stop & Think: Great investing isn’t about being smarter than everyone else. It’s about being more disciplined. Admiral is a test of discipline—theirs and yours.

So, let’s circle back to where we started.

Right now, the market is screaming at you to buy the future. It wants you to pay 50x earnings for a company that might mine asteroids or solve general intelligence in 2032. It’s exciting. It’s sexy. It makes for great dinner party conversation.

But remember Shelby Cullom Davis.

He didn’t build his dynasty by guessing the future. He built it by betting on the things that never change. People will always crash their cars. They will always need someone to pay for it. And they will always hate paying their premiums.

Admiral is the embodiment of that “unsexy” truth.

While the rest of the world is looking at the stars, Admiral is staring at the spreadsheet. While tech bros are talking about “total addressable markets in the trillions,” Admiral is arguing about how to save £5 on a bumper repair.

At ~11x earnings, you aren’t paying for a dream. You are paying for a machine.

Is Admiral going to be the next Nvidia? No. But Shelby Davis didn’t need Nvidia. He just needed a company that compounded capital, respected the downside, and didn’t get high on its own supply.

The question you have to answer this weekend isn’t about interest rates or AI. It’s much simpler: Are you willing to be bored enough to get rich?

References

1. Admiral Group plc — HY25 consolidated pack (results presentation + financial statements), dated 14.08 (PDF).

2. Admiral Group plc — ADM-Admiral Group PLC.xlsx (company-provided financial statement spreadsheet used for cash flow and supporting line items).

3. Admiral Group plc — UKI (PDF) (business model / UK Insurance operational context as presented).

4. Admiral Group plc — Video transcript (PDF)

Disclaimer

This Deep Dive is an educational breakdown of a public company based on information available in the materials provided (e.g., annual/quarterly reports, investor presentations, earnings transcripts) and my interpretation of that information. It is designed to be a “bolt-on” intelligence layer to your own due diligence — not a replacement for it.

Independence: I do not accept compensation of any kind from the companies discussed. My research is driven solely by my personal search for high-quality compounders.

Skin in the Game: Unless otherwise stated, assume the author may hold a long position in securities mentioned. Any position creates bias — treat this as commentary, not gospel.

Not Financial Advice: Nothing here is investment advice, a recommendation, or a solicitation. I am not a financial advisor. You are responsible for your own decisions. The stars rating is not a buy recommendation, but meant as a guide to understand the quality of the financial statement of the respective companies.

Error & Update Risk: Financial statements change, companies restate, guidance evolves, and I can be wrong. Verify key figures in the primary filings and consider reading the footnotes before deploying capital.

New to The Atomic Moat? This analysis of Duolingo ($DUOL) is a prime example of how we dissect high-quality compounders. If you want these deep dives sent to your inbox, join 900+ other investors below.

Join for Free

I remember the first time I ever tried the Duolingo app. It was to learn Italian back in 2019 when me and my girlfriend went to Rome one hot summer. I thought the app was a new and exciting way to learn the language back then.

And my Italian knowledge today is, ehm…

…alright, alright, I have forgotten every single word I learned, except “pane” (bread) for some reason.

But that’s not the point.

I recently picked it up again to learn Spanish, and I was really impressed with their new functions and how much it has improved since last time.

But do you want to know what is even more impressive? Their cashflow. And when I discovered this, I thought the company might deserve a thorough deep dive.

Subscribe now

So, here I present; The Duolingo deep dive.

It looks to me that Duolingo is what happens when a consumer app quietly figures out how to act like a serious cash-generating business, without losing its sense of humor (or that owl-shaped menace that haunts your lock screen).

If You’re In a Rush

• What they actually do: Duolingo sells language learning (and now they’re branching into other “subjects”) primarily through subscriptions. They sprinkle in a smaller mix of ads, in-app purchases, and the Duolingo English Test.

• Why people hate it: The whole story sits on two fragile things that don’t show up neatly on a balance sheet: user engagement and distribution control. They live on rented land: Apple, Google, and Stripe processed 62.2%, 23.3%, and 10.1% of total revenues respectively. That’s a lot of power in someone else’s hands.

• What fixes it: They need to keep those Daily Active Users (DAUs) and paid subscribers compounding while keeping a lid on AI and hosting costs. We don’t want those costs eating the lunch.

• Atomic Take: Duolingo’s operating machine looks real. The only question is whether the engagement engine stays “unfair” as AI costs rise and that dependence on the app stores gets louder.

• Falsifier: If DAUs fall below 30% YoY for two consecutive quarters (metric: company-reported DAUs YoY %), the thesis is broken.

Disclaimer: I own shares in Duolingo (and be sure to have this in mind when you read the deep dive).

Subscribe now

The Setup

Here’s the “why now” in plain English: Duolingo crossed 50.5 million DAUs (Q3 2025, users) while printing $271.7 million of revenue.

Even better, they generated $77.4 million of free cash flow. That is a lot of cash for a language app.

Under the hood, the money-making side kept climbing, too.

Paid subscribers hit 11.5 million and the penetration rate (the number of monthly users who actually pay up) rose to 9.0% of LTM MAUs.

Stop & Think: Getting 9 out of every 100 casual users to open their wallet for a free app is harder than it sounds. That number is the pulse of the business.

Why Smart People Disagree

Revenue grew 41% YoY to $271.7 million and total bookings grew 33% YoY to $281.9 million.

The bull case? Bulls see a subscription-first consumer product with a widening moat built on “habit + brand + constantly tweaking the product.” They see a profit margin that can expand even while they invest in the future.

And let’s not forget the bear case: Bears see a platform-risk consumer app where three things go wrong:

1. Apple and Google take their heavy toll.

2. AI costs push that gross margin down.

3. Engagement is a fickle god that demands constant sacrifice. Users get bored.

What to MONITOR: Watch subscription bookings growth versus DAU growth. If the bookings momentum disconnects from the engagement, the “pricing power” story weakens. For the record, Q3 2025 subscription bookings were $240.3 million, up 36% YoY.

Atomic Take: This is a quality story with one rule: the KPIs are the product, and the product is the KPIs.

The Falsifier: If Paid subscribers decline sequentially (metric: period-end paid subscribers, company-reported), we have a problem.

How the Business Actually Makes Money

Let’s start with the boring truth: Duolingo is a revenue recognition story masquerading as a meme brand.

They generate cash primarily from subscriptions (monthly and annual; plus the family plan annual), plus advertising, in-app virtual goods, and the Duolingo English Test.

In Q3 2025, subscription revenue was $229.5 million out of total revenue $271.7 million. Do the math, and that’s roughly ~85% subscription by revenue mix.

Share

“Other revenue” is real, but let’s be honest, it’s not driving the reactor:

• Advertising revenue: $21.0 million

• Duolingo English Test: $9.6 million

• In-App Purchases: $11.1 million

So, what is “bookings” doing here?

It’s the cash signal that matters when you sell time-based subscriptions, especially annual plans. Deferred revenue is the accounting proof that cash is arriving before the GAAP revenue does. It’s the best kind of liability—money you’ve already collected for work you haven’t done yet.

Deferred revenue ended at $441.7 million.

And the deferred revenue bridge shows just how subscription-heavy this business is: “Deferral of revenue” was $754.1 million while “Recognition of deferred revenue” was $(345.3) million.

They are stacking cash faster than they can recognize it.

Cost Structure Reality Check: The cost of revenues is heavily weighed down by payment processing + hosting + generative AI.

This is why the distribution “tax” matters:

Apple/Google/Stripe collectively process the majority of revenue. They are the toll booth collectors on Duolingo’s highway.

The Moat (What’s Actually Hard to Copy)

Engagement is scaling: 50.5 million DAUs on 135.3 million MAUs (Q3 2025, users).

That implies a ~37% DAU/MAU ratio, which is firmly in “habit product” territory. Most apps would kill for that ratio.

Also, monetization is scaling right alongside it: paid subscribers 11.5 million (period end Q3 2025) and paid subscriber penetration 9.0% of LTM MAUs.

HYPOTHESIS: The real moat is iteration velocity + brand distribution + personalization. AI becomes a retention amplifier here (not just a cost line).

MONITOR: Watch gross margin versus Max expansion. Management explicitly ties margin pressure to generative AI + hosting costs. We need to make sure the AI is worth the squeeze.

Atomic Take: Duolingo’s business model is “subscription first, everything else second,” and the compounding only works if engagement stays sticky enough to keep pushing users into paid.

Falsifier: If Subscription revenue growth falls below MAU growth for multiple quarters (metrics: subscription revenue YoY %, MAU YoY %, company-reported), the thesis is in trouble.

Subscribe now

The Fly in the Ointment (What Went Wrong)

When I say “wrong” here, I am not talking about a collapse. The building isn’t on fire. But there are some choices and optical illusions here that can spook investors who want consumer apps to behave like boring, predictable electric utilities.

1. The Margin Narrative Got Messier

The gross margin declined ~40 bps YoY to 72.5% (Q3 2025, GAAP). Why? Management explicitly attributed this to increases in generative AI and hosting costs.

Here is the trade-off: You want the app to be smarter (AI), but brains are expensive.

2. The Long Game vs. The Quick Buck

The company told you point-blank that it is shifting priorities. Management described leaning more into longer-term initiatives—investing proportionally more in “teaching better,” and prioritizing user growth over monetization in A/B tests (Q3 2025 letter).

That is strategically coherent. It’s what you want a founder to do. But it is optically terrifying to anyone allergic to the phrase “we’re choosing long-term over short-term.” On Wall Street, “long-term” is often code for “we missed the quarter.” Here, it looks like a genuine pivot to product quality.

3. The “Unhinged” Lever

Growth is still strong, but the story admits it’s controllable. They said DAU growth was slightly slower than Q2 partly because they posted less “unhinged” social content. Then, in their Q3 letter, they said they started posting the crazy stuff again and saw impressions rise.

Translation: Their brand-driven top-of-funnel is a lever, not a law of physics. They can literally turn a dial labeled “Chaos” to get more users.

4. The Optical Illusion of Profit

GAAP net income is not “clean” this quarter. If you glance at GAAP net income ($292.2 million, Q3 2025) and conclude “profitability just went to the moon,” you are misreading the map. That number included a one-time income tax benefit of $222.7 million from releasing a valuation allowance on deferred tax assets.

Stop & Think: This is an accounting event, not a cash event. If you strip that tax benefit out, the business looks different. Do not value this stock based on a P/E ratio calculated using that $292.2 million figure.

The Audit: The operating story is better told through operating income ($35.2 million, Q3 2025, GAAP), Adjusted EBITDA ($80.0 million, Q3 2025, non-GAAP), and cash flow. Focus on those.

The “AI can change what’s possible” push is simultaneously the growth upside and the source of your margin anxiety.

Atomic Take: The “break” isn’t demand—it’s optics. AI costs are pressuring gross margin while management openly chooses longer-term product bets over near-term neatness.

Falsifier: If Gross margin declines more than 100 bps YoY in a quarter while bookings growth decelerates (metrics: gross margin %, total bookings YoY %, company-reported), run for the hills.

The Rebound Catalysts

The good news is that Duolingo has a lot of controllable knobs, and they just told you exactly which ones they are turning.

Subscribe now

Catalyst 1: User Growth Momentum Persists at Scale

People are still showing up.

• FACT: DAUs 50.5 million (Q3 2025) and MAUs 135.3 million (Q3 2025).

• MONITOR: Watch if DAUs YoY stays in the 30s while MAUs keep compounding (Q3 2025 DAUs +36% YoY; MAUs +20% YoY). As long as the top of the funnel is wide, the machine works.

Catalyst 2: Monetization Outpaces the Funnel

They are getting better at squeezing juice from the orange.

• Subscription bookings $240.3 million (Q3 2025, USD), up 36% YoY; paid subscribers 11.5 million (period end Q3 2025).

• Management attributes a 7% YoY increase in subscription revenue per average paid subscriber (ARPU) to a mix shift toward higher-priced tiers including Max and the family plan (Q3 2025).

• Keep an eye out for: Watch paid subscriber penetration (currently 9.0% of LTM MAUs, Q3 2025). We want this trending up without DAU growth falling off a cliff.

Catalyst 3: Operating Leverage (The Magic Trick)

They are spending more on AI, but still expanding margins elsewhere. That is discipline.

• Adjusted EBITDA margin 29.5% (Q3 2025, non-GAAP) vs 24.7% (Q3 2024).

• FY 2025 Adjusted EBITDA margin guidance midpoint 29.0% (FY 2025, non-GAAP).

• Watch this, though: if GAAP operating expenses grow slower than revenue while product investment continues (Q3 2025 revenue $271.7 million; operating expenses $161.8 million, GAAP).

Catalyst 4: “Event-Driven” Growth (Q4 Seasonality)

New Year’s Resolutions are basically the Super Bowl for Duolingo.

• Management called out Year in Review and the New Year’s Promotion as high-impact Q4 events (Q3 2025 letter).

• But keep an eye on this: Q4 bookings and revenue land in the guided ranges (Q4 2025 bookings $329.5–$335.5 million; revenue $273–$277 million, guidance).

Atomic Take: The bull case doesn’t require miracles—it requires the boring continuation of DAU + subscriber compounding while AI costs stop widening as a percentage of revenue.

Falsifier: If FY 2025 revenue lands below $1,027.5 million or FY 2025 bookings land below $1,151 million (metrics: company guidance ranges), the thesis is broken.

Financial Quality Rubric (The Scorecard)

I like to grade my companies like a strict teacher. Here is the report card.

Growth Durability: 4/5

It is hard to grow fast when you are already big, but they are doing it.

• DAUs +36% YoY to 50.5 million (Q3 2025) while lapping Q3 2024 DAU growth of 54% YoY per management commentary. That is impressive resistance to gravity.

Monetization Quality: 4/5

• Subscription revenue $229.5 million (Q3 2025, GAAP), +46% YoY; subscription bookings $240.3 million (Q3 2025), +36% YoY.

Margin Structure: 3/5

• Gross margin 72.5% (Q3 2025, GAAP), down YoY due to generative AI + hosting costs.

Cash Conversion: 5/5

This is where the rubber meets the road.

• Free cash flow $77.4 million (Q3 2025) and free cash flow margin 28.5% (Q3 2025), company-defined non-GAAP.

Balance Sheet Resilience: 5/5

A fortress.

• Cash & cash equivalents $1,011.5 million and short-term investments $109.9 million (as of Sep 30, 2025, GAAP) with total liabilities $578.1 million (as of Sep 30, 2025). They have more cash than debt/liabilities.

Accounting/Definition Cleanliness: 3/5

A bit messy this quarter.

• Adjusted EBITDA definition expanded beginning in Q3 2025 to include integration costs related to acquisitions (non-GAAP comparability note).

• The Audit: The quarter’s GAAP profitability headline is radioactive with a smiley face—$222.7 million of one-time tax benefit (Q3 2025, GAAP) inflates net income, so “earnings quality” needs to be judged via operating income, Adjusted EBITDA, and cash flow instead.

Atomic Take: This is high-quality financially, but you must grade the core engine with operating metrics and cash—not the tax-boosted GAAP net income headline.

Falsifier: If Free cash flow margin drops below 20% for two consecutive quarters (metric: company-defined free cash flow margin), the cash machine is sputtering.

Share

The Statements

You can listen to the story management tells you on the conference call, or you can look at the numbers. I prefer the numbers. The numbers don’t have a PR team.

Balance Sheet (The Geiger Test)

As of Sep 30, 2025 (GAAP), Duolingo shows the kind of balance sheet you want when you’re funding product ambition.

It’s what I call a “sleep well at night” balance sheet:

• Cash & cash equivalents: $1,011.5 million (as of Sep 30, 2025, USD, GAAP).

• Short-term investments: $109.9 million (as of Sep 30, 2025, USD, GAAP).

• The Big Picture: Total assets $1,885.6 million against total liabilities $578.1 million (as of Sep 30, 2025, USD, GAAP).

• The Secret Weapon: Deferred revenue $441.7 million (as of Sep 30, 2025, USD, GAAP).

Remember, deferred revenue in a subscription model is an obligation, sure, you owe them the service, but it’s also a demand signal. It’s cash in the bank that hasn’t hit the P&L yet.

The only “don’t be asleep at the wheel” note here is the lease obligations. They have to rent office space somewhere. Operating lease liabilities are $97.3 million total. It’s manageable, but it’s there.

Atomic Take: The balance sheet looks like optionality—cash-rich, liability-light, and fueled by deferred revenue instead of debt.

Falsifier: If Cash & cash equivalents fall below $500 million while deferred revenue falls YoY (metrics: cash line; deferred revenue line, company-reported), the fortress is crumbling.

Cash Flow (The Turbine)

Profit is an opinion; cash is a fact. Here is the clean bridge, using the company’s own definition:

Profit → CFO → “capex-like” → Free cash flow

• Net cash provided by operating activities $84.2 million (Q3 2025, USD, GAAP).

• Also, “capex-like” investments were capitalized software + intangibles $(2.0) million and PPE $(4.9) million (Q3 2025, USD, GAAP), per the company’s free cash flow reconciliation.

• Free cash flow $77.4 million (Q3 2025, USD, company-defined non-GAAP) and free cash flow margin 28.5% (Q3 2025).

Stop & Think: A 28.5% free cash flow margin is elite. For every dollar that comes in the door, nearly 29 cents ends up in the owner’s pocket. That is rare.

Subscribe now

Now, the footnote that matters: 9M 2025 net income includes that one-time tax benefit we talked about earlier.

The cash flow statement shows a large deferred income tax adjustment $(239.5) million (9M ended Sep 30, 2025, USD, GAAP).

That isn’t “bad” cash flow. It’s just a reminder that tax accounting can torque the profit-to-cash bridge in weird ways.

The tax benefit was a paper gain; the adjustment removes it to show you the real cash.

Atomic Take: Cash generation looks sturdy, and the capex-like spend is modest relative to CFO—just don’t confuse a tax-accounting inflection with a permanent profit step-change.

Falsifier: If CFO declines YoY while deferred revenue also declines YoY (metrics: net cash from operating activities; deferred revenue, company-reported), the turbine is stalling.

Income Statement (The Reactor)

Q3 2025 (GAAP) is the cleanest snapshot of what the model can do at scale, as long as you keep your hands off that tax headline.

• Revenue $271.7 million, +41% YoY.

• Gross profit $196.9 million and gross margin 72.5%.

• Income from operations $35.2 million.

• Adjusted EBITDA $80.0 million and Adjusted EBITDA margin 29.5%.

And here is the cost story in plain English: Gross margin pressure is explicitly tied to generative AI costs (Max expansion) and hosting costs (Q3 2025).

They are paying for brains.

Operating expenses grew, but management frames it as headcount-driven investment, while still delivering operating leverage (Q3 2025).

Atomic Take: The income statement is doing the rare thing in consumer: scaling growth and scaling profitability at the same time—while admitting the AI cost line is real.

Falsifier: If Operating income turns negative while revenue growth remains above 30% YoY (metrics: income from operations; revenue YoY %, company-reported), they are losing control of the spend.

What about the price tag? (Valuation)

You don’t need a complex spreadsheet to figure this out. You just need the current stock price and the promises the company made for the year. Let’s look at what Mr. Market is asking for this business as of today (January, 2026).

• The Share Price: $173.87/share (as of January 8th 2026).

• The Slice of the Pie: Total estimated diluted shares 49.9 million (as of Sep 30, 2025, shares).

• The War Chest: Cash & cash equivalents $1,011.5 million + short-term investments $109.9 million (as of Sep 30, 2025, USD, GAAP). That is $1.12 billion in pure liquidity.

• The Promise (Guidance): FY 2025 revenue $1,027.5 million–$1,031.5 million and Adjusted EBITDA $296.9 million–$300.2 million.

What You’re Actually Paying:

• Implied Equity Value: ~$8.68 billion (Jan 9, 2026 price × 49.9M diluted shares).

• Implied Enterprise Value (EV): ~$7.55 billion (equity value minus that massive pile of net cash).

The Multiples (The “Expensive” Meter):

If we measure the price against the midpoint of what they promised for 2025:

• EV / FY 2025 Revenue: ~7.3x.

• EV / FY 2025 Adjusted EBITDA: ~25.3x (company-defined non-GAAP).

The Audit: This is a very different conversation than we were having just a few months ago. With the stock trading at $173.87, the massive “premium” (that extra charge you usually pay for elite growth) has compressed significantly.

Is it significant enough?

Well, think about it: The business hasn’t changed, but the price tag for the operating engine has been cut effectively in half compared to just a short time ago. At ~25x EBITDA, the market is no longer pricing this for “absolute perfection.” It’s pricing it for “just don’t break it.”

I like this, and that is why I own shares in it at the time of writing, but if this is reason enough for you to consider it? That is your own responsibility to determine.

Atomic Take: The valuation just moved from “priced for miracles” to “priced for execution.” That is where the opportunity usually hides.

Falsifier: If FY 2025 Adjusted EBITDA margin lands below the guided midpoint range (metric: FY 2025 Adjusted EBITDA margin, company-defined), even this lower multiple will feel expensive.

The Meltdown List (Risks)

Every great business has a few trapdoors. Here are the ones that could swallow your capital.

1. You Are Living on Rented Land

Duolingo doesn’t own the road; they just drive on it.

• Apple/Google/Stripe processed the majority of revenues.

• The Risk: If the landlord raises the rent (app store fees) or changes the rules, Duolingo pays the price.

• Keep an eye on this: Any disclosure change in processor concentration or changes in cost of revenues.

2. Brains Are Expensive (AI Costs)

• Management explicitly ties gross margin pressure to generative AI + hosting

• The Risk: AI is the new electricity. If the cost of computing (Max expansion) rises faster than the price they can charge for it, margins get crunched.

• Keep watching this: Gross margin trend; Max expansion versus margin.

3. The Fickle God of Engagement

• DAUs and MAUs are the engine (Q3 2025 DAUs 50.5 million, MAUs 135.3 million).

• The Risk: If people stop opening the app, the whole machine stops. There is no contract locking these users in.

• And watch this: DAU growth and the DAU/MAU (derived) stability.

4. The Dollar Diet (FX Risk)

• Over half of bookings come from outside the U.S.

• The Math: Management estimates each 1% change in the dollar versus their currency basket is about a $1 million impact on bookings for the remainder of the year (Q3 2025 letter).

• Watch: Reported vs constant currency growth deltas.

5. Optical Illusions (Bookings vs. Revenue)

• Deferred revenue $441.7 million (as of Sep 30, 2025) and large “deferral of revenue” flows in 9M 2025.

• The Risk: If bookings slow down, revenue might look fine for a while because of the backlog. Don’t be fooled.

• Watch: Deferred revenue growth and the bookings trajectory.

6. The Dilution Drip

• Stock-based compensation expense $101.2 million (9M ended Sep 30, 2025, USD, GAAP).

• The Outlook: Management expects fully diluted share count to increase about 1% in 2025 (assuming Q3 quarter-end share price).

• Keep an eye on this: Diluted share count and SBC as a % of revenue.

7. Moving the Goalposts?

• Adjusted EBITDA definition expanded beginning in Q3 2025 to include integration costs related to acquisitions.

• The Risk: Whenever a company changes how they calculate “profit,” you should squint a little harder.

• And watch: Further changes to add-backs and reconciliation stability.

8. Sneaky Risk: Live by the Meme, Die by the Meme

• Management explicitly described dialing back “unhinged” content and seeing growth effects, then resuming (Q3 2025 letter).

• HYPOTHESIS: If brand/funnel tactics become a major driver, growth may get choppier than the market wants to tolerate.

Atomic Take: The risks are mostly about control: platform tolls, AI input costs, and whether engagement remains reliably compounding.

Falsifier: If Processor concentration rises above current levels while gross margin declines YoY (metrics: revenue processor %; gross margin %, company-reported), the business is losing leverage.

Subscribe now

The Atomic Verdict

Status: ★★★★☆ (4.25/5)

I like this business. I respect the hustle. And I like the recent price drop. That does not mean it wont’t continue dropping, btw. As Peter Lynch said; “How much lower can it get?”. Well, sentiment and narrative are important short term drivers.

Reason 1 (The Good Kind of Boring)

The engagement engine is still compounding at scale—50.5 million DAUs and 135.3 million MAUs (Q3 2025), with 11.5 million paid subscribers (period end). That is a fortress.

Reason 2 (The Money Part)

The model is throwing off real cash—free cash flow $77.4 million and free cash flow margin 28.5% (Q3 2025), company-defined.

Atomic Take: Duolingo looks like a rare consumer compounder with legitimate cash output; the only “don’t get cocky” clause is that the valuation assumes KPI perfection.

Upgrade Triggers

I’m pulling the trigger if I see:

1. Paid subscriber penetration rises above 9.0% of LTM MAUs (metric: company-reported penetration).

2. Gross margin stabilizes or improves sequentially while Max continues expanding (metrics: gross margin %, management commentary on AI costs).

3. FY 2025 results land at or above the midpoint of Adjusted EBITDA margin guidance (29.0%, FY 2025, non-GAAP).

Downgrade Triggers

I’m selling if:

1. DAUs drop below 30% YoY growth for two consecutive quarters (metric: DAUs YoY %, company-reported).

2. Gross margin declines another 100+ bps YoY while subscription bookings growth decelerates (metrics: gross margin %; subscription bookings YoY %).

3. Free cash flow margin falls below 20% for two consecutive quarters (metric: company-defined free cash flow margin).

What Would Increase Confidence?

I need evidence that AI-driven features raise retention/ARPU without continuing to pressure gross margin (watch: subscription revenue per average paid subscriber commentary; gross margin).

I also need to see consistent delivery against Q4/FY 2025 booking and revenue guidance ranges (watch: Q4 and FY 2025 guidance versus actuals).

Final Falsifier: If FY 2025 bookings land below $1,151 million while FY 2025 Adjusted EBITDA margin lands below the guided midpoint (metrics: company guidance ranges; actuals), the thesis is invalid.

Duolingo has managed to do what very few apps do: become a utility that feels like a game. The financials back it up: cash rich, growing fast, and profitable.

Now, a little task for you: Go to the comments and answer me this:

Are you worried about the AI costs eating into the gross margin, or do you view that as the necessary cost of building a better moat?

References

• Duolingo, Inc. — Form 10-Q for the quarter ended September 30, 2025 (SEC filing; file: “0001628280-25-049743.pdf”).

• Duolingo, Inc. — Q3 FY2025 Shareholder Letter (as of September 30, 2025; file: “Q3FY25 Duolingo 9-30-25 Shareholder Letter Final.pdf”).

• Duolingo, Inc. — Q3 FY2025 Press Release (as of September 30, 2025; file: “Q3FY25 Duolingo 9-30-25 Press Release.pdf”).

• Duolingo model support file (provided spreadsheet; file: “DUOL-Duolingo Inc.xlsx”).

• Market price reference (used only for the updated valuation section): DUOL last price on January 9, 2026 from the finance quote tool.

Disclaimer

This Deep Dive is an educational breakdown of a public company based on information available in the materials provided (e.g., annual/quarterly reports, investor presentations, earnings transcripts) and my interpretation of that information. It is designed to be a “bolt-on” intelligence layer to your own due diligence — not a replacement for it.

Independence: I do not accept compensation of any kind from the companies discussed. My research is driven solely by my personal search for high-quality compounders.

Skin in the Game: Unless otherwise stated, assume the author may hold a long position in securities mentioned. Any position creates bias — treat this as commentary, not gospel.

Not Financial Advice: Nothing here is investment advice, a recommendation, or a solicitation. I am not a financial advisor. You are responsible for your own decisions. The stars rating is not a buy recommendation, but meant as a guide to understand the quality of the financial statement of the respective companies.

Error & Update Risk: Financial statements change, companies restate, guidance evolves, and I can be wrong. Verify key figures in the primary filings and consider reading the footnotes before deploying capital.

New to The Atomic Moat? This analysis of Sea Ltd ($SE) is a prime example of how we dissect high-quality compounders. If you want these deep dives sent to your inbox, join 900+ other investors below.

Join for Free

For a long time, Sea was the poster child for “Profitless Tech.” It was the stock you owned in 2021 because your neighbor did, and the stock you sold in 2022 because you were tired of hearing about “adjusted” metrics while the cash burned.

But a funny thing happened while we were looking away.

The “mess” turned into a machine.

If You’re In a Rush

• What they do: Sea Limited is three companies wearing one trench coat. You’ve got Shopee (the shopping mall), Monee (the bank), and Garena (the arcade).

• The Headache: Wall Street hates complexity. This stock is three different underwriting problems wrapped in one ticker: retail wars (Shopee), credit risk (Monee), and the rollercoaster of gaming cycles (Garena). Plus, the accounting optics make people argue over what’s actually “real” profit.

• The Fix: The numbers are finally shutting up the critics. Shopee is cranking up monetization without scaring away shoppers. The bank (DFS) is growing its loan book while keeping the credit quality boring (boring is good in banking). And Garena? It’s surging again, proving it’s not just a “one-hit wonder.”

• Atomic Position: High Conviction (with caveats). I like it, but we have to watch the credit quality like a hawk. If the bank starts making bad loans, the whole story changes.

While the market was busy arguing about interest rates, Sea quietly tightened the screws. They stopped lighting money on fire to chase growth and started demanding that the business actually work. And in Q3 2025, the scoreboard lit up in a way that is impossible to ignore.

This isn’t a “hope” trade anymore. This is a company that just posted nearly $6 billion in revenue and almost $900 million in EBITDA in a single quarter.

Why are we digging in now?

Because the market is still treating this like a disjointed collection of businesses—a retailer, a bank, and a game studio glued together by duct tape. They charge it a “complexity tax.”

But what if the complexity isn’t a bug? What if it’s the moat?

We are going to take this thing apart, piece by piece. We’re going to look at the Shopee demand engine, the DFS credit book (the scary part), and the Garena cash cannon. We aren’t going to rely on Wall Street jargon or “adjusted” fluff. We are going to look at the cash, the loans, and the users.

Put the coffee on. It’s time to move Sea out of the “Too Hard” pile and onto the operating table.

Subscribe now

What they do:

Sea Limited is basically three companies wearing one trench coat. You’ve got Shopee (the shopping mall), Monee (the bank), and Garena (the arcade).

The Headache: Wall Street hates complexity. This stock is three different underwriting problems wrapped in one ticker: retail competition (Shopee), credit risk (Monee), and the rollercoaster of gaming cycles (Garena). Plus, the accounting optics make people argue over what’s actually “real” profit.

The Fix: The numbers are starting to shut up the critics. Shopee is cranking up monetization without scaring away shoppers. The bank (DFS) is growing its loan book while keeping the credit quality boring (boring is good in banking). And Garena? It’s surging again, proving it’s not just a “one-hit wonder.”

Atomic Position: High Conviction (with caveats). I like it, but we have to watch the credit quality like a hawk. If the bank starts making bad loans, the whole story changes.

The Setup: No More Hand-Waving

For a long time, the debate on Sea Limited was based on vibes and promises. That era is over. Sea’s Q3 2025 print is the kind of report that forces the market to put down the coffee and look at the scoreboard.

The machine is printing money. Real money. At scale.

The Scoreboard:

• Revenue: $5,986.0m

• Operating Income: $475.9m

• Net Income: $375.0m

• Total Adjusted EBITDA: $874.3m

Stop & Think: Just a few years ago, people were asking if this business model could ever turn a profit. Now they just posted nearly $900 million in adjusted EBITDA in ninety days. The conversation has shifted.

These four numbers matter because they change the debate. We aren’t asking “Can this work?” anymore. We are asking “What exactly are we buying here?”

Three Engines, One Quarter

Here is the variant perception—the difference between what the Street sees and what I see.

The Market View: Sea is a “conglomerate discount” waiting to happen. It’s hard to value because the risk profile shifts every quarter. Is it a tech platform? A lender? A gaming studio? It gives analysts a headache.

The Bull View (My view at the time of writing): The bundle isn’t the problem; the bundle is the moat. Shopee brings the people in. Monee (DFS) lends to them. Garena prints the cash to fund the growth. It’s a self-feeding machine.

The Bear View: The bundle is a trap. Bears say the shipping subsidies never really go away (a permanent tax), Garena is just a cyclical boom-bust business, and eventually, the “Digital Financial Services” (DFS) arm gets so big that the stock trades like a risky bank book rather than a tech stock.

The “Why Now”: I’m looking at three dials on the dashboard, and for the first time in a while, they all turned green at the exact same moment.

Dial 1: The Mall (Commerce) The mall is getting bigger. Shopee GMV hit $32.2bn and gross orders clocked in at 3.6bn. The demand engine isn’t slowing down; it’s accelerating.

Subscribe now

Dial 2: The Bank (DFS) The loan book is scaling up. Consumer + SME loans principal outstanding reached $7.9bn (at end-Sept). And here is the “Geiger counter” for risk: NPL90+ (non-performing loans over 90 days) was disclosed at 1.1%. As long as that number stays low, we’re in business.

Dial 3: The Arcade (Garena) When this hits, it hits hard. Bookings came in at $840.7m. This is your margin cannon. It prints cash that hides a lot of sins elsewhere.

The Duration Question

This is starting to look like a Nick Sleep / Nomad-style case study in “duration.” Duration isn’t just a fancy finance word; it’s a practical question: How long can this system keep compounding before one part breaks?

If this were just one business, it would be easy. But Sea asks you to bet on a system. Systems can grow brutally fast, but they can also fail in weird, non-linear ways. That’s why the market charges them a “complexity tax.”

FACT: Segment revenue was $4,294.8m (e-commerce), $989.9m (DFS), and $653.0m (digital entertainment).

HYPOTHESIS: If Shopee can charge more (monetization) while the DFS credit book stays boring, that “complexity tax” disappears and the stock re-rates higher.

MONITOR: Shopee take rate; DFS NPL90+; Garena bookings.

How the Business Actually Makes Money

Forget the segments for a second. Let’s pop the hood. You understand this company by looking at the “knobs” management can turn.

• Shopee’s knobs: Volume, conversion, and ads (this is the big one).

• DFS’s knobs: Who do we lend to? How much cash do we have?

• Garena’s knobs: Are people playing? Are they paying?

1. Shopee: The Optics vs. The Reality

Management breaks e-commerce revenue into marketplace revenue ($3.8bn)—which is split into core marketplace ($3.1bn) and value-added services ($0.7bn).

Now, catch this detail: They explicitly said value-added services (mostly logistics) declined year-on-year. Why? Because of increased shipping subsidies.

Here is the plain English translation: They are eating the shipping cost to keep you happy. They are taking a bruise on the “optics” to keep the competitive edge sharp.

Is it worth it? The answer lies in monetization, specifically ads.

Sea told us ad revenue grew “over 70%” year-on-year and the ad take-rate improved by “more than 80 bps.” They also disclosed a total revenue take rate of 13.3%.

This is the classic playbook: Build density, get everyone addicted to the service, and then sell the high-margin ads. But—and this is a big but—it only works if the product actually gets better.

Management said ShopeeMall GMV more than doubled year-on-year in Brazil and delivery speeds improved by two days. That tells me the quality of demand is getting better. We aren’t just attracting bargain hunters; we’re attracting brand shoppers.

The Atomic Take: Shopee can afford ugly accounting optics (subsidies) as long as the monetization curve (ads) keeps bending upward faster than the costs.

Falsifier: If the take rate stalls out at 13.3% but they keep complaining about shipping subsidies, run for the hills. That means the moat is leaking.

Subscribe now

2. DFS (Monee): The Silent Giant

This is the segment that keeps me up at night—in both good and bad ways. It deserves your attention.

DFS Revenue: $989.9m DFS Adjusted EBITDA: $258.3m

The loan book is $7.9bn (split $6.9bn on-book and $0.9bn off-book). The NPL90+ is 1.1%.

Those are clean numbers. But here’s the trap: Growth is fine until it isn’t. Management bragged about adding 5 million first-time borrowers and hitting 34m active users.

In the lending business, growth signals are great, right up until the moment they become “underwriting errors.”

DFS is the best business in the bundle right now because it monetizes the ecosystem. But if credit quality drifts—if that 1.1% NPL starts creeping up to 2% or 3%—the market will stop viewing Sea as a tech platform and start viewing it like a risky subprime lender.

Atomic Take: Watch the NPL line. If it stays boring, the stock works. If it moves, the narrative breaks.

3. Garena: The Swing Factor

Garena is the “Quality Upside.”

Bookings: $840.7m Adjusted EBITDA: $465.9m Paying Ratio: 9.8% ARPPU: $12.8

Management called this the “best quarter since 2021” and guided for 30% year-on-year bookings growth for 2025.

Here is the thing about gaming: It’s volatile. “Best since 2021” reminds us that there were some bad times between 2021 and now.

The constraint here is repeatability. Can they keep the paying ratio at 9.8% and ARPPU at $12.8? If those numbers hold up, Garena is a cash machine that funds the rest of the war. If the payers leave, Garena becomes a drag.

Atomic Take: Don’t just look at the bookings number. Look at the payer behavior. We want steady addicts, not just a one-time splash.

Share

What Went Wrong (Or: Why the Market Gets Indigestion)

Let’s be clear: Sea didn’t crash the car. “What went wrong” isn’t about an operational disaster. It’s about a pattern of making smart business moves that look ugly on a spreadsheet.

The market hates ugly spreadsheets. It responds to “annoying optics” by demanding a discount.

1. The “Optics Bruise” on Purpose First off, Shopee is taking a punch to the face voluntarily. They explicitly tied the year-on-year decline in “value-added services revenue” to increased shipping subsidies.

Think of it like a restaurant giving away free appetizers to get you to buy the steak. The “appetizer revenue” looks terrible, but the restaurant is packed.

• The Optic: “Value-added services revenue is down! Quality is deteriorating!”

• The Reality: They are subsidizing shipping to drive throughput.

• The Result: Investors spend all day arguing about the direction of the business when they should be arguing about the trade-off.

2. The “Gap” Where the Bears Live When you hand the skeptics a stick, don’t be surprised if they hit you with it. Sea’s earnings framing gives the bears plenty of wood.

Look at the spread:

• Operating Income (GAAP): $475.9m

• Total Adjusted EBITDA: $874.3m

Stop & Think: That is a massive gap. Almost double. When the “real” number and the “adjusted” number are that far apart, people get nervous.

The bridge between those two numbers isn’t trivial. It’s filled with heavy hitters:

• D&A: $92.2m

• Share-based Compensation (SBC): $154.2m

• Deferred Revenue Changes (Net Effect): $151.9m

Even if every single one of those adjustments is 100% legitimate, a bridge that big makes “profitability” feel like a debate about definitions instead of a debate about economics. In a bundled model, you don’t just underwrite the results; you underwrite what the management wants you to look at. If Adjusted EBITDA is the headline and the bridge is the footnote, the skeptics will assume the footnote is where the bodies are buried.

3. The Three-Headed Monster Problem The bundle creates a permanent identity crisis. Sea is asking you to believe three impossible things before breakfast:

1. Commerce can scale while monetizing.

2. DFS (the bank) can scale while keeping credit boring.

3. Garena can contribute cash without turning into a volatile cycle trap.

When those three beliefs sit under one ticker, the market doesn’t average the risks. It prices the scariest one. If credit looks shaky, the whole company is priced like a bad bank. If gaming slows, the whole company is priced like a washed-up studio.

Atomic Take: Sea can execute perfectly and still get punished because complexity creates three different ways for confidence to break. The market prices the break, not the base case.

Falsifier: If GAAP operating income trends down materially from $475.9m while that EBITDA bridge stays huge, the “profitability” story falls apart.

Subscribe now

The Catalysts (The Watch List)

You don’t need a crystal ball here. Sea gives you the gauges. You just have to watch them.

Catalyst 1: Shopee Monetization (The Cash Register)

• The Test: Can they keep charging more?

• The Baseline: Take rate is 13.3%. Ad revenue grew “over 70%” year-on-year. Ad take-rate improved “more than 80 bps.”

Catalyst 2: Quality Control

• The Test: Are people happy with the service?

• The Baseline: In Brazil, delivery speed improved by about two days year-on-year. ShopeeMall GMV (the branded stuff) more than doubled year-on-year.

Catalyst 3: The Boring Bank

• The Test: Can DFS get bigger without blowing up?

• The Baseline: Loans at $7.9bn. NPL90+ at 1.1%. (Boring is beautiful).

Catalyst 4: Garena’s Legs

• The Test: Is the game still fun?

• The Baseline: Bookings $840.7m, payer ratio 9.8%, ARPPU $12.8. Management reiterated expectations for more than 30% bookings growth for 2025.

FACT: Management expects FY2025 Shopee GMV growth to be more than 25% and 2025 Garena bookings growth to be more than 30%.

HYPOTHESIS: If Shopee grows that fast while monetization improves and the credit book stays clean, the “complexity discount” shrinks.

MONITOR: The three dials (Monetization, Credit Quality, Bookings). These change the lens.

Falsifier: If NPL90+ drifts materially above 1.1% while they keep bragging about loan growth, run.

Financial Quality Rubric (1–5)

Here is the scorecard. No fluff.

Scale + Growth Durability: 4/5

Shopee moving $32.2bn in GMV and 3.6bn gross orders is serious volume. Consolidated revenue is knocking on the door of $6bn. This isn’t a startup anymore; it’s a giant.

Profitability Under Reinvestment: 4/5

Posting $475.9m in GAAP operating income while eating the cost of shipping subsidies is impressive. It proves the model works even when they are spending to fight.

Cash Generation Quality: 3/5

This one is tricky. Read carefully. Operating cash flow (9M) was $3,548.1m. Looks great. But you have to look at where the cash went. Sea explicitly says the investing outflow was driven by a $3,355m increase in loans receivable.

Translation: They aren’t burning cash on fancy offices; they are “spending” cash to lend it out to customers. That’s inventory for a bank. It’s different from lighting money on fire, but it still eats cash.

Balance Sheet Simplicity: 3/5

It’s messy. You have $3,067.8m in cash plus restricted cash (current $1,963.2m / non-current $35.8m). But you also have the banking plumbing: $3,484.2m in deposits payable and $2,775.7m in escrow payables. It’s a real bank balance sheet now.

Credit Discipline Risk: 4/5

1.1% NPL90+ on a $7.9bn book is a clean baseline. The only reason it’s not a 5/5 is because they are growing the book so fast. Rapid growth is the enemy of underwriting.

Accounting / Optics Noise: 3/5

The “Gap” we talked about. The Adjusted EBITDA bridge has big numbers ($154.2m SBC, $151.9m deferred revenue). Plus, the “unallocated expenses” line invites people to cherry-pick the narrative.

Atomic Take: Sea is high quality, but it is not simple. The market charges a “simplicity tax.” Sea has to pay that tax until they prove they can do this for four quarters in a row.

Falsifier: If Operating Cash Flow trends down materially from $3,548.1m (9M) while loan growth is still the main driver and credit quality gets worse, the engine is stalling.

Share

The Statements

Alright, it’s time to put on the reading glasses. We need to separate what’s liquid from what’s plumbing.

1. Balance Sheet (The Geiger Test)

In a normal retail business, you look at the cash pile and call it a day. But Sea isn’t normal. Sea reports cash & cash equivalents of $3,067.8m and restricted cash totaling $1,999.0m (current + non-current).

If this were just an online store, that would be a fortress. But it’s not just a store; it’s a bank. And banks have plumbing. Current liabilities include deposits payable of $3,484.2m and escrow payables/advances from customers of $2,775.7m.

Stop & Think: Those liabilities aren’t debts in the traditional sense. They are the plumbing of a financial system. But they mean you can’t just look at the cash and say “Wow, look at all that spending money.”

The Audit: Investors love to treat “cash” like a single bucket. But DFS turns cash into a labeled ecosystem. Some cash is for fighting wars (operating flexibility), and some cash is just sitting there because regulators or customers insist it sits there.

Atomic Take: The balance sheet is strong enough to support compounding, but only if the plumbing stays boring. In banking, “boring” is the highest compliment. It means stable credit and stable funding.

Falsifier: If deposits/escrow liabilities rise materially but profitability tanks and credit quality rots, the plumbing is backing up.

2. Cash Flow (The Turbine)

This is where the amateur mistakes happen. Operating cash flow for the first 9 months was $3,548.1m. Investing cash flow was a negative ($3,255.1m). Financing cash flow was $617.6m. The net increase in cash was $985.2m, ending at $5,066.8m.

The “cash burn” debate ends right here. The turbine is spinning. The operating cash is strong. But wait, why is investing cash flow so negative? Is management on a shopping spree?

No. Sea explicitly says the investing outflow was primarily driven by an increase in loans receivable of $3,355m. PPE purchases (actual capex) were only $351m. (There was also $508m of net proceeds from treasury items).

Here is the translation: The company is generating cash from operations, and then choosing to recycle that cash into loan growth. The common error is to see “Investing Outflow” and scream “Capex Binge!” That’s wrong. The biggest absorber is loans receivable. That is DFS scaling up. It’s not spending; it’s allocating capital to inventory (loans).

Atomic Take: The turbine is working. The only debate is whether the power is being routed into compounding (good loans) or risk (bad loans).

Falsifier: If CFO (Cash Flow from Operations) stays high but NPL90+ rots while the loan book keeps growing, then we have a problem. Strong cash flow paired with bad underwriting is a trap.

3. Income Statement (The Reactor)

On the surface, it looks clean:

• Revenue: $5,986.0m

• Operating Income: $475.9m

• Net Income: $375.0m

But check the bridge. Total Adjusted EBITDA was $874.3m. How do you get from $475m to $874m? You add back the “adjustments.”

• D&A: $92.2m

• SBC (Share-Based Comp): $154.2m

• Deferred Revenue Effects: $151.9m

The Audit: In a multi-engine business like this, adjustments can become a crutch. If you only read the headlines, you see the EBITDA. If you read the footnotes, you see the bridge. The reconciliation is your checklist.

Atomic Take: The reactor is running hot. The argument is just about how much of the glow is structural profit and how much is accounting adjustments.

Falsifier: If GAAP operating income declines materially from $475.9m while adjusted EBITDA stays high because the bridge items got bigger, the “quality” of the earnings is degrading.

Subscribe now

Valuation: What Are We Paying For?

At the time of this writing, Sea’s market cap is around $75.5bn.

Let’s do some napkin math using Q3 as a run-rate (arithmetic, not a forecast):

• Net Income Run-Rate: $375.0m × 4 = $1.5bn.

  • Implies ~50x equity value / run-rate net income.

• Adjusted EBITDA Run-Rate: $874.3m × 4 ≈ $3.5bn.

  • Implies ~22x equity value / run-rate adjusted EBITDA.

This is “Duration Math.” You aren’t paying 50x earnings for what they did yesterday. You are paying for the belief that the flywheel will spin for a long, long time.

For that multiple to make sense, three things must be true at once:

1. Shopee keeps monetizing without killing demand.

2. DFS keeps the credit boring while getting bigger.

3. Garena proves it’s not a one-hit wonder.

Atomic Take: At $75.5bn, you are not paying for “one good quarter.” You are paying for synchronized durability.

Falsifier: NPL90+ moves materially above 1.1% while loans principal outstanding keeps growing from $7.9bn AND GAAP operating income trends down from $475.9m. If that happens, the duration assumption breaks, and the multiple collapses.

Risks (The Meltdown List)

I keep this list taped to my monitor.

1. Credit Cycle Risk (The Fast Repricer)

   • Gauge: NPL90+ vs loans principal outstanding (baseline: 1.1% on $7.9bn).

2. Monetization Overshoot (The Tax Man)

   • Gauge: Take rate (13.3%) vs GMV ($32.2bn) and gross orders (3.6bn). If they tax the ecosystem too hard, the users leave.

3. Subsidy Creep (The Bad Habit)

   • Gauge: Continued shipping-subsidy talk without offsetting improvements in ads momentum.

4. Garena Cyclicality (The Fade)

   • Gauge: Bookings ($840.7m) vs payer ratio (9.8%) and ARPPU ($12.8).

5. Definitions Risk (The Bridge)

   • Gauge: Watch the gap between Operating Income ($475.9m) and Adjusted EBITDA ($874.3m). If the bridge grows, the quality drops.

6. Bundle Correlation Risk

   • The Nightmare Scenario: Shopee slows down, people stop paying back loans, and gamers stop playing, all at once.

The Atomic Verdict

Status: High Conviction (with caveats). Rating: 4.1 / 5.0 Stars.

Why it earns the rating: The bundle is finally profitable at scale.

• Revenue: $5,986.0m

• Operating Income: $475.9m

• Net Income: $375.0m

• Total Adjusted EBITDA: $874.3m

Shopee is showing it can monetize (13.3% take rate) while still fighting for market share. DFS is getting huge ($7.9bn loan book) while behaving responsibly (1.1% NPL).

What would increase my confidence? Give me two or three more quarters of this. If the loan book keeps growing without the NPLs spiking, and if Shopee keeps the ads machine humming, this stock goes higher.

What would make me wrong fastest? If NPL90+ moves materially above 1.1% while they keep pumping the loan book. That is the classic bank failure recipe.

Upgrade Triggers:

• NPL90+ stays around 1.1% while loans scale past $7.9bn.

• Shopee grows GMV past $32.2bn and orders past 3.6bn while keeping the take rate up.

• Garena keeps bookings near $840.7m with stable payer ratios (9.8%).

Downgrade Triggers:

• NPL90+ rises materially above 1.1%.

• Operating Income trends down from $475.9m while the “Adjusted” numbers stay high (fake profit).

• Shipping subsidies keep rising, but ad revenue slows down.

I’ve laid out the case. It’s a complex machine, but a profitable one.

Question for the comments: Do you trust a tech company to run a bank? Or is the lending arm a ticking time bomb inside the portfolio?

Reference Notes

Core quarter scoreboard, segments, KPIs

• Sea Third Quarter 2025 Results (press release PDF), p.2: revenue 5,986.0m; operating income 475.9m; net income 375.0m; total adjusted EBITDA 874.3m; Shopee GMV 32.2bn; Garena bookings 840.7m; loans 7.9bn; NPL90+ 1.1%.

• Sea Third Quarter 2025 Results Deck (PDF), p.7: segment revenue (4,294.8m / 989.9m / 653.0m) and segment profitability tables.

• Sea Third Quarter 2025 Results Deck (PDF), p.10: Shopee GMV 32.2bn and gross orders 3.6bn chart.

• Sea Third Quarter 2025 Results Deck (PDF), p.9: ads growth language (“over 70%”), ad take-rate improvement (“more than 80 bps”), and Brazil delivery speed / ShopeeMall mix highlights as presented.

• Sea Third Quarter 2025 Results Deck (PDF), p.16: Garena bookings 840.7m; paying users 65.9m; paying ratio 9.8%; ARPPU 12.8; adjusted EBITDA 465.9m.

• Sea Q3 2025 earnings call transcript: e-commerce revenue component commentary and subsidy explanation; management narrative on momentum and guidance language.

Cash flow and investing drivers

• Sea Third Quarter 2025 Results report (PDF), p.16: CFO 3,548.1m; investing (3,255.1m); financing 617.6m; net increase 985.2m; ending cash+CE+restricted 5,066.8m; investing drivers (loans receivable +3,355m; PPE 351m; net proceeds 508m).

Balance sheet plumbing

• Sea Third Quarter 2025 Results report (PDF), p.14: deposits payable 3,484.2m; escrow payables/advances 2,775.7m; cash and restricted cash labels (as of Sept 30, 2025).

Adjusted EBITDA reconciliation bridge

• Sea Third Quarter 2025 Results report (PDF), reconciliation page: D&A 92.2m; SBC 154.2m; deferred revenue effect 151.9m; operating income to adjusted EBITDA bridge.

Disclaimer

This Deep Dive is an educational breakdown of a public company based on information available in the materials provided (e.g., annual/quarterly reports, investor presentations, earnings transcripts) and my interpretation of that information. It is designed to be a “bolt-on” intelligence layer to your own due diligence — not a replacement for it.

Independence: I do not accept compensation of any kind from the companies discussed. My research is driven solely by my personal search for high-quality compounders.

Skin in the Game: Unless otherwise stated, assume the author may hold a long position in securities mentioned. Any position creates bias — treat this as commentary, not gospel.

Not Financial Advice: Nothing here is investment advice, a recommendation, or a solicitation. I am not a financial advisor. You are responsible for your own decisions.

Error & Update Risk: Financial statements change, companies restate, guidance evolves, and I can be wrong. Verify key figures in the primary filings and consider reading the footnotes before deploying capital.

New to The Atomic Moat? This analysis of MercadoLibre ($MELI) is a prime example of how we dissect high-quality compounders. If you want these deep dives sent to your inbox, join 900+ other investors below.

Join for Free

The 30-Second Elevator Pitch

What They Actually Do

Think of MercadoLibre as the central nervous system for Latin American commerce. They aren’t just selling stuff. They run the marketplace, the trucks delivering the boxes, the ads you see, the payments for the goods, and the credit card you swipe to buy them. They’re trying to spin all those plates at once into one giant, compounding flywheel.

The Headache (The Challenges)

Now, the results are real, but looking at them can be messy.

1. Margin Squeeze: They’re offering free shipping like it’s going out of style, which hurts profits in the short run.

2. The Banker’s Worry: Their credit business is growing fast, which always makes Wall Street whisper, “What happens if people stop paying?”

3. The Argentina Factor: It adds a layer of what I call “accounting radiation”—hyperinflation math and currency noise that makes the spreadsheets look radioactive.

The Fix The solution is simple but hard: Keep the volume pedal to the floor while keeping the loans boring and the trucks full. If logistics get cheaper per package, the math works. Basically, keep the flywheel spinning without the loan book turning into a horror movie.

My Position: High Conviction. But let’s be honest, it’s heavy on the “overhang” if the credit quality slips or Argentina acts up.

Subscribe now

The Setup: Q3 2025

MercadoLibre’s message for Q3 2025 wasn’t exactly a whisper. It was a shout: The machine is accelerating again.

Let’s look at the scoreboard:

• Net Revenues & Financial Income: $7,409m

• Income from Operations: $724m (That’s a 9.8% operating margin).

• Net Income: $421m (That’s a 5.7% net income margin).

Why is this happening now? Three dials are turning to the right at the same time:

1. Commerce: People are buying more stuff. GMV hits $16.5bn.

2. Fintech: Money is moving faster. TPV hits $71.2bn.

3. Credit: The loan book is getting thicker. Credit portfolio hits $11.0bn.

The Market’s Argument

The market isn’t arguing about whether this is a good business. Everyone knows it is. The argument is: “What kind of risk is this?” If you’re just a platform connecting buyers and sellers, you get one price tag. If you’re a platform plus a lender, you get a cheaper one.

Atomic Take: Q3 2025 looks like great execution on the ground, but the stock’s story is still stuck in the “Credit + Argentina + Optics” narrative.

The Falsifier (How we know I’m wrong): If Consolidated Net Revenues & Financial Income growth falls below 30% YoY for two quarters in a row, the thesis is broken. (Metric: Company-reported YoY % on the same line item).

How This Business Actually Makes Money

I like to think of MercadoLibre as a two-core reactor with a “third rail” that powers the whole grid.

• Core 1: Commerce (The Marketplace + Logistics + Ads).

• Core 2: Fintech (Mercado Pago payments, acquiring, + Credit).

• The Third Rail: Integration. Each part exists to make the other part necessary.

The Logistics Gamble

On the commerce side, management is willing to “spend” money on shipping to make the market bigger. It’s counter-intuitive, but effective. They accelerated growth “despite a major increase in free shipping and lower seller shipping charges, which reduced shipping revenue” (Source: Q3 2025 shareholder letter).

Stop and think for a second. Most companies try to charge you more for shipping to pad the bottom line. MELI is charging less. Why?

This isn’t an accident. It’s a strategic knob they are turning. Lots of companies talk about “ecosystems” like it’s a buzzword. MELI actually pays for theirs. That free shipping isn’t a promotion or a coupon; it’s Capex in disguise. Except instead of building a factory, they are building a habit. The question isn’t “Did it hurt margins this quarter?” The question is “Did it make MercadoLibre the default choice for a family in Brazil?”

The Brazil Case Study

Brazil is the cleanest example of this. Management highlighted reducing the free shipping threshold from R$79 to R$19. That is a massive drop. Here is what you hope happens next (and what they say happened):

• Conversion rates went up.

• Retention went up.

• Suppliers started adding more cheap items (the newly subsidized band).

• The network swallowed the higher volume without breaking a sweat.