AppLovin

IThe business

AppLovin runs a market with two sides. App makers, mostly mobile game studios and since 2024 also online shops, pay AppLovin to find new users or customers by showing ads inside other apps. Other app makers, the publishers, offer the ad space in their apps for sale through MAX, AppLovin’s auction tool, and AppLovin pays them their share.⁶ AppLovin books as revenue what advertisers pay, less what it pays publishers.⁵ That matters when I compare it with other companies: its 84% adjusted EBITDA margin is on this net revenue, and Meta’s margin is on gross ad revenue.

It has one reportable segment. Four products make it up. The first is the ad engine: AppLovin Ads, called Axon Ads Manager in the 10-K and the first-quarter 10-Q, which the 10-Q says is “the vast majority of revenue.” The other three are MAX, Adjust (measurement software sold by subscription) and Wurl (connected-TV distribution); the 10-Q calls revenue from services other than the ad engine “not material.”⁵¹³⁶ So MAX matters as a position, because it lets AppLovin see the auctions, and not as a source of revenue. The company calls its ad models Axon. Advertisers set the return they want on their spend, and the system finds the users who meet it, charging by the result and not by the click.⁶

The company used to own game studios too. It sold them to Tripledot on 30 June 2025 for $715.6M in total: $430.6M in cash and 596.9M Tripledot shares it valued at $285.0M, about 20% of that company.⁵¹⁴ It took a $188.9M goodwill write-down on the studios before the sale.³ Tripledot now buys ads from AppLovin, as any studio does: that was $44.7M of net revenue in the first nine months of 2026, under 1% of the total.⁵ The stake sits on the balance sheet at $298M and I do not count it in my values.³

It is a small company for its size. At the end of 2025 it had 898 employees, 380 of them in research and development, and 60% were outside the United States.⁶ Revenue over the last twelve months was $7.6B, or about $8.5M for each employee C (calculated). The 10-K says no customer was 10% or more of revenue in 2023, 2024 or 2025.⁶ The concentration the filings do describe is in the category: the 10-K says the company depends on mobile gaming, on the Apple and Google app stores, and on Meta, which is both a large advertiser and a rival.⁶ The 10-Q says its clients run from indie studios to “some of the largest global internet platforms, such as Meta and Google.”⁵

How the revenue grows

The 10-K and 10-Q split each revenue change into two parts: how many installs AppLovin delivered, and what it earned per install. In 2024 installs rose 50% and net revenue per install 22%. In 2025 installs rose 3% and revenue per install 72%. In the first quarter of 2026 installs fell 18% and revenue per install rose 93%; in the third quarter installs rose 2% and revenue per install 55%.¹⁵⁶¹³⁵ Over nine months installs are down 7% and revenue per install is up 68% C (calculated).⁵ In plain words, the models are getting more from each user, not finding more users. That is what “better AI” would look like if the claim is true, and it is also what a change in the mix of ads, in prices or in which advertisers are allowed in would look like. The filings do not separate them.

The new market: online shops

Since 2024 AppLovin has taken advertisers that sell on the web and not through an app, mainly e-commerce brands.⁵ The filings give no revenue for them. The company’s own number is older: in a note to shareholders in February 2025 the chief executive said that in December 2024 the category ran at about $1B a year of advertiser spend from around 600 customers, and that a minimum monthly spend was there to cover manual onboarding and would be lifted over the year.¹¹ A trade-press account of the 4 November 2026 webinar quotes the chief financial officer that spend from consumer advertisers was up double digits on the quarter and triple digits on the year, and says no dollar figure was given. The same article repeats a May report that the ad engine was scheduled to open to all advertisers in June 2026; the call did not confirm that it did.¹⁷ I did not hear the call. I cannot size this business, so I do not value it separately, and my growth path treats it as part of the whole.

The latest quarter, and the record against guidance

AppLovin reported the third quarter on 4 November, after the close. Revenue was $2,185M, up 55%, and adjusted EBITDA $1,842M, an 84% margin. Net income was $1,439M, or $4.30 a share. Free cash flow was $1,418M. The company bought back and withheld shares worth $1.4B in the quarter, and the board added $3.2B to the repurchase authorization, bringing it to $3.6B.³ The stock closed at $271.75 that day and $356.10 the next, a rise of 31%.¹ The chief executive said the new ad models were only partly rolled out, according to the trade-press summary of the call.¹⁷

The record has been uneven. Revenue beat the top of the range by $67M in the first quarter and by $100M in the third, but landed inside it in the second. Costs are moving too: research and development rose 132% in the third quarter to $102M, almost all of it stock pay for engineers, which went from $16M to $68M a quarter.⁵ Stock pay in total is $84M a quarter, up from $34M a year earlier.⁵ The next report, for the fourth quarter and the full year, is expected in February 2027: the 2025 results came on 11 February 2026 and the 2024 results on 12 February 2025. No date has been announced.⁴ The fourth-quarter guide, if repeated for four quarters, is about $9.5B a year, against $8.3B for 2026 C (calculated).³

So what. This is a very profitable auction layer whose revenue is growing because each install earns more. What I can know is the reported numbers, which are audited and show real cash. What I cannot know from the filings is how much of the growth is the models, how much is the mix of advertisers and ads, and how big online shops already are.

IIThe moat

The protection here is a loop. Each ad AppLovin shows teaches the model something about which user installs which game and what that user is worth. The 10-K says it directly: more advertisers bring more data on users and engagement, which sharpens the model, which makes the ads work better for the next advertiser.⁶ MAX adds to it, because an auction run by AppLovin lets it see the bids in thousands of apps. That is hard to copy in a quarter. AppLovin was founded in 2011 and its models and auction have been tuned for a decade.⁶

It is thinner than it looks, for three reasons. First, the clients have no long-term contracts with AppLovin, and advertisers normally buy from several platforms at once, so they can move spend in a week.⁶⁵ Second, the loop runs through other people’s rules. Apple and Google decide what data an ad system may collect and how an app is installed, and the 10-Q says their policy changes “can drive rapid change” across the mobile app and advertising businesses.⁵ Third, the rivals are close. The 10-K names Meta, Google, Amazon and Unity.⁶ Unity’s ad business, Vector, is reported to have reached a $1B annual run rate two quarters early, and a Wedbush analyst said after a court ruling in October that stronger Unity bidding could pressure AppLovin’s take rate.²⁰ I am not able to test the claims about Unity.

The record suggests the loop has held so far. If rivals were catching up, I would expect AppLovin to lose volume and then price. Installs are down 7% in nine months while revenue per install is up 68%.⁵ But that is also what it looks like when the models earn more from a smaller base of users, and a lower volume leaves less to learn from. A rival with a better model would show up first as falling revenue per install. In the third quarter it was up 55%.

Against the best comparable here, Meta, the protection is narrower. Meta owns the audience, the habit and the data, and it competes with AppLovin for the same advertisers while also buying its ads. AppLovin owns an auction and a model, and both sit on top of platforms it does not control. I would call it wide against today’s ad rivals, and narrow against Apple, Google and Meta acting together.²¹⁶

So what. The moat is real and it is young: it is a data loop that has paid off since 2023 on rules that others can change. Revenue per install is the number to watch; if it keeps rising without installs falling further, the loop is holding, and if it stalls, a rival or a rule change is the likelier reason than the model.

IIIThe competitive advantage

An advantage is something you can measure against a rival. AppLovin’s is cost. Of every dollar of revenue in the third quarter, 11 cents went to cost of revenue, mostly cloud computing, 3 cents to sales and marketing, 5 cents to research and 3 cents to general costs, leaving 78 cents of operating income.⁵ Meta’s margin is measured on gross ad revenue, a different base, so I do not compare the two margins directly.²¹

The cost line to watch is research. Research and development was $102M in the third quarter, up 132%, and almost all of the rise is stock pay.⁵ The 10-Q lists data-centre costs and engineering talent as the two drivers of cost.⁵ Cost of revenue rose 38% in the third quarter against 55% for revenue, so the leverage is still there.⁵

The advantage over Meta is different. Meta sells to anyone and sees the whole feed. AppLovin is paid on results in games, where an install can be measured within hours, so the model gets fast feedback. That is the strongest part of the case: a fast feedback loop in a category where the result is a number. It is weakest where the result is slower or shared. The February 2025 short reports argued the e-commerce results were partly credit taken from Meta’s ads, Muddy Waters estimated in March 2025 that only 25% to 35% of the sales were incremental, and the company replied that it drives real, incremental value.⁹⁴³¹¹ I cannot test that. The trade-press summary of the November call quotes the chief financial officer that spend from consumer advertisers rose by double digits on the quarter, which shows advertisers kept spending, though spending is not the same as incrementality.¹⁷

Put next to the other advertising businesses on this site, the price asks the least of the three. The like-for-like way to compare is how far each base case falls short of the return I ask. AppLovin falls short by 1.3 points, Meta by 1.6 and Alphabet by 4.6.²¹²² That ranks them. It does not make AppLovin cheap: all three prices are above my base values.

So what. The advantage is cost and speed of feedback in games, and it shows in an 84% margin on net revenue. What I cannot tell is how much of it carries to new categories, where the results are slower to measure and where large rivals already stand.

IVManagement

Start with who decides. AppLovin has Class A shares with one vote and Class B shares with twenty. Adam Foroughi, the co-founder and chief executive, holds 27.9M Class B and 2.4M Class A, about 9% of the economic value and 61.6% of the votes. With Herald Chen and their family trusts, who are bound by a voting agreement, the figure is 66.9%.⁸ The company calls itself a controlled company and says it does not use the exemptions; a majority of the board is independent, and since 2 April 2026 so is the chair.⁸²³ The controlling holders can still decide anything that goes to a vote. On 29 October 2026 holders of about 67% of the votes approved, by written consent, moving the company’s legal home from Delaware to Nevada, and the company will send a Schedule 14C to all holders.²⁴ I have not seen that document. The bylaws now send derivative suits and fiduciary claims to the Delaware Court of Chancery, so the move changes where such cases would be heard; the 10-Q says written consent stays available while Class B has a majority of the votes.⁵ At the June meeting, 63.1M votes were cast against the pay plan and 753.0M for; officer exculpation passed with 80.4M against.²⁵

Then how they are paid. The base salary of the four named officers has been capped at $400,000 since before the 2021 listing, and there is no cash bonus.⁸ Everything else is stock. In October 2025 each of the four got the same one-year grant of restricted stock, worth $12.6M at grant, and the company says it chose one-year vesting to deliver value despite the share price’s volatility.⁸ In 2023 Mr Foroughi was granted $82.9M of performance stock with market conditions, fully earned in 2024; the proxy’s “compensation actually paid” for 2024 was $386M, a figure it says is heavily influenced by the share price.⁸ What that pays them to do is lift the stock price within a year or two. It does not pay them to watch the long term, and the company’s own pay-versus-performance table names no financial measure.⁸ Against a median employee paid $133,808, the chief executive’s pay ratio is 97 to 1.⁸ A system like this gets the effort it pays for; it does not pay anyone to say no to a bad year.

Their own money is in, and a lot of it has been taken out. The chief executive’s stake is worth about $12B at the price C (calculated).⁸ He sold $231M of stock in 2025 and $97M so far in 2026, which is under 2% of his holding in each year, and Mr Chen sold $719M in 2025. Directors and officers together sold about $1.3B in 2025 and $0.4B in 2026 to October; I found no open-market purchase by an insider in either year C (calculated).¹⁰ One director, Eduardo Vivas, has 550,000 shares pledged under a prepaid variable forward, and the company’s policy lets employees pledge up to 10% of their stock with approval.⁸ Selling is normal for founders and is not a signal by itself. The sales in 2026 averaged about $483 a share, against $395.51 now, and the lowest close in the last year was $228.50, where I found no insider purchase.

Then what they did with capital over the cycle. Since 2022 the company has spent $7.5B on buying its own stock, in a year when the stock was $9.30 (the low close, 27 December 2022) and in a year when it was $733.60.²⁶⁵¹ In the first nine months of 2026 it bought back 7.3M shares for $2.9B, an average of about $397, close to today’s price.⁵ In the third quarter it paid $443 a share in July, $317 in August and $311 in September, and spent $1.4B of the quarter’s $1.4B of free cash flow on repurchases and tax withholding.⁵³ The record is mixed. Buying at $10 to $40 in 2022 and 2023 added much more per dollar than buying at $400 or more in 2025. The company also owned game studios, took a $189M goodwill write-down on them in 2025 and sold them that year for $716M.³⁵ Two senior officers are leaving this year, the chief technology officer and the chief legal officer, and one director is not standing again.²³

So what. The people running it are paid in stock and own a lot of it, which is a strong reason to want the price up, and the pay system rewards that over a short horizon. They also decide alone: two holders can approve nearly anything, and they have been steady sellers at prices above today’s. I read that as capable owners with no one to say no, and what I would want to see is a purchase.

VFinancials

The record is short for a company this size. AppLovin listed on 15 April 2021.²⁷ I show the advertising business, which is the part that remains, from 2019, and the whole company as it was first reported next to it. The two differ until 2025, because the game studios were in the second set until the sale.⁶

Three things stand out. Revenue has grown 74% a year since 2019, 70% in 2025 and 55% in the quarter just reported C (calculated).²⁷⁶³ The adjusted EBITDA margin went from 67% in 2023 to 84% in the latest quarter, because the models earn more per install while the cost to serve them grows slower than revenue.¹⁵⁶³ And cash arrived in step: operating cash flow went from $0.2B in 2019 to $4.0B in 2025.²⁶

What an owner can take out

I value the cash left for owners after stock pay. For AppLovin that number is $4,544M for the last twelve months, against $5,013M of net income and $6,412M of adjusted EBITDA C (calculated).⁵³⁴ Capital spending is almost nothing: $9M in nine months, because the computing is rented and shows up in cost of revenue.⁵ The gap between adjusted EBITDA and free cash flow is cash tax, interest and receivables. Cash paid for taxes was $926M in the first nine months of 2026, against $167M a year earlier, and the 10-Q ties the higher tax charge to higher profit, more foreign income and smaller stock-pay deductions.³⁵ Receivables rose to $2.39B from $1.82B at the end of 2025, a rise of $569M in nine months, as revenue grew.³ Stock pay is the cost the company leaves out of adjusted EBITDA. It was $334M in the last twelve months, 7% of free cash flow, and it is rising: $84M in the third quarter against $34M a year earlier.⁵ Net income also includes a $97M unrealized gain on private holdings in nine months that is not cash.⁵

Debt, cash and what the company bought

At 30 September 2026 AppLovin had $3.17B of cash and $3.55B of senior notes at face: $1.0B at 5.125% due 2029, $1.0B at 5.375% due 2031, $1.0B at 5.5% due 2034 and $550M at 5.95% due 2054.³⁶ Net debt is therefore $376M C (calculated), under a tenth of one year’s adjusted EBITDA. It has an undrawn $1B revolving credit line, and the 10-K says the notes are rated investment grade.⁶ Interest cost is about $204M a year.⁵ The company spent $7.5B on buying its own stock from 2022 to September 2026 and the share count fell 12% from the end of 2021; since the end of 2023 it has spent about $6.0B and the count has fallen 2.5% C (calculated).²⁶⁵

What the market paid

The price per dollar of revenue has been a bigger mover than revenue. At the end of 2022 the market value was 3.8 times that year’s advertising revenue, at the end of 2023 7.4 times, at the end of 2024 34 times and at the end of 2025 42 times. Now it is 17.2 times the last twelve months C (calculated).¹²⁶ From the high close of $733.60 on 22 December 2025 the stock fell 69% to $228.50 on 30 October 2026 and has since risen 73%.¹ The index inclusions came in the middle: S&P Dow Jones Indices said on 5 September 2025 that AppLovin would join the S&P 500 before the open on 22 September 2025, and Nasdaq lists it in the Nasdaq-100.²⁹³⁰

So what. The financials are the best part of the case: growth, an 84% margin and cash that matches the profit. What they cannot show is whether it lasts. The multiple on revenue moved from 4 times to 42 times and back to 17 in four years while revenue grew about seven-fold, so the price has moved with expectations more than with results.

VIValuation

First, how this goes badly wrong

Before the numbers I asked what would most likely hurt an owner at this price, and worked backward. Three things stand out.

The price per install stops rising, or gets pushed back. Almost all of the growth since the start of 2025 is revenue per install, up 72% in 2025 and 68% in nine months of 2026, on installs that are flat to down.⁶⁵ If an Apple or Google rule, a court order or a regulator restricts the way ads trigger installs, or the way data is gathered, that number is where it would show. Cutting my base revenue path by a quarter from 2027 takes the value from $336 to $254, and by half to $171 E (estimate).

A rival wins the auction. AppLovin is paid by results and advertisers can move spend in a week. If Unity, Meta or Google bid better, AppLovin either loses volume or pays publishers more to keep it. That is my bear case: revenue falls to $6,497M by 2030 and the margin to 58%, worth $88 a share. Each five points off the margin in every year costs about $24 a share in the base E (estimate).

The new models and online shops stall. At $395.51 the market pays 17 times revenue, so the price needs several more years of fast growth. Half the growth I assume takes the base to $222.

None of these needs anything exotic. A fourth risk is different in kind: that the reports of short sellers or the lawsuits turn into findings. I treat each claim, and what it would change in my numbers, in Section VIII. Here I note only that it sits inside the first two risks: the claims are about how installs are generated and who gets credit for them.

What an owner could really take out

I value cash after stock pay. For AppLovin that is $4,544M over twelve months, 71% of adjusted EBITDA and 91% of net income C (calculated). It is close to what the company reports, because the business needs almost no capital and its stock pay is small beside its profit. The distance from adjusted EBITDA is mostly tax, which has grown fast. In my base the cash tax rate is 20% of profit after stock pay, against a book rate of 16% in nine months and cash taxes of $926M.⁵³ Owner cash is 3.5% of the market value. An 84% margin does not mean a high yield: the cash is 3.5 cents a year for each dollar paid.

What the price implies

Put the other way round: what does $395.51 need? The market value is $131.5B on 332.5M shares, and with $376M of net debt the business is priced at $131.9B, or 29 times the last twelve months of owner cash. At an 11.5% discount rate and my base margins and spending, the price needs revenue of $29.7B in 2036, growing 13.5% a year from this year’s $8.3B. My base has $24.3B and 11.3% a year. The other way to read it is that my base cash flows return 10.2% a year at the price, against the 11.5% I ask. The price therefore asks for about a fifth more growth than my base over ten years.

What I can’t judge

• How much of revenue per install is the model and how much is mix. The filings split volume and price, not the cause of the price.
• How large online shops are. There is no revenue figure in a filing, and the company’s last public number is for December 2024.¹¹
• Whether the install practices alleged by short sellers are within platform rules. The company says its ads follow app-store policies. No court has ruled either way. The SEC, according to the company’s chief financial officer, closed its inquiry with no recommended action; I have not seen the SEC’s letter.¹¹⁴⁹
• Who stands behind one large holder. Angel Pride Holdings holds 5.6% of the Class A shares and the proxy names no person behind it. Culper says it is linked to an investor whom a Schedule 13G/A filed on 30 January 2025 names as holding 4.0% at 31 December 2024, and who denies, in court, the criminal allegations Culper makes about him. I found no later filing from these holders, and I cannot check the link.⁸⁴²⁵¹
• Who the advertisers are. No customer was 10% or more of revenue, but the filings do not rank them.⁶
• What the move to Nevada changes for holders, because I have not seen the Schedule 14C.²⁴
• Where Apple, Google and Meta stand on AppLovin’s methods. I found no public statement from any of them.

Because I can’t know these, I show a range and where the price sits in it, and I don’t treat $336 as more exact than it is. A fair reading of my work is that the base is somewhere from $285 to $410 a share, depending on whether the right discount rate is 13% or 10% E (estimate).

The discount rate

The ten-year Treasury yield was 4.98% on 10 November, and the implied equity risk premium I use across the site is 4.20%.³¹³² AppLovin’s weekly beta against the S&P 500 ETF is 2.44 over five years, 2.80 over three and 2.24 over two C (calculated).¹³³ On the five-year beta the cost of equity is 15.2%. I use 11.5%, well below the formula, because a beta measured through a 70-fold rise and a 69% fall mixes two different companies, the studio owner of 2021 and the advertising business of today. It is one point above the 10.5% I use for Meta, because AppLovin depends on two app stores, has no long-term client contracts, and is controlled by two holders.²¹ At 10% my base is $410 and my bull $648; at 13% they are $285 and $439; at the formula’s 15.2% the base is $233.

Three cases

Bear

The price gains stop and then reverse. Revenue grows 6% in 2027 to $8,831M, which is below the fourth-quarter pace, then falls 12% a year in 2028 and 2029 and 5% in 2030, reaching $6,497M that year, and grows 2% a year after. The cause could be a platform ruling on installs, a rival such as Unity or Meta taking the auction, or both. The adjusted EBITDA margin falls from 83% to 58% by 2030 and 52% by 2036 as AppLovin pays publishers and compute more to keep the volume, stock pay is 6% of revenue and cash tax is 22%. Value $88: owner cash of $4,992M in 2027 shrinks to $2,643M in 2030 E (estimate).

Base

Revenue keeps growing, more slowly each year: 28% in 2027 to $10,664M, 20% in 2028, 16%, 13%, 10%, and 4% by 2036, when it is $24,333M. The margin slips from 83% to 72% as rivals and publishers take more. Stock pay is 4.5% of revenue and cash tax 20%. Owner cash grows from $6,383M to $12,949M. Value $336.

Bull

The new ad models and online shops both work. Revenue grows 38% in 2027, 30% in 2028 and 24% in 2029, reaching $22,054M in 2030 and $37,736M in 2036. The margin holds at 84% to 2029 and eases to 77%. Stock pay is 4% and cash tax 19%. Value $525.

Other uses of the same dollar

The cleanest alternative is a Treasury note. Ten-year paper yielded 4.98% on 10 November with no risk of the kind above.³¹ At $395.51 my base cash flows return 10.2% a year, 5.2 points more than a Treasury, for a business that has to keep growing price per install, and my bull returns 14.0% against 11.5% that I ask E (estimate). The best comparison on this site is Meta, the larger ad business: its base returns 8.9% a year against 10.5%, a shortfall of 1.6 points, against AppLovin’s 1.3.²¹ Alphabet’s base returns 5.4% against 10.0%.²² By these numbers AppLovin is the smallest shortfall of the three. That ranks them; it does not make the price one I would call low. There is also the company’s own use of the dollar. In the third quarter it bought its stock at $311 to $443, and it has authorized $3.6B more.⁵³

Two different questions

The first is how good the business is. By the numbers it is a very strong business: 74% a year growth since 2019, an 84% margin, almost no capital, $4.9B of free cash flow, and net debt of under a tenth of a year’s EBITDA.⁵³ Against that, it has a short record, with revenue growing 55% for a few quarters on a rise in price per install that no one outside can explain, and it sits on rules that others set. The second question is how good the price is. At $395.51 the price is 18% above my base value and 25% below my bull, and it needs growth about a fifth faster than my base for ten years. A great business can be a poor purchase at the wrong price. This looks like a very good business at a price that is higher than my base and lower than the case where everything goes right, and the market may be seeing something I cannot: that the new models are a second step as large as the first. I show what that would have to be worth.

The margin of safety

I set it from the bear case: I want a price at which my $88 bear value is at least three quarters of what I pay, which is $117. That is 30% of today’s price and about half of the lowest close in the last year, $228.50 on 30 October 2026.¹ Net debt per share is $1.13, so at that price the market would value the business at $39B, or 9 times owner cash. I am not forecasting a move to $117. I am saying where the margin of safety begins.

So what. My base value is $336 and the price is 18% above it, so the market is paying for a few more years of fast growth than I model, or for a lower risk than I allow. What would settle it is whether revenue per install keeps rising without installs falling further, and whether platform rules and the open claims move the way the company says. I cannot judge either from outside.

VIIThe case for the price

What most people miss

Three facts cut in AppLovin’s favour and get little attention. First, size: it earns $7.6B of revenue with 898 employees, and its capital spending was $9M in nine months, so growth does not need cash.⁶⁵ Second, the company used the fall to buy. In the third quarter it spent $1.4B on repurchases and tax withholding, nearly all of its $1.4B of free cash flow, at average prices of $443 in July, $317 in August and $311 in September, and the board added $3.2B to the authorization.⁵³ Third, the guide is an exit rate: $2,380M in the fourth quarter, repeated four times, is $9.5B a year, 14% above this year’s $8.3B before any growth in 2027 C (calculated).³

The bull case

I found no dated statement about AppLovin by a named investor that I could open and check, so I give the case as the people who run it make it, and as one analyst reads the risk. On 4 November the company said its record third quarter had “industry-leading” margins, and a trade-press summary of the call says the chief executive described the new ad models as only partly rolled out, with no date for the rest.³¹⁷ On 2 October, after the court refused AppLovin’s request against Unity, a Wedbush analyst, Alicia Reese, called it a sign of a competitive shift but said she did not expect the dispute to affect the engine’s performance.²⁰ A securities suit filed on 16 September 2026 alleges that earlier statements about the pace of improvement of the ad models were misleading; that is an allegation, and the company says it lacks merit.⁵⁵⁴

In plain words: the models are getting better at pairing an advertiser’s goal with a user, the improvement is only partly in the numbers, the same engine now sells to online shops, and the company is buying its own stock hard. The record supports part of this. Revenue beat the top of the guide by $67M in the first quarter and by $100M in the third, and net revenue per install rose 55% in the quarter.¹⁸³⁵

Against reality

If AppLovin keeps lifting revenue per install by 25% to 30% a year while installs hold, brings online shops in at scale, and holds the margin at 80% or better through 2028, the price is possible. That is close to my bull case, which is worth $525 at 11.5% and $648 at 10%. Today’s numbers are part of the way there. The bull needs revenue of $11,497M in 2027 and $22,054M in 2030, 2.6 times this year’s $8.3B; a margin of 84% through 2029, which is where the latest quarter is; and cash after stock pay of $13,618M in 2030 against $4,544M over the last twelve months E (estimate). Only the margin is in hand.

It helps to multiply it out. The price needs six things together: revenue per install still rising by 25% or more in 2027, 60%; installs no longer falling, 80%; online shops reaching a large share, 50%; the margin staying at 80% or better through 2028, 65%; no adverse ruling on the install and data practices, 75%; and no loss of share to Unity, Meta or Google, 70%. Multiplied, that is about 8%, or one chance in 12. These are my guesses, they are not independent, and I do not use them in the values E (estimate).

What a 10x would take. Ten times $395.51 is $3,955, about $1.3 trillion of market value on today’s shares. At 25 times owner cash that needs about $53B a year. At a 55% owner cash margin, below today’s 60%, that is $96B of revenue, 11.5 times this year’s. By 2036 that means growth of about 28% a year for ten years, and by 2031 about 63% a year for five C (calculated). AppLovin’s revenue has grown 74% a year for six years, and its best year was 2021, at 225%, from a small base.²⁷⁶ Tenfold moves in this stock have happened: it rose from $10.53 at the end of 2022 to $673.82 at the end of 2025, 64 times, and is down 46% from its high since.¹ A repeat from here needs revenue of $96B, about 41% of what Meta’s ad business earns now ($235.6B in twelve months).²¹ That is close to impossible for a company of this size, and I would not build a case on it.

Where I come down

What moved me is cash that matches the profit, a margin near 84% for four quarters running, and a company that bought hard when the stock fell. The strong third quarter did not move my base, because the growth came from price per install, which I cannot explain from outside, and because the open claims did not move either way. What would raise my values is two more quarters of rising revenue per install with installs no longer falling, or an insider buying. A ruling for the company or a closed inquiry would take away a tail, but it would not lift my base, which already assumes the install practices continue.

So what. The price is possible if revenue per install keeps rising by a quarter or more with installs steady, online shops grow into a large business and the margin stays at 80% or better. Today only the margin is in hand. What I can know now is the guide, the margin and the cash. What I cannot know is how long the rise in price per install lasts, or how much of it the new models explain. The next dated checks are the fourth-quarter release and 10-K in February 2027, the Talbot lead-plaintiff deadline on 16 November 2026, and the ruling on the motion to dismiss in the securities case.³⁵⁵⁵

VIIIRisks, and what would change my view

The three ways this goes badly wrong are at the start of Valuation. These are the rest.

• Platform rules. The 10-Q says the app stores have discretion over fees, over which apps they promote and over how much consumer information ad networks receive.⁵
• Category concentration. The 10-K says gaming is the base of the business and that Meta and Google are both partners and rivals.⁶
• Control. Two holders vote 66.9% and can decide anything put to a vote; on 29 October they moved to change the company’s legal home to Nevada by written consent.⁸²⁴
• Litigation. Four securities and derivative matters are open and the company cannot estimate a loss; a county suit over ads in children’s games was reported as filed on 5 October.⁵⁵⁰
• Stock pay. It was $84M in the third quarter against $34M a year earlier, and most of the rise is research staff.⁵
• Tax. Cash taxes were $926M in nine months against $167M, and the 10-Q ties the rise to higher profit and smaller stock-pay deductions.³⁵
• People. The chief technology officer and chief legal officer are leaving, and the chief executive gave up the chair.²³

The short reports

Four outside firms have published critical reports on AppLovin: Culper Research (February and June 2025), Fuzzy Panda Research (February 2025), Muddy Waters Research (March and May 2025) and CapitalWatch (January 2026). Since then the company has faced class actions, a county suit, an SEC inquiry that it says is closed, and two defamation suits filed against Culper by people and companies the June report names. Read this block with one rule. What a report says is an allegation, and what a company or a lawyer says is a statement. Neither is a finding unless a court or a regulator has made one.

Both sides have a reason to speak. A short seller profits if the price falls. Culper’s disclaimer says it is likely to trade the shares right after publishing, and Muddy Waters says it begins covering most of its short when it publishes.⁹⁴³ AppLovin’s management and insiders are paid in stock: the chief executive owns about $12B of it, and insiders sold about $372M in 2026.⁸¹⁰ Neither side is a neutral reader of the facts.

I opened the reports myself where I could. I read Culper’s February 2025 report on its hosted PDF and Fuzzy Panda’s post on its own site. I read Culper’s June 2025 report and Muddy Waters’ two reports through Internet Archive copies of the PDFs the firms host, because Culper’s site showed a bot check that I did not try to get around and Muddy Waters’ site did not load.^937424344 I could not open CapitalWatch’s report, so that row is secondhand. Culper’s founder, Christian Lamarco, signed the note at the front of the February report, which says Culper shared some research with Fuzzy Panda and that each firm presents its own views.⁹

The stock moved on several of these days. By Nasdaq close it fell 12.2% on 26 February 2025 (the two February reports), 20.1% on 27 March 2025 (Muddy Waters), 0.7% on 12 June 2025 (Culper), 14.0% on 6 October 2025 (the day Bloomberg reported the SEC inquiry) and 5.8% on 21 January 2026 (CapitalWatch). It rose 13.2% on 9 February 2026, after the CapitalWatch retraction.¹ A price move shows what traders did that day. It does not show whether a claim is true.

Since 2025 the claims have also moved into courts and filings. The second table has each matter with its source. A complaint is a party’s allegations.

Now the part that matters for value. I ran each claim through my model, to see which changes the answer and by how much. I ran the findings that would go against the company and, in the last two rows, the ones that would go for it.

The pattern is that a finding on install practices matters most, because it goes to the one number that has carried growth. A $1B settlement matters little: it is $3 a share against $336. The claims are not all about growth, though. Fuzzy Panda’s points on revenue quality and subsidiaries, and Culper’s on disclosure, are about how far to trust reported figures and who controls the company. The filings are audited by Deloitte and show $4.9B of free cash flow in twelve months, and no regulator or court has said otherwise.²⁵⁶³ A dismissal or a closed inquiry would not lift my base, because my base already assumes the practices continue; it would remove a tail.

So what. For someone valuing the company, these reports and suits ask two things. The first is whether the price per install is real and repeatable. If it is, the cash is worth what the model says. If part of it rests on practices that platforms or courts stop, the revenue path falls, by $33 to $165 a share on a 10% to 50% cut. The second is about who controls the company and what it has disclosed, where I have no model input and only the discount rate as a lever: a point higher is $36 less, a point lower is $46 more. What I can know now is the filings and the court record. They show suits filed and no finding, an SEC inquiry that the company says was closed without action, a partial retraction by one critic and, according to Mr. Tang’s complaint, Culper’s June report uncorrected in June 2026, and a company that confirms its Direct Download experience while saying every install is a choice. What I cannot know is how much revenue comes from the methods alleged, what the SEC examined, and whether the ownership claims go beyond the shareholdings in the filings. I give neither side’s claim more weight than a filing, a regulator or a court gives it. Watch next, with dates: the Talbot lead-plaintiff deadline on 16 November 2026;⁵⁵ a ruling on the motion to dismiss in Brownback, N.D. Cal. docket 3:25-cv-02772, no date set;⁵⁵³ the legal note in the 10-K, expected February 2027; the S.D.N.Y. dockets 1:26-cv-04930 and 1:26-cv-05018;³⁸³⁹ and the San Diego Superior Court docket.⁵⁰

The whole page

So what. The numbers show a very profitable business. What is open is how long revenue per install keeps rising, and whether the rules it depends on stay as they are. For the price to return the 11.5% I ask, growth has to run about a fifth faster than my base for ten years. My base could be too high if an app store changes a rule, if Unity or Meta takes the auction, or if a court finds against the install practices. The short sellers’ account needs install flows that break platform rules or rest on inadvertent clicks, and credit in online shops that is overstated. Management’s account needs the new ad models to be the reason for the rise in price per install. I can see the filings, the price per install and the buyback prices. I cannot see what share of revenue comes from the methods alleged, what the SEC examined, how the courts will rule, or how much of the rise in price per install is the model. Watch next: the Talbot lead-plaintiff deadline on 16 November 2026;⁵⁵ the Schedule 14C for the Nevada move, after which the conversion cannot take place for 20 days;²⁴ the fourth-quarter results and the 10-K’s legal note, expected in February 2027; and the ruling on the motion to dismiss in Brownback, fully briefed since February 2026.⁵ I do not lean either way.