Latest / Deep Values: Value Investing Deep Dives / Why Record Q2 profits tanked AppLovin‘s Stock
Transcript
- 0:00I want you to imagine a machine where like you feed in $0.12 and
- 0:03it immediately hands you back a dollar.
- 0:06That sounds like a pretty good machine.
- 0:08Right. Now imagine that exact same
- 0:09machine is running constantly, I mean 24 hours a day, eventually
- 0:13generating over $1.2 billion in a single quarter.
- 0:17You'd be thrilled. You would think the people who
- 0:19own that machine would be celebrating in the streets, but
- 0:22when the market found out about this incredible cash generation,
- 0:26investors just absolutely panicked.
- 0:29The company's stock plummeted 16% overnight.
- 0:33Which I know sounds completely irrational on the surface, but
- 0:38when you look at the underlying mechanics of how that machine
- 0:41actually operates, the panic suddenly makes a, well, a
- 0:45terrifying kind of sense. Welcome to another deep dive.
- 0:48Today we are opening up a massive 53 file 2026 financial
- 0:53and research package detailing app Love Incorporation.
- 0:56And this isn't just a couple of summary spreadsheets for you to
- 0:58gloss over, by the way. Yeah, definitely not.
- 1:01We are looking at a comprehensive Q22026 earnings
- 1:04analysis, their 2025 annual report, some really extensive
- 1:08deep value investing checklist and incredibly detailed
- 1:12discounted cash flow models. A.
- 1:13Lot of data. It is, and our mission today is
- 1:16to figure out how a former mobile game developer
- 1:19transformed into this artificial intelligence ad tech juggernaut
- 1:23with these staggering profit margins.
- 1:25And to determine whether a tech giant like this actually
- 1:27qualifies as a deep value investment in today's.
- 1:30Exactly but before we unpack exactly how they built this
- 1:33digital money machine, I want to share something with you really
- 1:36quickly to access more advanced researches, applications and our
- 1:40latest deep values agents. Welcome to explore and join deep
- 1:44values dot a track total market temperature.
- 1:47Welcome to search and download iOS app Buffett indicator from
- 1:50Apple App Store or Bubble watch from Google Play.
- 1:54Disclaimer All analysis is based on publicly available sources
- 1:57and may be subject to revisions or different interpretations.
- 2:01All contents from this channel is for research and
- 2:03entertainment purposes only and does not constitute any
- 2:05investment or financial advice or service.
- 2:08Conduct your own due diligence before making investment
- 2:10financial decisions. Oh, it's good to keep in mind.
- 2:12Right. OK, let's unpack the source
- 2:14material because to really understand this company, we have
- 2:17to start with the whiplash of their August 5, 2026 earnings
- 2:20report. Oh yeah, that report perfectly
- 2:23illustrates the Razor's edge of AI driven market expectations.
- 2:27It really does. The numbers from this Q2 release
- 2:29are frankly they're mind blowing.
- 2:31Total revenue hit $1.924 billion.
- 2:35Which is a 53% year over year increase.
- 2:38Massive net income reached 1.267 billion, but here is the number
- 2:43that really jumps off the page for me.
- 2:45They achieved an 84% Adjusted EBITDA margin, which is.
- 2:50Just an astronomical figure for any company operating at that
- 2:53scale. Right.
- 2:53And for those of you listening who don't spend your weekends
- 2:56reading balance sheets, Adjusted EBITDA is essentially a way of
- 2:59measuring a company's pure operating cash profit.
- 3:02Yeah, it strips out things like taxes, interest and, you know,
- 3:05the depreciation of asset. Exactly, giving you a look at
- 3:08the raw cash the core business is actually generating.
- 3:11So keeping $0.84 of operating profit for every dollar that
- 3:15comes in the door is almost unheard of.
- 3:17It's basically printing money. It is, plus their cash on hand
- 3:20swelled to over 3 billion. So I have to ask you, how does a
- 3:24business grow its revenue by 53%, maintain an 84% operating
- 3:28profit margin and then watch its stock crashed down to $350.00 in
- 3:33after hours trading? I mean, are investors just
- 3:35completely detached from reality?
- 3:37What's fascinating here is that the market's reaction wasn't
- 3:40about the cash they printed today, it was entirely about the
- 3:43mathematical trajectory of the engine printing it.
- 3:46OK. The trajectory.
- 3:47Right. And that engine is Axon 2 Point
- 3:50O, which is 11's proprietary artificial intelligence
- 3:54algorithm. The brain of the whole
- 3:55operation. Exactly.
- 3:57Axon is the redictive model that matches mobile ads to the users
- 4:01most likely to click on them. The market noticed that the AI
- 4:05models performance improvements during this specific quarter
- 4:08were well slightly lighter than they had historically observed.
- 4:11Slightly lighter. Yeah.
- 4:12I mean, it's still improved, right?
- 4:14And it's still generated massive revenue, but the rate of
- 4:16acceleration showed a very slight deceleration in momentum.
- 4:20Meaning the algorithm didn't get smarter quite as fast as Wall
- 4:23Street had modeled it too. Precisely because of that
- 4:26microscopic friction, they marginally missed the most
- 4:29optimistic Wall Street consensus, which was looking for
- 4:321.94 billion in revenue. And they hit 1.924 billion,
- 4:36right? So Wall Street didn't panic over
- 4:39the current cash flow, they panicked over the consistency of
- 4:42the hyper growth. Because in the AI space,
- 4:44investors aren't just buying current growth.
- 4:46No, not at all. They are buying the compound
- 4:49interest of an ever improving algorithm.
- 4:52It's like it's not just about how fast the rocket is going,
- 4:55it's about escape velocity. That is a great way to put it.
- 4:57In AI, if you're compounding, learning rate slows down even a
- 5:01fraction of a percent. The math basically dictates that
- 5:04you might eventually fall back to Earth rather than reaching
- 5:07orbit. Yeah.
- 5:08So Wall Street wasn't looking at the current altitude, they were
- 5:10looking at the trajectory. That is a perfect way to
- 5:13visualize it. If the compounding slows even a
- 5:16fraction of a ercent, the forward-looking valuation models
- 5:19just shatter. Wall Treet is asking is this a
- 5:23temporary blip or has Axon 2.0 reached the limits of its
- 5:27predictive capability? Wow.
- 5:29But this intense, almost singular focus on the Axon
- 5:33algorithm didn't happen by accident, did it?
- 5:35No, it was very deliberate. Because if we want to understand
- 5:38how they achieved those 84% margins in the 1st place and why
- 5:41AI performance is now the only metric Wall Street cares about,
- 5:45we have to look back at the massive strategic pivot revealed
- 5:49in their 2025 annual report. The 10K, yeah, that pivot is the
- 5:53whole story. The sources detailed detail how
- 5:55App Levin used to literally make its own mobile games.
- 5:58Right, they had an in house studio.
- 6:00But in June 2025, they took their Apps gaming portfolio and
- 6:05sold it off to a company called Triple dot Studios for $400
- 6:08million in cash plus a 20% equity stake.
- 6:13Which was an absolute masterstroke of corporate
- 6:15strategy. By divesting the Games, they
- 6:17removed what the industry calls channel conflict.
- 6:20Wait, let's break that down for the listener.
- 6:21Channel conflict is essentially when you are competing against
- 6:25your own customers. Right.
- 6:26Yes, exactly. Think about it from the
- 6:28perspective of a third party mobile game developer.
- 6:31You're trying to find an ad network to monetize your new
- 6:35puzzle game run. Yeah.
- 6:36If Affluent also makes puzzle games, are you really going to
- 6:39plug your app into their network and give them all your valuable
- 6:42user data? Definitely not.
- 6:43I wouldn't hand my competitor the exact demographic profile of
- 6:47my best players. I mean, they would just use that
- 6:49data to target those players for their own games.
- 6:52Exactly. So by selling off the game
- 6:54studios, Apple Oven became a completely neutral technology
- 6:58platform, right? They transitioned from a
- 7:00competitor into a pure software as a service and programmatic
- 7:04marketplace. It's like they realized that
- 7:06building the fastest race cars year after year was too
- 7:10exhausting. And far too risky.
- 7:11Yeah, because consumer tastes change so fast.
- 7:15So they just sold their cars and bought the only toll booth on
- 7:18the only highway. Yep.
- 7:19The toll booth. Model and the sources breakdown
- 7:21this toll booth into two main sides.
- 7:23You have app discovery on the demand side, which brings in
- 7:26about 80% of their revenue. That's basically advertisers
- 7:29bidding to acquire users, right? Then you have Macs on the supply
- 7:33side bringing in roughly 10% of revenue, which is the software
- 7:37letting publishers run real time options for their ad space.
- 7:40And operating that pure software marketplace is exactly what
- 7:44allowed their gross margins to explode to nearly 89%. 89%,
- 7:48yeah. Let's look at the mechanics of
- 7:50what happens when you open a game on your phone in the
- 7:53milliseconds before an ad loads. The Mac software holds a real
- 7:57time auction. OK.
- 7:59Exxon 2 Point O predicts exactly how much an advertiser should
- 8:03bid on your specific attention based on your past behavior.
- 8:07When that transaction clears, Apple Oven takes a massive cut.
- 8:11And what's crazy is that the actual cost of their revenue
- 8:14dropped to a mere 12% of total sales.
- 8:17It's incredible. Because when you sell physical
- 8:19goods, you have to buy raw materials.
- 8:20When you develop video games, you have to pay teams of
- 8:24engineers and artists. That's overhead.
- 8:26Right, but when you are simply routing advertiser demand to
- 8:29publisher supply using an AI algorithm, every incremental
- 8:33dollar of revenue cost you almost nothing to generate.
- 8:36The infrastructure is already built.
- 8:39That is the Holy Grail of operating leverage.
- 8:41Your fixed costs stay relatively flat, while your revenue scales
- 8:45infinitely with computing power. If they are printing this much
- 8:48cash with almost zero overhead, my immediate thought is like I
- 8:51need to buy this stock. But when I look at the research
- 8:54applying traditional value investing metrics to this
- 8:57company, it basically tells me to run for the hills.
- 9:00Totally. Why the massive disconnect?
- 9:03I mean, can an AI hyper growth stock actually be a value play?
- 9:07Well, depends entirely on whose rulebook you are reading.
- 9:10If you look at traditional Benjamin Graham metrics, you
- 9:13know, the old school foundational rules of value
- 9:15investing. It's a bloodbath.
- 9:17A complete failure. Let's actually look at the
- 9:19numbers. The trailing price to earnings
- 9:21ratio, or PE is 26.6 X, right? And the PE ratio just tells you
- 9:26how much you are paying for every dollar of the company's
- 9:29current earnings. A strict value investor usually
- 9:32wants that under 15. So 26.6 looks pretty expensive,
- 9:35Definitely. But the real shocker is the
- 9:37price to book ratio. It is a massive 37.1 X which
- 9:42brutally fails the strict 1.5 X benchmark of traditional value
- 9:46investing. To understand why it fails so
- 9:48spectacularly, we have to define what book value actually is.
- 9:51Yeah. Let's do that.
- 9:52Benjamin Graham designed these metrics in an era of railroads,
- 9:57you know, textiles and factories.
- 9:59Book values literally the net value of a company's physical
- 10:02assets if you liquidated them today.
- 10:04Right. Like if a factory goes bankrupt,
- 10:06you can sell the land, you can sell the machines, you can sell
- 10:09the steel. That's tangible book value.
- 10:11Exactly. But Applevin doesn't own
- 10:14factories. Right.
- 10:15It's just software, right? Traditional accounting simply
- 10:17does not know how to value a digital toll booth.
- 10:20Applevins true assets aren't capitalized on a balance sheet
- 10:24in a way Benjamin Graham would recognize.
- 10:25Because their asset is the algorithm.
- 10:28It's the Axon 2 point O AI algorithm and crucially, the
- 10:32seven petabytes of localized daily data it processes.
- 10:37You cannot put a massive market share of human attention on a
- 10:39traditional balance sheet. You really can't.
- 10:42So when you look at a price to book ratio of 37, it looks
- 10:45wildly expensive, but only because the denominator, the
- 10:47book value, doesn't reflect the company's actual economic Moat.
- 10:51But wait, I have to push back on that a little bit.
- 10:53Go ahead. If their only real asset is an
- 10:56algorithm and seven petabytes of behavioral data, isn't that data
- 10:59Moat incredibly fragile? I mean, a factory is a physical
- 11:03reality. You can't just delete a steel
- 11:04mill, but a data stream isn't that dead book value metric
- 11:08actually a terrifyingly accurate reflection of their fragility?
- 11:12If the data stops flowing, their book value really is 0.
- 11:15That is the multibillion dollar question, honestly, and it is
- 11:19exactly why the market reacts to violently to those slight
- 11:23misses. We talked about it.
- 11:24You are absolutely right that a digital mode is theoretically
- 11:28more fragile than a physical one, but while that data flows,
- 11:32it produces cash at a rate physical factories could only
- 11:34dream of. Which brings us to the other
- 11:36side of the valuation debate. Because when you flip the page
- 11:39in the research, Aplivan absolutely crushes the cash
- 11:42generation tests. They destroy them.
- 11:44They boast a 55.4% free cash flow margin, generating $3.78
- 11:50billion in free cash over the trailing 12 months.
- 11:53Which is absurd. And their return on invested
- 11:56capital or ROIC is a phenomenal 62.5, 5% For you listening.
- 12:01ROIC is simply a measure of how efficiently a company turns
- 12:05investor money into profit, right?
- 12:07If you give a company $100 and their ROIC is 62%, they are
- 12:11generating $62.00 of operating profit from your hundred.
- 12:15That is elite capital efficiency.
- 12:16It really is, and the research points to a different metric to
- 12:20justify the value, one that's way better suited for growth,
- 12:23the Peter Lynch PEG ratio. Peter Lynch.
- 12:27Yeah, the legendary mutual fund manager, he realized that
- 12:29looking at a PE ratio in isolation penalizes fast growing
- 12:33companies. So he divided the PE ratio by
- 12:35the companies earnings growth rate.
- 12:36And Apple Vens PEG ratio is .82 right?
- 12:39Anything under 1.0 is generally considered undervalued by growth
- 12:43investors, so relative to their 32.2% net income growth, the
- 12:47stock is historically cheap. They are growing earnings faster
- 12:50than the market is pricing those earnings.
- 12:52And this exact dichotomy has completely split the smart
- 12:54money. The sources show how divergent
- 12:57the institutional strategies are right now.
- 12:59It's wild to see. Value investor Stephen Mandel,
- 13:02who is known for finding compounding value, increased his
- 13:05position by 88% during AQ One dip.
- 13:09Because he looks past the book value, he sees the elite free
- 13:12cash flow yield on that PEG ratio.
- 13:15Right. But conversely, Ray Dalio's
- 13:17Bridgewater cut their position by 98.7%, which indicates a
- 13:23massive macro driven portfolio rebalancing, perhaps seeing the
- 13:27exact fragility I just mentioned.
- 13:29When you have two financial Titans taking exact opposite
- 13:32sides of a trade, you know the underlying valuation narrative
- 13:35is wildly complex. For sure, it forces you to look
- 13:38forward. If traditional metrics look
- 13:41backward and the value of this company is entirely dependent on
- 13:44its future cash flow, we need to look at the discounted cash flow
- 13:47models, the DCF models provided in the research.
- 13:50OK, let's talk about the DCF models.
- 13:51ADCF model is basically a mathematical attempt to
- 13:54calculate what a stock is worth today based on a prediction of
- 13:58all the cash it will ever generate in the future.
- 14:00Right, discounted back to its present value.
- 14:01Because a dollar today is worth more than a dollar tomorrow.
- 14:04Exactly. And the DCF models in this 53
- 14:07file package are wild. They showcase the extreme
- 14:11variance of tech valuations. Perfect.
- 14:13They really do. They run a conservative slow
- 14:15death scenario. Yeah.
- 14:16This model is a catastrophic 22% annual decline in cash flow for
- 14:21a decade. Ouch.
- 14:22Yeah, in that scenario, the stock is mathematically worth
- 14:25just $21.31 today. Compare that to its current
- 14:29$350.00 price tag. That's a huge drop.
- 14:32Massive. Then they run a nominal model to
- 14:35justify a 30% margin of safety. To get a buy price of $436, you
- 14:40have to assume they will grow free cash flow by 25% every
- 14:44single year for five years, right?
- 14:46And then tapering down to 15% growth for the five years after
- 14:49that which? Requires near flawless execution
- 14:51and a perfectly stable environment.
- 14:53But the environment is anything but stable.
- 14:54Here's where it gets really interesting.
- 14:56To hit those massive cash flow numbers, they have to navigate
- 14:59some terrifying structural risks.
- 15:01Oh, absolutely. Even with their massive data
- 15:03mode, Aklavin doesn't own the smartphones.
- 15:07They don't own the operating systems.
- 15:09They are entirely at the mercy of ecosystem gatekeepers.
- 15:14The sources specifically cite Apple's App Tracking
- 15:17Transparency, Google's Privacy Sandbox and the EU's Digital
- 15:21Markets Act. These are permanent existential
- 15:24threats. Right, and let's explain how
- 15:25this actually works mechanically for you.
- 15:27Listening AI algorithms like Axon require a constant, massive
- 15:33diet of user behavior data to know which ads are effective.
- 15:37They have to feed the machine. Right, they need to know that
- 15:39you looked at an ad for a puzzle game, clicked it, downloaded it,
- 15:42and then spent $3 on in game hints.
- 15:45Exactly. If Apple updates its App
- 15:47Tracking Transparency rules to completely block the mechanism
- 15:50that connects your ad Click to your app install, the AI goes
- 15:53blind. And if the AI goes blind, the
- 15:55prediction accuracy plummets. If accuracy plummets,
- 15:58advertisers stop bidding high prices in the Max auctions and
- 16:01revenue collapses. So my question is, how does
- 16:04management navigate a landscape where Tim Cook or Sundar Pichai
- 16:08could functionally turn off their business model with a
- 16:10software update? This raises an important
- 16:13question about the structural foundation of the business
- 16:15itself, which extends into governance and diligence risks.
- 16:19OK, governance risks. To navigate an environment that
- 16:22volatile a company requires absolute agility, and the
- 16:27sources point out that CEO Adam Fourier has engineered exactly
- 16:30that by controlling 61.6% of the company's voting power.
- 16:34Wait, 61.6%? Yeah, that is an absolute
- 16:38supermajority, but they have massive institutional ownership
- 16:41by heavyweights like Vanguard, BlackRock and KKR.
- 16:44It doesn't matter really, the company operates on a dual class
- 16:47share structure. The shares the founders hold
- 16:50have vastly more voting weight than the shares the public or
- 16:53the institutions hold. Oh wow, so even if BlackRock,
- 16:56Vanguard and every other retail investor teamed up, they
- 16:59mathematically cannot outvote him?
- 17:01That is incredible centralization of power.
- 17:03It really is. At the recent annual meeting, a
- 17:06shareholder proposal was introduced for separate class
- 17:08voting. OK, what would that do?
- 17:10This would have given public shareholders more granular
- 17:13visibility and influence over major corporate decisions.
- 17:17It failed massively, with 119 million votes for it and
- 17:21696,000,000 against. The dual class structure keeps
- 17:25control heavily, heavily concentrated at the very top.
- 17:29Which introduces serious diligence risks.
- 17:31In fact, the report impartially notes an SEC investigation that
- 17:35remained active as of February 2026.
- 17:38Right now we need to be clear, the source is very careful to
- 17:41define this strictly as an unresolved diligence risk, not
- 17:45evidence of any adjudicated violation.
- 17:47We are simply conveying the information contained in the
- 17:50risk matrix here. Exactly, it's just a data point
- 17:52in the research. But it forces you to look at the
- 17:54broader picture. Is this extreme centralization
- 17:57of power and the opacity that comes with it simply a necessary
- 18:01trade off for algorithmic dominance in today's tech
- 18:04landscape? It's the classic founder LED
- 18:06tech dilemma, right? You are betting on the jockey
- 18:08just as much as the horse and in this case, the jockey holds all
- 18:11the reins, all the carrots and all the sticks.
- 18:14But you have to look at the track record.
- 18:15So far, the jockey has executed A flawless pivot.
- 18:19The management team recognized early on that owning the
- 18:22infrastructure of the mobile ad economy, being the digital toll
- 18:25booth, was infinitely more profitable and scalable than
- 18:29trying to churn out hit mobile games year after year in a super
- 18:34fickle consumer market. That's a really fairpoint.
- 18:37The governance structure is exactly what allowed them to
- 18:40make that ruthless, aggressive pivot without getting bogged
- 18:43down by activist investors demanding short term dividends
- 18:46or questioning the sale of their gaming studios.
- 18:48Right if. They had a traditional board
- 18:50structure, constantly answering to a diverse group of
- 18:53shareholders with different timelines.
- 18:55They might never have successfully transitioned from a
- 18:57gaming company to an AI ad tech giant.
- 19:00Which brings us back to you, the learner trying to wrap your
- 19:03head. Or wine, modern valuations and
- 19:05how to evaluate a company's intrinsic worth.
- 19:08The App Levin story proves that competitive moats in the 21st
- 19:11century are no longer built with bricks and mortar or factories
- 19:15or railroads. No, not at all.
- 19:17They are built with compounding data advantages, aggressive
- 19:20technological pivots, and ruthless operating leverage.
- 19:23They have a gross margin of nearly 89%, not because they
- 19:27manufacture the best product, but because they own the
- 19:30transaction flow. They own the invisible highway.
- 19:33It is a profound shift in economic reality.
- 19:36If you stick rigidly to the valuation rule books written in
- 19:391950, looking only at tangible book value, you will completely
- 19:44miss the most dominant cash generating machines of the
- 19:46modern era. You really will.
- 19:48You have to understand the mechanics of the algorithm, the
- 19:51data mode and the structural risks of the ecosystem they
- 19:55operate within. And Speaking of understanding
- 19:57those mechanics, I want to leave you with one final mind bending
- 20:00detail from the source's discounted cash flow analysis.
- 20:03Oh, this part is crazy. It is.
- 20:05We talked about those complex DCF models and how they attempt
- 20:08to predict the future. Well, when the researcher was
- 20:11building those models, they looked back at App Levin's
- 20:13actual historical average free cash flow growth rate over the
- 20:17past 10 years to establish a baseline.
- 20:19That actual historical average was 89% year over year growth.
- 20:24Which is an almost unfathomable metric to sustain for a decade.
- 20:28But here is the crazy part. The researcher had to
- 20:31artificially constrain their nominal predictive model.
- 20:35Because if you didn't. Because if you simply followed
- 20:37the math right and projected that historical 89% growth rate
- 20:41forward for another 10 years, Apple ovens annual free cash
- 20:45flow would reach $2.2 trillion by year 10.
- 20:49Trillion with AT. Yes, $2.2 trillion in pure free
- 20:54cash flow in a single year. That is just.
- 20:57It's hard to even picture. That is a figure exceeding the
- 20:59entire gross domestic product of most sovereign nations.
- 21:03It just leads you to ponder how AI hyperstaling and the
- 21:06compounding nature of these algorithms might literally break
- 21:09the mathematical bounds of traditional financial modeling.
- 21:12They absolutely break the math. The models simply aren't built
- 21:15to comprehend a business that requires almost zero capital to
- 21:18double its size. The treasure map isn't just
- 21:20written in invisible ink, you know.
- 21:22The treasure chest itself is expanding faster than our
- 21:25ability to measure it.