AppLovin: 1.6 billion users—you’re one of them

"If you do the right things on the top line, the bottom line will follow."

– Steve Jobs, co-founder of Apple (1955–2011; in an interview in 1997)

AdTech.

Or, advertising technology.

The plumbing behind almost everything you get free on your phone.

For most of the last century, advertising was a real estate business. A page in a newspaper, thirty seconds before the weather, a billboard on the road to the airport. You paid for the space, and whether it worked was faith. Half the budget was wasted, and nobody could tell you which half.

The internet spent fifteen years moving that model onto a screen. Better, because you could count who saw it. Still faith.

Around 2010 it stopped being a negotiation and became a market. When an app loads, the space where the ad will go is auctioned in real time. Advertisers bid, the highest wins, the ad appears. A tenth of a second, billions of times a day.

The winner was whoever knew most about the person on the other side of the screen.

Which is exactly what broke in 2021.

When Apple let users refuse tracking, the identifiers the industry was built on quietly disappeared. Targeting by «who you are» stopped working. What replaced it was prediction: modelling what someone who behaves like you tends to do next.

That is a machine learning problem.

So, the modern advertiser does not choose an audience at all. It hands over three things — a budget, a region, a target return per dollar — and steps back. No longer buying space. Buying outcomes.

Which calls for an AI-driven advertising company, built for mobile.

Enter AppLovin.

Chart 1: The flywheel — distribution feeds data, data feeds the model, the model feeds results

AppLovin’s Flywheel, and Why You Are Already In It

Ever played a puzzle game and been offered another between levels?

Watched a short video for an extra life?

Not?

But this one, my friend, will have caught you: bought something only after the very same product trailed you from app to app for a week, like a puppy that refuses to go home?

Et voilà. That relentless little shadow is the part I quietly admire most. Not luck, not stalking. Just a model that worked out, from millions of people who behaved like you, that you were three days and one price cut from buying.

Sit with that for a second. It did not know you. It had never met you. And it still guessed the exact moment your resistance would break — because ten thousand people just like you had already broken at that same moment. That is the uncomfortable magic of this business: it does not persuade you. It predicts you.

A bit scary, I must confess…

Thus, you have almost certainly been through this auction — in touch with AppLovin, probably hundreds of times. The platform reaches some 1.6 billion daily active users, and almost none of them have heard the name.

Four pieces do the work:

1. AXON is the engine, the model that takes those three inputs and decides who sees the ad.
2. MAX runs the auctions for publishers selling their space.
3. Adjust measures what happened after the click.
4. Wurl extends the whole thing onto television.

Put them in a circle and you have Chart 1. MAX brings the supply and the signal, Adjust reports the outcome, AXON learns from both. Sharper predictions, better returns, more spend, more data, sharper again.

And the model keeps everyone honest: AppLovin earns a share of the spending it drives, when it works. An advertiser who is not getting a return turns the tap off.

In mid-2025 AppLovin sold its own games portfolio. What is left is one software platform, run by roughly 900 people, producing close to $6.8 billion of revenue. No factories, no inventory, no fleet. The heaviest asset in the building is the model. Asset light, par excellence.

Which brings us back to a sentence from 1997.

Apple was around ninety days from running out of cash. Steve Jobs had just walked back into the building he had been thrown out of, and someone asked him what his plan was. He did not talk about cost cuts.

«If you do the right things on the top line, the bottom line will follow.»

He meant something specific.

Get the strategy right, the people right, the culture right, and the product follows. Do that, and the profit comes looking for you. It does not work in reverse. We made that case at length, with Apple and Amazon as the evidence.

And it has a shape (Chart 2).

Nearly every business that ever worked walks the same six stages. It starts up. Then it grows fast and burns money doing it — the red line, below zero, the years most investors avoid.

Somewhere in the third stage the lines cross. Break-even — the most consequential moment in a company’s life, because of what comes next.

Revenue keeps climbing, but the costs are already built and paid for. So the profit line does not merely turn positive. It accelerates.

And notice what the chart does not say. The profits did not produce the revenue. The revenue came first, for years, while the profit line sat underwater and everybody explained why the company would never make money.

AppLovin is that curve with a stock ticker attached.

And once through that curve, things move fast. Sometimes very fast.

Chart 2: The six stages of the business growth cycle

AppLovin’s Six Years of Losses, Then the Switch Flipped

AppLovin is the theory brought to life.

Look at the blue line (net profit margin) in Chart 3, and find the stretch where it sits below zero. That is most of six years. Minus 23.6% net margin in 2020, minus 18.4% in early 2022. Stage two, «hyper growth», red line, under water — exactly where the curve says the years nobody enjoys are supposed to be.

Because AppLovin was doing the unglamorous thing.

It was building the top line and paying for it: Adjust, MoPub, Wurl — the acquisitions that turned it into a real advertising platform. The revenue was there, just not yet big enough to cover a cost base built for a far larger company.

Then, in 2023, AXON 2 shipped. And the lines crossed.

From this point on, the blue line no longer acts like a brake, but rather like an accelerator: minus 0.6%, then eighteen, then forty-two, then 65.8% by this summer. Net margin, not gross. Break-even to stage four in three years.

Nothing was added underneath.

The same nine hundred people, the same platform. New revenue ran through machinery already paid for and fell almost straight to the bottom. That is operating leverage — the reason software, when it works, is the finest business model ever invented.

And the stock?

From about $10 at its 2021 low to more than $700. A seventy-bagger, in four years.

Not because it chased profit, but because it did the right things on the top line — and the bottom line, and the share price, came looking for it. Jobs, drawn as a chart. A prime example.

But every great story needs a villain. And here comes ours.

Valuation.

Since the IPO the median price-to-earnings ratio has been 25.8. Today it is 16.8. The free cash flow yield averaged 2.5%; today it is 4.3%. And the long-term growth estimate sits at 37.4%.

Wait!?

«Patrick», you say, «you are telling me it is cheaper than its own history, and growing at nearly 40%?»

Deep value and hypergrowth in the same ticker. Yes. That is exactly what I am telling you.

So, what is the issue?

Start with what is not the issue. The moat is real, and it has held. Meta and Google have thrown a decade and untold billions at mobile-game advertising and never broken AppLovin's grip on it. The numbers are not the issue either — growth has cooled from 70% to the mid-fifties, a deceleration the market hates, but a number most companies on Earth would kill for.

The issue is one word: gaming.

Because the whole trick so far has been played inside a single arena — advertising within mobile games. Enormous, wildly profitable, and, in the market's eyes, close to saturated. The bull case was never really about gaming. It was about carrying the same prediction machine into e-commerce and connected TV, and turning a big market into an enormous one.

And that is exactly what AppLovin has not yet shown. The e-commerce push is real but barely out of the gate, and the last quarterly call did little to prove the market bigger. Until that proof arrives rests every fear on the same unanswered question.

There is a second, quieter worry too. The model is a black box nobody outside the building can forecast, which loops back to the question we have put to software in general: what happens when AI makes software cheap to build?

A wonderful set of numbers, then. But you know our rule — wonderful numbers are not automatically a wonderful investment.

So what does Mr. Market say?

Time for the «Good Chart».

Let’s dig in.

Chart 3: Quarterly revenue, gross margin and net margin — the inflection after AXON 2

AppLovin’s «Good Chart»: The More the Damage, The Longer the Repair

It is brutal.

From a high above $745, the shares now trade near $315. A drawdown of almost 60%, and the red bars (maximum drawdown) along the bottom of Chart 4 show it has happened more than once.

What triggered the latest leg is almost comic. And typical for high growth stocks.

In August, AppLovin reported revenue up 53% and an 84% EBITDA margin — and missed the midpoint of its own guidance by a hair. The first real miss since it went public. Model improvements, management said, had simply been lighter than usual that quarter.

Tens of billions of market value went out the window on a timing problem.

Which makes sense only once you see what was in the price. A stock compounding at 60% and beating every quarter is not valued on this year’s earnings, but on the assumption the beating continues. The moment that assumption cracks — not breaks, cracks — the multiple comes apart, and a falling multiple can swamp rising earnings for years.

Now the chart, rather than the income statement.

The price has broken below its long-term moving average, and that average (200 day) has rolled over and started to point down. That is trend deterioration. The line of least resistance, which pointed relentlessly up from 2023 through 2025, has changed direction.

That’s a lot of damage to repair.

But there lies an opportunity.

If the trend turns, the setup is rare: 16.8 times earnings against a 37% growth estimate, a depressed multiple re-rating while earnings compound underneath. That is how the biggest returns are made, and why we watch stocks like this rather than ignore them.

But a 60% drawdown is not repaired by a good week. Everyone who bought higher is waiting to get out at break-even, and that supply has to clear first. A base has to form. It takes time, and proof.

The «Good Story» is intact. Our screening still ranks it among the highest quality businesses in the whole database.

But Mr. Market’s verdict on exactly those stocks is clear.

The more the damage.

The longer the repair.

Chart 4: AppLovin’s stock price since IPO with long-term moving average and maximum drawdown