How AppLovin Built an Ad Empire Without Owning the Audience

The advertising company that does not own the audience

The most powerful digital advertising businesses usually begin with ownership. Google owns the search box. Meta owns the feed. Amazon owns the store and the checkout. Each company controls a place where attention gathers, then sells advertisers access to it.

AppLovin built something different. It does not own a social network, a search engine or a large consumer marketplace. Most people who encounter its technology have never heard of the company. Yet AppLovin says its MAX platform reaches more than one billion daily active users, while advertisers are spending more than $11 billion a year through its system.

That sounds like an advertising network. The economics suggest something more interesting.

AppLovin sits between the businesses buying customers and the mobile apps selling attention. MAX helps publishers auction their advertising inventory. AppLovin Ads brings demand into that auction. AXON predicts which impression is valuable to which advertiser and how much should be paid for it. The company does not need to own the audience because it has built a privileged position inside the transaction.

The central question is whether that position can become a durable platform beyond mobile gaming, or whether AppLovin has perfected a highly profitable machine inside a market that will prove harder to escape than investors expect.

What AppLovin actually became

AppLovin spent years looking like two companies joined together. One side owned mobile game studios. The other sold the advertising and monetisation technology used by game developers. Owning games gave the company inventory, operating knowledge and data, but it also made the business harder to understand. AppLovin was simultaneously a supplier, a customer and a competitor inside the same ecosystem.

The strategic break came in 2025. AppLovin sold ten game studios to Tripledot Studios for $400 million in cash and an equity stake that represented roughly 20% of Tripledot at closing. The divestiture removed the consumer-app business from AppLovin's continuing operations and left a much cleaner company behind.

Today, substantially all revenue comes from AppLovin Ads, according to its latest regulatory filing. That matters because it changes the way the business should be judged. AppLovin is no longer primarily a collection of mobile games supported by advertising software. It is an advertising system whose original laboratory happened to be mobile gaming.

The distinction explains both the extraordinary margins and the ambition. Games taught AppLovin how to price attention in an environment where performance could be measured quickly. The company is now attempting to carry that skill into e-commerce, subscriptions and other consumer categories with far larger advertising budgets.

The machine has three connected parts

MAX sits on the supply side. Mobile publishers use it to run real-time auctions among advertising buyers and improve the revenue generated by each impression. A publisher does not want to guess which network will pay the most. MAX allows bidders to compete for the opportunity.

AppLovin Ads sits on the demand side. Advertisers define the return they want, and the platform deploys their budgets across available inventory. The advertiser is not buying a fixed placement in the traditional sense. It is asking AppLovin to find customers at an economically acceptable price.

AXON provides the prediction layer. It estimates the likelihood that a user will engage, install, purchase or produce enough lifetime value to justify a bid. AppLovin describes the system as operating across auctions at enormous scale and in microseconds. The better those estimates become, the more confidently the platform can bid for valuable impressions without overpaying for the rest.

The three parts reinforce one another. More publishers create more inventory and more opportunities to observe behaviour. More advertisers create stronger demand and more competition inside the auction. Better predictions allow AppLovin to win the impressions where it sees value that other bidders miss. Successful campaigns encourage advertisers to spend more, which makes the auction more valuable to publishers.

This is why describing AXON as the moat is incomplete. A model can be copied, improved or overtaken. The harder asset to reproduce is the operating loop around it: access to supply, advertiser demand, outcome data and enough transaction volume to keep learning.

The economics are exceptional

AppLovin's latest results look less like traditional advertising technology and more like a software business with unusual pricing power. In the second quarter of 2026, revenue reached $1.92 billion, up 53% from a year earlier. Net income was $1.27 billion, while adjusted EBITDA reached $1.61 billion. Free cash flow was $863 million for the quarter.

The company guided to an adjusted EBITDA margin of 83% for the following quarter. That figure is non-GAAP and should not be confused with net profit, but it still reveals how little incremental cost is required when more advertising spend moves through the same infrastructure.

The most useful detail sits beneath the headline growth. AppLovin reported that net revenue per installation increased 58% year over year in the quarter while installation volume declined 2%. Across the first half, net revenue per installation rose 75% even as installation volume fell 10%.

Growth therefore did not come from simply processing more activity. It came from making each unit of activity more valuable. That could reflect stronger advertiser demand, better prediction, improved pricing or some combination of the three. Whatever the mix, it is the signature of a platform getting better at allocation rather than merely getting larger.

AppLovin also spent $551 million on repurchases and tax withholding connected with vested awards during the quarter. Buybacks can be sensible when cash generation is strong and shares are attractive, but investors should still separate genuine share-count reduction from purchases that offset employee compensation.

Gaming created a cleaner feedback loop than commerce

AppLovin's strength in mobile gaming is not accidental. Games produce frequent, measurable events. A user sees an advertisement, installs an app, opens it, plays, watches more ads or makes a purchase. The time between the original impression and the economic outcome can be short enough for a model to learn quickly.

Commerce is messier. A person may watch an advertisement today, visit a website tomorrow, purchase on another device next week and return a product later. The advertiser may also be running campaigns through Meta, Google, email, influencers and affiliates at the same time. Assigning credit becomes harder because the path to purchase is longer and shared across more channels.

That difference is the heart of the AppLovin investment case. The company has already proved that its system can allocate advertising capital inside gaming. Its larger opportunity depends on proving that the same predictive advantage survives when the feedback becomes slower, noisier and more contested.

Management has been deliberate about opening the system. AppLovin launched its self-service platform by referral in October 2025, then removed the referral requirement in June 2026. For the first time in fourteen years, any advertiser could sign up without an existing relationship with the company.

Opening the platform expands the addressable market, but it also changes the customer mix. Large advertisers can arrive with sophisticated creative teams, clean data and enough budget for the model to learn. Smaller businesses may have weaker creative, limited conversion history and less tolerance for an expensive learning period. A self-service product can remove sales friction without removing those economic constraints.

A closed loop can still depend on someone else's rules

AppLovin's system feels vertically integrated because it connects publishers, advertisers, auctions and prediction. It is not fully independent.

The audience still belongs to third-party apps. Distribution still runs largely through mobile operating systems controlled by Apple and Google. Measurement depends on the data those platforms and publishers permit AppLovin to collect. A policy change affecting identifiers, privacy disclosures, software development kits or attribution can reduce the information available to the model.

AppLovin acknowledges this risk directly in its filings. Apple's tracking-transparency framework has not materially damaged the overall business so far, according to the company, but future restrictions could make its advertising solutions less effective. That is an important distinction. AppLovin may have built the best bidder in the auction without controlling the venue in which the auction ultimately exists.

Publishers also need to believe the auction serves them fairly, while advertisers need confidence that reported returns reflect incremental sales rather than generous attribution. Those incentives are not automatically aligned. The platform becomes stronger when both sides trust the measurement and weaker when either begins to question how value is being assigned.

This makes transparency more than a reputational issue. It is part of the product. Performance advertising works because results can be measured. If participants lose confidence in the measurement, the technical quality of the prediction model matters less.

The valuation is really a portability question

AppLovin closed on September 25 at $310.75, giving the company a market value of roughly $104 billion. The shares were more than 50% below their 52-week high, even though trailing revenue and earnings had continued to grow quickly. That combination makes the stock look inexpensive compared with its recent history.

A lower share price is not the same thing as a lower-risk investment. At roughly 24 times trailing earnings, the market is still assigning meaningful value to future growth. The multiple looks modest only if the current earnings base is durable and the advertising engine continues expanding beyond its original market.

The bull case is not difficult to understand. AppLovin has an asset-light platform, exceptional margins, strong cash generation, a shrinking share count and a model that appears to improve the economic value of each installation. If e-commerce and other consumer categories adopt the platform at scale, gaming may turn out to have been the training ground for a much larger advertising business.

The bear case begins with the same facts. Current margins leave little room for the economics to improve, while the next market is harder to measure than the first. The company depends on platforms it does not control, and advertiser spending can disappear quickly when returns weaken. If AXON's advantage proves less portable outside gaming, AppLovin may remain an excellent business without becoming the much larger business its best expectations require.

That is why the valuation question cannot be answered by comparing the current share price with the former high. The relevant comparison is between the cash flow AppLovin already produces and the amount of future commerce growth embedded in the price today.

My read

The most impressive thing about AppLovin is not that it uses artificial intelligence. Every serious advertising platform now makes some version of that claim. The impressive part is that AppLovin has connected prediction to a market structure where better decisions can be monetised immediately.

MAX gives it a view of supply. AppLovin Ads brings demand. AXON decides where capital should go. When the system works, publishers earn more, advertisers acquire profitable customers and AppLovin keeps a valuable position between them. That is a stronger business than a standalone tool that merely recommends which advertisement to run.

I would still resist calling the moat complete. AppLovin does not own the consumer relationship, the operating system or the app store. Its advantage survives only while the platform keeps producing returns that advertisers can verify and revenue that publishers cannot match elsewhere.

The next phase will be decided outside the market that taught AppLovin how to win. If the company can price a shopper as accurately as it learned to price a game installation, the opportunity is much larger than mobile gaming. If it cannot, the current business remains exceptionally profitable, but the story becomes one of depth rather than endless reach.

That is what makes AppLovin worth studying. It built an advertising empire without owning the audience. Now it has to prove that the machine understands people beyond the game.