A Late Model Release Cost AppLovin Three-Fifths of Its Value

On August 5, AppLovin reported second-quarter revenue of $1.924 billion, up 53% year over year . Against its own guidance range of $1.915–1.945 billion , that landed roughly $6 million — about a third of one percent — below the midpoint. Adjusted EBITDA of $1.614 billion at an 83.9% margin came in just under the bottom of the guided $1.615–1.645 billion band . The company generated $863 million of free cash flow in the quarter . For that, the market has taken the stock from a 52-week high of $745.61 to trade around $305.80 intraday on September 9, within a few percent of the $297.50 low set in late August, and capitalized the business at $102.7 billion — close to three-fifths of the equity value gone since early October 2025.

The gap between the size of the operational event and the size of the repricing is the argument. AppLovin is not a hyperscaler and not a diversified ad conglomerate: it is a single-engine machine that matches advertiser demand to in-app inventory through an auction run by a recommendation model called Axon, and it earns money only when that model gets better and advertisers respond by pushing more budget through it at their target return. Management's account of the quarter is that the model got better later than usual — the next step up in performance landed just after quarter end — while MAX publisher earnings grew double digits sequentially and the company's share of publisher waterfalls held . My view is that the disclosure supports that account, that the Q3 guide is constructed in a way that makes it a floor rather than a forecast, and that at roughly 15 times next year's expected earnings the market is now paying for a business it has decided is saturated rather than one that skipped a release cycle. I think that is wrong, and I think the composition risk that would make it right — growth now coming entirely from price — is real enough to argue at length and still not decisive.

Anatomy of a Miss That Was Not a Demand Miss

The mechanism matters because it determines whether the miss is repeatable. Revenue at AppLovin decomposes into two observable variables the company discloses in its MD&A: installation volume and net revenue per installation. In the June quarter, revenue rose 53% because net revenue per installation increased 58% against a 2% decline in installation volume ; across the first half the same split was +75% and −10% . Nothing in that decomposition depends on winning new advertisers. It depends on the model producing better matches, which raises what advertisers will pay per outcome, which raises the price AppLovin clears on a roughly fixed pool of impressions.

Under that mechanism, a light quarter is a research-cadence...