APP · Q2 2025 Earnings

Five distinct shifts in posture, all pointing the same direction: management is moving from "prove the model works in gaming" to "scale the model across every digital surface."

The most concrete shift is on advertiser onboarding. For multiple prior quarters, AppLovin constrained new advertiser intake to protect model performance during the e-commerce pilot. That gate is now opening on a defined schedule: "We ended up constraining the advertiser onboarding process for a couple quarters while we made sure that the product was at the level that we wanted to get it to...now we're looking at, in Q4, talking about a referral-based opening...and then opening up the platform entirely to the world in the first half of next year." This is a binary change — closed to open — with calendar dates attached.

Second, management is explicitly planning to spend on customer acquisition for the first time. The prior model was organic and word-of-mouth: agencies and brands found AppLovin. Now: "Once Axon is fully open next year, we plan to begin paid marketing to recruit new advertisers, which will drive predictable, compounding growth." For a company with 81% EBITDA margins to commit to performance marketing signals conviction that incremental advertiser LTV/CAC is structurally favorable — and that the constraint on growth has been demand-side discoverability, not supply-side capacity.

Third, the platform is getting its own brand. Spinning "Axon" out as a distinct product identity — rather than keeping it as AppLovin's underlying tech — implies management sees a TAM large enough to justify separate go-to-market and possibly separate organizational identity: "The opportunity is so big that we will be launching the platform under its own brand, Axon."

Fourth, e-commerce is no longer framed as an experiment alongside gaming. The architecture is dual-model: "the two models are distinctly different because on the one hand, a user is going to a website, and on the other hand, a user is going to an app store...we need to have visibility into consumer behavior across every category." The framing has moved from "can it work outside gaming" to "we are building two parallel optimization stacks."

Fifth, supply is being deliberately diversified beyond mobile gaming inventory: "There's absolutely no reason why we wouldn't want to plug into other properties, even if they're large social networks, music apps, news apps, sports apps, websites." This is a meaningful expansion of the addressable inventory pool and a precondition for the data flywheel management argues will compound model quality each quarter as cross-category consumer behavior data accumulates.