AppLovin CEO says ads are ML 1.0, pushing game UA toward model-based buying

In an interview published on the All-In Podcast’s official feed, AppLovin chief executive Adam Foroughi discusses “Surviving a 92% Drawdown, Ads as ML 1.0 & the $50B Game Ad Market,” and says the company built its mobile gaming advertising platform without early venture funding. This is single-source — a media interview released by the All-In Podcast, with no accompanying filings or corroborating disclosures — and the headline figures are presented as topics of discussion, not audited data. For marketing, growth and procurement leaders in gaming, the claim that “ads are ML 1.0” is the material point: if the ad market is defined by model performance rather than media placement, the way budgets are bought and sold changes. All-In Podcast

If ads are framed as machine learning systems, buyers will price the model, not the media

Treating advertising as “ML 1.0” recasts the transaction from buying inventory (placements, impressions) to buying a model’s ability to predict and deliver outcomes for a specific app’s user acquisition or monetization goals. That has two procurement implications. First, the seller’s differentiation becomes its training data and feedback loops — not the size of its network alone. Second, pricing gravitates toward variable, performance-linked structures that expose buyers to spend volatility but make model quality auditable. For a head of growth at a game studio, this means fewer hour-long media plans and more runbooks that stress-test model performance on a defined cohort, with controlled spend, over a fixed period, and clear stop-loss rules. It’s also a shift in accountability: creative, data, and procurement have to agree to a contract where “what we’re paying for” is the model’s incremental lift on defined outcomes, not the cheapest CPM. All-In Podcast

The $50bn game ad market figure is a big stake — but the denominator is missing

The episode title references a “$50B Game Ad Market,” but does not define whether that refers to global in-app game spend, performance budgets only, or any mix of brand spend routed through gaming channels. Without a defined perimeter (geography, channels, formats, gross vs. net of platform take rates), buyers and investors can’t benchmark share or growth. If you’re a CMO or CFO weighing platform lock-in, the question becomes: where, precisely, is that $50bn moving — into which exchanges, SDK networks, and publisher groupings — and what portion is effectively toll to intermediaries versus working media? Until a clear basis is disclosed by the source, treat the number as directional framing, not a procurement input. All-In Podcast

“Surviving a 92% drawdown” is a volatility warning to buyers as much as to investors

A 92% drawdown, mentioned in the episode title, signals severe cyclicality and repricing risk in an ad-tech model. For advertisers, platform volatility has operating consequences. When a platform’s economics swing, discounting can spike and inventory mix can change. That shows up in your campaigns as sharper pacing, more aggressive auto-optimization, or shifting incentive structures. Heads of procurement should respond by hard-coding escape hatches: minimum viable test windows, spend caps tied to observed retention curves, and service-level triggers that allow pausing or reallocating budget when the platform’s own economics appear to be driving bid behavior more than your targets are. In a model-first market, vendor risk management becomes part of the media plan. All-In Podcast

Building without early VC funding reframes the sales motion: prove it in the model, not the deck

The episode summary says AppLovin built “a massive advertising platform within the mobile gaming ecosystem without early VC funding.” In go-to-market terms, that typically correlates with a product that was forced to earn revenue early by solving a narrow, high-frequency problem for a specific buyer: in this case, user acquisition managers at game studios. The commercial lesson for both sides is straightforward. Vendors built under capital constraints often embed fast, programmatic proof loops into their product — APIs and SDKs that let a game push events, get a model’s decisions back, and judge whether to scale within days. Buyers should lean into that product truth: insist on test accounts, access to model reporting at the granularity your data team can validate, and clean separation between model outputs and any promotional credits that blur incrementality. When the sales motion is model-first, the renewal depends on retention curves, not relationships. All-In Podcast

Why the obvious read — “AI makes ads better” — misses the procurement shift

The flattering read is that deeper learning models simply boost return on ad spend for game marketers. That’s incomplete. Model-based advertising moves cost from a fixed line (pre-bought inventory, guaranteed placements) to a variable line whose performance depends on the vendor’s black-box training corpus and inference choices. Marketers don’t just get better outcomes; they inherit model risk. Procurement should therefore negotiate for performance accountability: clear definitions of success, rights to run holdout tests, data usage terms that prevent your first-party data from training rivals’ models, and exit clauses if model performance regresses. If ads are ML 1.0, enterprise buying needs to look more like buying cloud compute — metered, observable, and easy to scale down. All-In Podcast

The skeptical read: model performance claims need a denominator and an audit path

No one in the reported packet is on the record with a third-party validation of the “ML 1.0” claim’s impact on outcomes, and the episode page does not define the basis for the $50bn figure or the mechanics that tie model architecture to pricing. A reasonable skeptic will ask for three things before moving budget: (1) a like-for-like test against your current best-performing channel with a pre-registered measurement plan; (2) explicit data terms on what user and event data trains the vendor model and whether you can restrict cross-customer learning; and (3) transparent pricing that maps spend to observable model decisions (for example, inference bill rates or performance bands) rather than blended take rates. Until those are visible, treat any “AI lifts ROAS” claim as a starting hypothesis. All-In Podcast

What changes next for sales and marketing teams on both sides of the deal

If large portions of the game ad market are priced on model performance, sellers’ revenue teams will need to restructure around customer success engineers and measurement specialists who can run tests, explain model behavior to non-ML buyers, and keep cohorts green through the first 90 days. The classic enterprise account executive who sells a multi-quarter media commitment will be less effective than a team that can land a small test, expose the right diagnostics, and scale on proof. On the buy side, CMOs and CFOs should expect to fold more of the UA budget into contracts with performance-linked fees and clearer variable spend guardrails — and to pull legal into earlier negotiation on data rights, since the model’s training corpus will outlast any individual campaign. The org-chart consequence is that growth, data science, and procurement sit closer together, with campaign performance reported alongside model health, not just spend. All-In Podcast

Signals to watch: disclosures, contracts and who gets hired to sell

Because this is an interview with no supporting filings, watch for how platforms operationalize the claims. Two obvious markers over the next two quarters will be whether leading mobile ad platforms disclose model-level performance diagnostics in their public materials and whether they introduce pricing schedules tied explicitly to outcome bands rather than blended take rates. A third, softer signal sits in hiring: if you begin seeing more “solutions architect” and “incrementality lead” roles in commercial teams, that is a tell that the sales motion has shifted from media selling to model selling. If those signals don’t show up, the safer read is that the rhetoric hasn’t yet forced a change in how budgets are bought and retained. All-In Podcast