Adam Foroughi, Applovin CEO: Surviving a 92% Drawdown, Ads as ML 1.0 & the $50B Game Ad Market

1. Key Themes

Advertising as "ML 1.0" — the Original Application Layer for Deep Learning

Foroughi frames advertising not as a marketing function but as the first real-world proving ground for the deep learning technology now powering AI broadly. This reframes AppLovin as an applied AI company rather than an ad-tech company.

"Advertising is like ML 1.0, but really was the first implementation of all these technologies that now are driving AI today... a lot of the research that's being done in the space in the large language model space can port to recommendation systems. And vice versa." 00:03:34

The Real Value of Advertising is Discovery, Not Closing Loops

Foroughi draws a sharp distinction between "bottom of funnel" search advertising (which merely intercepts existing intent) and "discovery" advertising (which creates net-new economic activity). He argues this is why AppLovin and Meta are structurally different from — and less threatened by — LLM-based search ads.

"When you show a consumer an ad for something that they had no idea existed, they didn't know they needed to buy... this is what makes Meta so amazing in their ad business and what we aspire to do. You create that discovery moment... you create economic expansion." 00:07:10

Mobile Gaming Ad Inventory is a Vastly Underestimated, ~$50B Market

The scale of the in-game advertising ecosystem is presented as a hidden, underappreciated market comparable to the early days of social advertising.

"We disclosed last January... that on our own platform, there was $11 billion a year of ad spend. Since then, we've grown 60% year over year, roughly. And so if you gross that up... you get $20 billion... there's probably about $50 billion of advertising being spent every single year in this mobile gaming ecosystem." 00:01:40

Public Market Mispricing Can Be a Gift, Not a Crisis

AppLovin's near-total collapse (down 92%+ from IPO) is used as a case study in how a founder can exploit market irrationality via aggressive capital allocation rather than trying to talk the stock up.

"We had no demand and a lot of supply... the multiple went from fairly high... to something that was absurdly low, sub four times... guess what? We generated a ton of cash. Let's start buying our own stock. Let's become our best investor." 00:11:14

Privacy Regulation, Once Codified, Becomes a Solvable Engineering Problem

Rather than viewing Apple's ATT and privacy regulation as existential, Foroughi frames regulatory clarity as ultimately beneficial — it lets deep learning models adapt and consumers still end up wanting relevant ads.

"You want the regulations to be clear. So once they're clear, technology can deal with them... What's happened since a lot of the privacy noise is a lot of calm. The rules were written. Technology companies have adapted. And deep learning networks are really powerful." 00:16:30

Agentic Commerce Will Be Real But Overestimated by Silicon Valley

Foroughi pushes back on the assumption that AI agents will replace discovery-based shopping, arguing that most consumers are not "extremely online" and enjoy the shopping process itself.

"I think we really over-index on the Twitter-verse and forget that the average shopper is not that... a dopamine hit from going through it is what they enjoy." 00:19:47

Winning Against Giants (Meta, Google) Through Focus and Leanness

AppLovin's competitive strategy against trillion-dollar incumbents is framed around narrow focus, speed, and paranoia rather than resource scale.

"We never think we won. We think every day we wake up and we're probably going to get screwed right now and we better work hard... there's this ability to take on giants if you're very focused, you remain lean, and you can just move faster than them." 00:20:28

2. Contrarian Perspectives

Agentic Commerce Is Overhyped for Mainstream Consumers

While much of Silicon Valley assumes agents will disintermediate discovery shopping, Foroughi argues the "New York Times audience" — not the "Twitter-verse" — represents the real shopper base, and that this majority actively enjoys the browsing/comparison process, making automation undesirable even when it's objectively more efficient.

"If you told them after the fact, hey, an agent could have done this for you and saved you 20%, I don't think that matters on a $50 transaction... I think we really over-index on the Twitter-verse and forget that the average shopper is not that." 00:19:21

Advertising Companies Do Not (Meaningfully) Track Location or Conversations

Against the popular "your phone is listening to you" conspiracy, Foroughi flatly denies this is technically or economically realistic, attributing the "creepy ad" phenomenon instead to trackable digital behaviors (searches, browsing) combined with social-graph adjacency.

"I don't think advertising companies can track locations. So we don't track location at all. It's a really heavy concept to track people's precise location to then render an ad... Not realistic." 00:09:00

Stock Price Collapse Was Not a Signal About the Business — It Was a Signal About Investor Quality

Rather than accepting the market's judgment during the 92% drawdown as informative about the company's health, Foroughi argues the price action reflected purely mechanical supply/demand dynamics from a bad COVID-era IPO investor base — while the underlying business (EBITDA) was actually strengthening.

"We went out in 2021, $600 million of EBITDA. $28 billion market cap... And then in 2022, the stock went down literally every day. We got to about a $3.8 billion market cap. And that year, we did a billion dollars in EBITDA." 00:10:47

High Margins Don't Necessarily Invite Destructive Competition

Conventional wisdom says 84% EBITDA margins should attract competitors willing to undercut on price. Foroughi argues that hasn't happened and won't, because the technical moat (proprietary data + model scale) is far more durable than margin alone would suggest — comparing it to why Anthropic isn't easily out-competed despite obvious financial incentive for rivals to do so.

"Because these technologies are really complex. And if you can innovate and you have differentiated data, you can build an advantage. I mean, by that token, Anthropic shouldn't be running away with the large language model space." 00:22:41

3. Companies Identified

AppLovin — Mobile advertising/marketing platform embedded across ~100,000+ mobile games, monetizing via deep-learning-driven ad recommendation and now expanding into e-commerce advertising. Mentioned as the episode's central subject and, per Friedberg, "of all those thousand plus IPOs, the number one most valuable."

"Our EBITDA margins I think are number one in the market. It's 84%." 00:21:33

Meta (Facebook/Instagram) — Held up repeatedly as the gold standard for discovery-based advertising and the model AppLovin aspires to emulate.

"If you talk to most people who shop today, most of their shopping recommendations are coming from Instagram. The ads have become very much like content." 00:04:46

Google — Referenced as the dominant "bottom of funnel" search advertiser, and as a peer whose engineering resources AppLovin nonetheless out-competed in its specific niche.

"That ads model is almost going to exclusively compete with the Google search business." 00:06:12

OpenAI / ChatGPT — Mentioned as building an ad product amid a freemium-driven need for monetization, discussed as a comparison point for how LLM-based advertising might evolve.

"Chat GPT has already said they're going to make it free." 00:05:59

Anthropic — Cited as an example proving that even massively resourced incumbents (Google, Meta) can be outcompeted by focused innovators with differentiated technology/data — used as an analogy for AppLovin's own market position.

"By that token, Anthropic shouldn't be running away with the large language model space. But the power of a model that then reaches a point of scale and gets adopted by a large scale community becomes something that is a moat." 00:22:41

Bending Spoons — Referenced by Friedberg as a company pursuing an aggressive acquisition strategy of slower-growth businesses overlooked by venture capital, drawing a comparison to AppLovin's earlier games-studio acquisition strategy.

"We have Bending Spoons coming on today to talk about their aggressive acquisition of not bad businesses, but let's call them slower growth businesses that maybe Venture isn't interested in." 00:17:32

Yahoo — Cited as evidence that a large, "unsexy" but massive user base still exists and represents AppLovin's real target demographic.

"There's still a ton of people using Yahoo properties every single day." 00:19:21

4. People Identified

Adam Foroughi — Founder/CEO of AppLovin. Identified as an under-the-radar operator who built one of the most profitable ad businesses in tech while avoiding media/investor attention, and who navigated a 92% stock collapse into a subsequent ~$250B market cap recovery through disciplined capital allocation (buybacks) and internal culture management.

"Adam is probably the best founder no one's ever heard of." 00:00:00 "We ended up going from nine to $750 a share in a matter of two and a half years." 00:15:17

AppLovin's China/Beijing engineering team — Singled out specifically for excellence and humility, described as a core competitive advantage.

"Chinese people are very humble. They're very, very hardworking. They're very sharp... when I sit in a room with some of the people on my team, I know I'm probably the dumbest person in that room." 00:23:08

5. Operating Insights

Turn a Market Crisis Into a Buyback-Driven Compounding Machine

When faced with irrational public market pricing, Foroughi's response was not to spend cycles courting investors but to redirect cash flow entirely into share repurchases, effectively making the company its own highest-conviction investor during the trough.

"I'm not going to talk to investors at all anymore. They're not buying our stock. It's a waste of time. But guess what? We generated a ton of cash. Let's start buying our own stock... we bought roughly $6 billion of the company's stock, retired 20% to 25% of the shares outstanding. At peak, that $6 billion was worth over $50 billion." 00:11:43

Extend Performance-Based Equity Deep Into the Organization During Crises, Not Just at the Top

Rather than only protecting executive incentives during the drawdown, Foroughi pushed a performance stock plan down to key non-executive employees to keep the team aligned and motivated through the "us against the world" period.

"We implemented a performance stock plan, which typically goes to CEOs. But we did it across key people in the company and said... if you dig in and we recover, you're going to make a ton on the upside." 00:13:15

Use Proprietary Assets as a Bootstrap, Then Divest Once the Flywheel Works

AppLovin bought its own game studios purely as a temporary data-generation mechanism to train its first deep learning model — not as a permanent strategic bet — and exited once the model succeeded and third-party publishers began supplying data directly.

"We bought them originally as a data play. When we built our first deep learning model, we needed to have data to train it. And game studios don't tend to want to share data to third-party companies. So we bought our own studios. We seeded the training data in our first model... Once we started doing that, third parties were coming in and that was that." 00:17:59

One Well-Timed Re-engagement With Public Investors Can Reprice a Company Overnight

Foroughi's decision to re-emerge and simply confirm the company's survival and performance — after a long self-imposed IR blackout — triggered an outsized repricing, indicating that information asymmetry/neglect, not fundamentals, was the binding constraint.

"I said, I'm going to start talking to investors because market cap's getting high enough... in that week, the stock went from 80 to 150. And I think it was like 28 billion to 55 billion." 00:14:10

6. Overlooked Insights

The Model Transition (Regression → Deep Learning) Was the Real Turning Point, Not the Buyback

While the buyback narrative dominates the conversation, Foroughi mentions almost in passing that the actual fundamental inflection was a quiet technical migration from a regression model to a deep learning model launched in April 2023 — before anyone was paying attention. The stock recovery followed performance which followed this architecture shift, not the reverse. This is arguably the single most important fact in the episode, yet it's delivered in one sentence sandwiched between stories about stock price.

"We went from ML 1.0... to ML 2.0. We went from a regression model to a deep learning model. And the outcome was we're driven by our advertising algorithm. The better it works, the better advertiser return is on our platform... So the company just started growing really quickly." 00:13:43

Advertisers Are Effectively Buying Consumers on a Fully Performance-Based Arbitrage Model

Buried in the margin-leakage discussion is a description of AppLovin's actual unit economics that reveals the business is structured as a real-time arbitrage on consumer acquisition cost versus transaction value — closer to a financial trading operation than traditional ad-serving — which explains both the extraordinary 84% margin and the durability against margin-competing rivals.

"Advertiser comes into our platform and they have a transactional model... We give them an arbitrage. They buy the — they from us are buying the consumer. That consumer transacts. And they cover the cost of the consumer immediately." 00:21:33