Today, AppLovin Corporation stands among the most valuable technology companies in the United States, poised to generate roughly $60 billion in cash this year, with its advertising platform now processing close to $200 billion in annual ad spend. Yet few recall that less than three years ago, the company's shares had plunged 92% from their peak.
At the 2026 All-In Summit, founder and CEO Adam Foroughi recently offered his most comprehensive account of that turbulent journey—covering everything from the post-IPO crash and AI transformation to the daring share buyback strategy that effectively turned the company's fortunes around.
Foroughi was blunt during the discussion, stating: "Advertising is ML 1.0—the first commercial realization of all the technologies that power AI today." And it's along this technological curve that AppLovin made the leap from mobile gaming ads to creating purchase intent in e-commerce.
An Ad Empire Hidden Inside a Billion Mobile Games
AppLovin Corporation isn't exactly a household name, a fact Foroughi concedes with a laugh: "That silly name didn't do us any favors either."
But its scale is anything but trivial. Foroughi explained during the interview that AppLovin is at its core a mobile gaming advertising firm—helping developers monetize their traffic. He shared a few telling numbers:
More than one billion people worldwide play casual mobile games daily. In early 2024, annual ad spend on AppLovin's proprietary platform had already hit $11 billion, growing at roughly 60% per year since—today that figure approaches $20 billion. The broader mobile gaming advertising ecosystem, he estimates, is worth approximately $50 billion.
"Not long ago, social media advertising was a $50 billion opportunity," Foroughi noted. "This space is expanding quickly."
What's more, AppLovin's EBITDA margin currently sits at an industry-leading 84%.
The Darkest Hour: $1 Billion in Profit, But Only $3.8 Billion in Market Value
The company's present-day success stands in stark contrast to where it stood three years ago.
In April 2021, AppLovin Corporation went public at a valuation of roughly $28 billion, briefly surging to $40 billion. That was followed by a cliff.
"In 2022, the stock fell almost every single day," Foroughi recalled. "We eventually dropped to a market cap of $3.8 billion." During that same year, the company generated $1 billion in EBITDA. The valuation multiple had collapsed from peak levels to under 4 times EBITDA.
The reason behind the crash, he explained, came down to the shareholder base: "Your price in the market is determined by the quality of your investors."
With the 2021 IPO wave, blue-chip investors simply didn't have the bandwidth to research a company with an odd-sounding name. Early private equity holders and employees were eager to cash out, creating a flood of supply against scarce demand—and the stock cascaded downward. During that period, Foroughi fielded worried calls from family and friends. "They asked me, 'Are you okay? Are you thinking about doing something drastic?'" he said. "I told them, we started with a penny, the stock is still around $10—it's still up a lot. But as CEO, you quickly realize your team is fielding the same calls—and they don't have your conviction or your stake."
The Self-Rescue: Stop Pitching, Start Buying
When the market turned its back, Foroughi made an unconventional call.
"I told my team, I'm not going to meet with investors anymore. They aren't buying our stock—meeting them is a waste of time. But we're generating enormous cash flow. So let's buy our own shares and become our own best investor."
AppLovin launched an aggressive buyback program. The result: approximately $6 billion in cumulative repurchases, retiring roughly 20%–25% of outstanding shares. At peak prices, that $6 billion buyback is now worth over $50 billion.
In parallel, the company introduced a performance stock plan extending to core employees—not just the CEO. Foroughi's reasoning: "We know this is tough. You think you have a house, and now it feels like it's gone. But if you hang in there, and we bounce back, you're going to make a lot."
The conviction behind his confidence came from a pivotal technical upgrade. In April 2023, AppLovin upgraded its advertising algorithms from regression-based models (ML 1.0) to deep learning models (ML 2.0). The effect was immediate—advertisers saw dramatically improved returns, and the platform's growth accelerated.
Still, the company remained silent externally. It wasn't until September 2023 that Foroughi made his first trip to New York to meet investors, when the stock had already rebounded to around $80.
"That week, the stock went from $80 to $150. The market cap went from $28 billion to $55 billion."
"All because I went to New York and said, 'Hey, we're still here. We survived.'"
Advertising as 'ML 1.0': Unpacking the Technical Logic
Foroughi has a clear framework for understanding AppLovin's technological positioning.
"Advertising is ML 1.0—it's the earliest commercial implementation of the technologies that are now driving AI," he said.
He distinguishes between two fundamental types of advertising. Search ads operate at the bottom of the funnel: users already know what they want and search to complete the transaction, a need that large language models are now directly displacing Google. Discovery ads, by contrast, operate at the top of the funnel: users don't yet know what they want, and the ad creates the demand. "You show them a recommendation and they think, 'Wow, that looks great, I'm going to go buy it'—that's what drives Meta's ad business, and it's what we're trying to do."
He argues that discovery advertising creates genuine economic value, whereas search advertising merely accelerates a transaction that would have occurred anyway.
On whether AI agents will disrupt the ad industry, Foroughi remains measured: "A subset of people will use agents to optimize recurring purchases—like auto-renewing supplement subscriptions each month. But discovery platforms work differently. The average shopper isn't the type of person glued to Twitter chasing the latest tech. Our audience is more like The New York Times reader. There are still vast numbers of people using Yahoo products every single day."
Why a Smaller Company Can Outmaneuver Meta and Google
With an 84% EBITDA margin sustained for years, how has AppLovin managed to stay ahead of two advertising behemoths that employ arguably the best engineers in the world?
Foroughi's answer is refreshingly simple: "We never think we've won. Every morning we wake up feeling like we might fail today—so we work extraordinarily hard."
His competitive playbook emphasizes focus, leanness, and rapid iteration: "If you stay extremely focused and lean, you can move faster than the giants."
Responding to scepticism about whether high margins will eventually be competed away, he counters: "These technologies are deeply complex. If you keep innovating and own differentiated data, you build a real moat. Look at Anthropic—by all logic, it shouldn't be running that fast in large language models. But when a model reaches scale effects and mass adoption, it forms a barrier that's incredibly hard to cross."
He also highlighted AppLovin's engineering presence in Beijing: "The Chinese people are incredibly hardworking and brilliant. Being able to work with them is one of this company's advantages. I sit in the room and often feel like I'm the dumbest person in the room—and honestly, that's what gets me excited to come to work every day."
Full Interview Transcript
Adam Foroughi, AppLovin CEO: Surviving a 92% Drawdown, Advertising as ML 1.0, and the $50 Billion Gaming Ad Market
September 21, 2026 · All-In Summit 2026
Chapter One: Adam Foroughi Joins the Besties!
Host: Adam is perhaps the most accomplished founder you've never heard of. There's an ad platform hidden inside a hundred thousand mobile games, quietly becoming the go-to choice for e-commerce brands, surpassing even Facebook ads. Among over a thousand IPO companies, the highest market cap belongs to AppLovin. The mindset of a founder must be one of wanting to win. They're going to generate around $6 billion in cash flow this year. In an irrational world, people assume you must be cheating—not realizing you've actually built one of the coolest technologies out there. Please welcome Adam Foroughi!
Host: Adam, thank you for being here. We thought this would be fascinating because you rarely make headlines. You run your company almost entirely outside the media spotlight—few interviews, little public commentary. But this is a remarkable business. Can you walk us through what AppLovin does and help us understand your market?
Adam Foroughi: Sure. I think the reason we were able to grow this large without early venture capital is precisely because we kept our heads down and operated quietly. Of course, the slightly ridiculous name didn't help.
At the core, we're an advertising company that helps mobile game developers monetize their traffic. A lot of people don't realize how massive the mobile gaming world has become—over a billion people play casual mobile games every day. These are adults, household decision-makers. The market opportunity is hard to grasp.
In January of last year, we disclosed that annual ad spend on just our platform had reached $11 billion. Since then we've maintained roughly 60% year-over-year growth—today that's nearly $20 billion. And we're not the only player. Other ad companies also monetize this space. All together, you're looking at roughly $50 billion in annual ad spend across the mobile gaming ecosystem.
Not long ago, the social media ad market was a $50 billion space. This sector is growing rapidly, and the audience actively chooses to watch ads—often for rewards. That mechanism creates real opportunity for "intent generation."
For most of our history, we've been creating intent through gameplay—drawing users naturally from one game to the next. What genuinely excites investors, and us, is that deep learning models are now powerful enough that we can use that same real estate to guide these adult users into consumer shopping. That opens up far larger economic sectors and creates much greater value. That's why our team is so passionate about what we do.
Chapter Two: Discovery vs. Search Advertising, and Is Your Phone Really Listening?
Host: The first wave of internet advertising ignited many of the foundational technologies that spread across the globe—what Google did with AdWords, AdSense, applied semantics, and so on. Is this generation of internet advertising having the same effect? Is something fundamental emerging here for the entire internet?
Adam Foroughi: Yes. Advertising is ML 1.0—the earliest implementation of all the technologies driving AI today. Large language models now create more economic value in society, but advertising remains an extremely profitable application of deep learning.
Recommendation systems are structurally different from large language models, yet they follow similar trajectories in many ways. A great deal of research from LLMs transfers to recommendation systems, and vice versa. Many of today's LLM researchers started their careers working on ad systems. The two fields are deeply interconnected.
The beauty of advertising—and any ad business—is that when you train a model, you're predicting a future outcome—whether an ad, a recommended post, or a piece of content—and you can immediately monetize the value of that prediction.
Host: Is there a universal behavioral pattern in how humans respond to ads in 2026 versus 2006? Is there a predictable trajectory?
Adam Foroughi: It's interesting. I entered this industry in 2005, and advertising back then was terrible—pure spam, with insufficient technology. Facebook then did something clever: it realized that combining all available data with strong technology could make advertising incredibly precise. Today, most shoppers will tell you their purchase inspiration comes from Instagram. Ads have become almost indistinguishable from content itself.
In our space, people genuinely enjoy the ads we show. You might think they wouldn't, but we see massive engagement with mini-game ads—they appear inside other games, and users actively try the previews because the technology is powerful enough to show everyone something genuinely relevant.
Host: There are many concerns about AI disrupting ad networks, particularly for Google. OpenAI now has an advertising product, and I'm sure you've been following that closely. When people start having five or six rounds of queries with chatbots, what will advertising look like? Obviously 95% of users won't pay $20 a month for the technology—they'll expect it free. ChatGPT has already said they'll be free. You tell us how their ad play unfolds. Will it be fewer ads per interaction but more usage overall, or will it beat Google's ad model outright?
Adam Foroughi: Advertising has two dimensions. There's bottom-of-funnel advertising—where consumers broadly know what they want and are researching to complete a purchase. That's Google's search business. If I want to buy leather shoes, I used to search on Google, research, and ads would guide me where I needed to go. Today, you can close that same loop with a large language model. So that form of advertising directly competes with Google's search business.
Where we operate is showing ads to users who haven't yet expressed intent. We create something that didn't exist before—show them a recommendation and make them think, "Wow, that looks great, I'm going to buy one." That's also the underlying logic powering Facebook's ad business.
Transactions facilitated by search or LLMs were going to happen anyway. Even if Google ads had never existed, as long as search existed, that transaction still occurs—just via a different path. So there isn't much incremental economic value being created there.
But when you show a consumer an ad for something they didn't know existed, didn't know they needed—that's pure discovery, a completely discovery-based experience. That's why Meta's ad business is so powerful, and it's the direction we pursue. You create that "discovery moment." Not only is it a pleasant experience for the consumer—they're excited about what they bought, waiting for the package, eager to open it—but you're genuinely creating incremental economic value.
Host: What about the arms race question? Some say: I mention something at a dinner with friends, and suddenly I see ads for it on Meta or elsewhere. Are we overreacting? Is this really happening? Does the impulse to sell more push ads to become increasingly intrusive?
Host: Yes, when it happens it feels genuinely creepy. What's actually going on—you say something, and then ads appear?
Adam Foroughi: I think you've done some other trackable behavior—a search, a site visit, a product lookup—that you're not conscious of, and then you talk about a related topic and start seeing related ads. It's not microphones being on, or some app actually eavesdropping.
Host: Though there's a theory that if we're all at dinner together, these apps know our geolocation and can group us. Say we all discussed wanting a certain car or watch, then Friedberg leaves and searches for that watch on his phone to save it—and the system tracks all our locations and serves that watch ad to all four of us. Could that be the logic?
Adam Foroughi: I don't believe ad companies can track location. We don't track location at all. Precisely tracking someone's location for ad targeting is extremely heavy. And if you were to listen to microphones, parse the audio, and convert it into ad instructions—the data volume makes that completely impractical.
Host: What if we're friends and associated with each other?
Adam Foroughi: We wouldn't have that data. But if you're on a social network, your social connections may well drive ad experiences. If a friend searches for something, you might see related content. There's nothing wrong with that.
There's definitely a "creepy factor" that feels unsettling, but with so many companies running ads at scale, the data collected is heavily regulated at this point.
What people often overlook is that precisely because these ads have become so accurate, they've created enormous economic value. The ad you see, you can relate to it—20 years ago you wouldn't have paid attention to it. And today, a significant portion of GDP comes from this digital advertising economy. The more advanced these technologies become, the faster GDP grows.
Chapter Three: The IPO Crash and Becoming Your Own Best Investor
Host: Adam, I find how you run the company genuinely fascinating. You're based in LA now?
Adam Foroughi: I'm in LA. The company started in Silicon Valley—founded in Palo Alto.
Host: Palo Alto. And you have a significant number of developers in China?
Adam Foroughi: Our engineering teams are in Palo Alto, Beijing, and Singapore.
Host: You didn't raise much venture capital, went public in 2021 at around $20 billion market cap?
Adam Foroughi: We went public via a joint IPO in April 2021, at roughly $28 billion.
Host: $28 billion. Then by 2023, where did the market cap fall to?
Adam Foroughi: The public market is interesting. At IPO in 2021, we had $600 million in EBITDA and a $28 billion market cap, peaking at $40 billion. But in 2022, the stock fell almost every day, eventually reaching around $3.8 billion in market cap. And that same year, we delivered $1 billion in EBITDA.
Host: That's incredible. Wait, let's get this straight—the market wasn't buying the business story for whatever reason. So what did you do?
Adam Foroughi: You quickly realize—and coming from a finance background, I understood this clearly—your stock price depends on the quality of your shareholders. Our early shareholders included private equity investors, co-founders, and other team members who chose to sell post-IPO. There was also a glut of companies going public simultaneously, and blue-chip investors had no time to figure out what this oddly-named company was. So we faced minimal demand against massive supply, creating the crash. The valuation multiple went from elevated—I wouldn't use 50 times EBITDA to value us—down to a ridiculous below 4 times.
So as someone with a finance mindset facing such a collapse, you have to recognize the flip side represents opportunity. I turned to the internal team and said: I'm not going to talk to investors anymore. They're not buying our stock; it's a waste of time. But we're generating massive cash flow—so let's use it to buy our own stock and become our own best investor. We launched an aggressive buyback program. Since then, we've repurchased around $6 billion of our shares, retiring roughly 20% to 25% of outstanding stock. At peak prices, that $6 billion buyback is now worth more than $50 billion.
So what looked like a deeply bleak moment was entirely convertible into a massive opportunity.
Host: Were you always thinking that way, or was there a low period where you thought, "My God, what's happening?"
Host: Yes, I was just about to ask—how do you maintain company culture when the stock drops 92%?
Adam Foroughi: It was genuinely difficult. Family and friends called to ask, "Are you okay? Are you thinking about doing something drastic?" I told them: look, we started from a penny, the stock is still around $10—in absolute terms it's still up a lot. But as CEO, you immediately realize something painful: your team members are getting the same calls. And they don't have your certainty, nor do they own as many shares.
Our response was to build a "us against the world" mindset—everyone had turned their back on us, so we'd double down with buybacks. Simultaneously, we introduced a performance stock program. Normally this is reserved for the CEO, but we distributed it to our core team, telling them: we know this is hard. You think you have a house, and right now it feels like nothing. But if you hold on, and we bounce back, you're going to be rewarded richly on the upside.
When investors eventually started paying attention again, we happened to be at the point of upgrading from ML 1.0 to ML 2.0—moving from regression models to deep learning. The results were striking. Our ad algorithms drive the entire business. The better the algorithms, the higher the return on ad spend for our advertisers, and everything is performance-based. So the company started growing rapidly.
Entering 2023, we launched the new model in April, but we still weren't communicating externally, so nobody knew yet. Around September 2023, I went to New York. The stock had already recovered to about $80. I said, it's time to start talking to investors because the market cap is high enough that we can't buy back as aggressively as before.
That week, the stock went from $80 to $150, and the market cap roughly doubled from $28 billion to $55 billion—just because I went to New York.
Host: Just because you went to New York and said, "Hey, we're still alive, we made it through"?
Adam Foroughi: Exactly. Sitting in those meetings, you could easily read the response—I've done this long enough. You could see people dialing outside and saying, "Buy, buy, buy, buy, buy." That moment felt absolutely fantastic.
Host: And the flip side? Once they're long, do they start asking, "Adam, how do we expand? How do we grow faster? Why just games? Why not e-commerce? Why not this or that?"
Adam Foroughi: Absolutely. It's a genuine dilemma—you can't win either way.
Most people aren't real contrarians—the deeper the crash, the higher the rally. Our stock went from $9 to $750 in two and a half years, with market cap moving from $3.8 billion to eventually $250 billion. Both extremes were extreme.
We've now settled into a more stable position and are genuinely excited about growth opportunities. But my observation is that public market investors and private market investors aren't fundamentally different—they all chase trends, just often later than you'd hope. The very best ones catch these trends earlier than the market. That's the gap between top-tier VCs and average VCs, and similarly between top public market investors and average ones.
Chapter Four: Privacy Regulation, Apple's Control, and How AI Agents Will Change Shopping Behavior
Host: Can we talk about privacy? Apple and the EU are indeed scrutinizing companies like yours, believing your data collection methods are too aggressive. Some game developers also don't want platforms collecting user data and have been tightening access. Zuckerberg faced this directly. So what kind of headwind does this represent for your business? How do you handle privacy when Apple is actively trying—to put it plainly—to destabilize your business?
Adam Foroughi: In any such domain, you want clear regulations. Once rules are clear, technology can adapt. Five years ago, you could precisely target an individual user on iOS. Today, if a user says "I don't want precise targeting," you place them in a cohort and serve them less accurate ads.
Interestingly, after Apple made that change, a large number of users actually complained: "Show me more relevant ads—what you're showing me now is garbage." So yes, privacy regulation is needed—it enables tech companies to operate as intended—but consumers also do need relevant advertising to discover products.
You sit in a game, watch a 30-second ad to earn an extra life—you're actually receiving something of monetary value. Would you rather watch 30 seconds of garbage ads, or 30 seconds of content relevant to you? Since these privacy rules landed, tech companies have adapted, and deep learning networks are remarkably capable.
Host: A quick follow-up on game acquisitions. You've previously tried acquiring games. We also have Bending Spoons here today, actively acquiring businesses that aren't necessarily bad but have flat growth and don't interest VCs. For you, is being a game studio sustainable? Does it create conflicts with existing partners?
Adam Foroughi: We've sold all those games. We originally bought them for a data strategy—when building our first deep learning model, we needed training data, and game studios typically won't share data with third-party companies. So we bought studios ourselves to fuel the first model's training data, built a model that performed exceptionally well, and the company began growing rapidly. Once third-party partnerships came in, we divested those studios.
Host: In a world full of AI agents, what will advertising look like? The human-machine interface might shift—no longer typing at a computer or browsing a phone, maybe Meta glasses or other devices—what role do ads play in "agentic commerce"? So many people are pushing to make agentic commerce a reality.
Adam Foroughi: I think the reality is that part of the world will indeed start using agents to optimize certain recurring shopping behaviors. For example, I might hand my supplement subscription to an agent to auto-optimize and deliver monthly. But discovery ad platforms work differently, and the typical consumer isn't the kind of person deeply using agents, active on Twitter, and chasing the latest technology.
Our audience is closer to The New York Times reader—there are still a massive number of people using Yahoo products daily. The average consumer wants to find a product, experience the shopping journey themselves, compare options, track shipping, and enjoy the whole transactional experience.
If you tell them afterward, "Hey, an agent could have done all that for you and saved 20%," honestly, for a $50 transaction, it doesn't matter—because the dopamine from going through that process yourself is the real enjoyment they're seeking.
So yes, a segment of early adopters will embrace agents, but I think we drastically overestimate how representative Twitter circles are and forget that ordinary consumers simply aren't like that.
Chapter Five: How a Lean Team Beats Giants, Profit Moats, and Building Teams in China
Host: Let me understand how you win—Meta and Alphabet/Google are the smartest companies in the world with the best engineers, generating most of their revenue from advertising. They've had their models built for a decade or two. How does a small company compete and actually win in this specific domain? And what operating model sustains your continued success?
Adam Foroughi: One thing that's helped us get here: we never believe we've won. We wake up every morning thinking something's likely to go wrong today, so we have to work incredibly hard. That's created a lean company with a concentration of genuine domain experts deeply focused on the mobile gaming experience and how to convert it into consumer behavior on the other end.
I believe if you stay focused enough and stay lean, you can move faster than the giants. That's the key capability for taking them on.
Host: So where are the "leaks" in your business? I mean looking at a P&L like I did with Amazon a decade ago—we found their leaks everywhere, and the insight was Amazon would absorb those leaks and turn them into new business lines. That was our long-term thesis for being long Amazon. For you, where are the leaks? Is it payments infrastructure or something else? Or to put it differently, where's the opportunity for margin expansion that gives people a reason to buy the story?
Adam Foroughi: Our EBITDA margin, I believe, is number one in the market at 84%. So I'm not sure how many more "leaks" we have left.
Host: Given that number, it's hard to argue there are leaks—but thinking differently: advertisers come onto our platform, and their model is transaction-based. Say they're selling lipstick. We offer them an arbitrage opportunity—they "buy" consumers from us, that consumer completes a transaction, immediately covering their acquisition cost. The consumer spends $20 on lipstick, they pay us less than $20 minus the cost of goods, they're satisfied, and they keep increasing spend. This performance model is highly repeatable.
Our "leak" is that we're not everything along the entire chain. We're not the advertiser itself—we want to empower advertisers to reach consumers. We operate extremely lean, highly focused on algorithms and automation, so there aren't many leaks.
Host: The other side of the question: when you have an 85% EBIT margin, people say "this company might be over-earning," and competitors say, "I can compete away Adam's margins—I'd be happy doing this at 60% or 50%." But—perhaps extending the David question—that hasn't actually happened. The margin has held remarkably steady. Why is that?
Adam Foroughi: Because these technologies are extremely complex. If you continuously innovate and own differentiated data, you build genuine advantage. In a similar vein, Anthropic shouldn't by rights be leading the charge in large language models, but once a model reaches a certain scale and achieves widespread adoption by the community, it forms a moat that's extraordinarily hard to cross.
Host: Tell us about your China team—how are they an advantage?
Adam Foroughi: The Chinese people are humble, incredibly hardworking, and very smart. Whether working with Chinese engineers in China, in the US, or anywhere else in the world, you're working with some of the smartest people on the planet.
When I started the company, one of my goals was working with smart people and figuring things out together. Now, whenever I sit among my team, I know I'm probably the dumbest person in the room—and that's what motivates me to get to work every single day.
Host: Genuinely, how does that feel?
Adam Foroughi: It works for me.
Host: Let's give Adam a round of applause! Adam, thank you so much.
Adam Foroughi: Thank you, guys. Great to meet you all.
Comments