Bootstrapping a Business to $5 Billion in Free Cash Flow | AppLovin’s Adam Foroughi

Summary

• AppLovin’s response to a 92% drawdown was what Foroughi called one of the more successful buybacks in the company’s history: with the stock at a $3.8B market cap against $1B+ of EBITDA, he levered up and deployed ~$6B repurchasing shares — negotiated directly from the private-equity holders he knew would sell — ultimately generating “somewhere in the neighborhood of 50, 60 billion dollars of actual proceeds.” His logic was brutally simple: “You look at a multiple of five times cash flow, and you go, ‘Okay, why don’t we just buy all the shares?’” — and “I never believed in saving cash for a rainy day.”
• Every top VC rejected AppLovin in spring 2012 at $1M for 25% of the company — then funded competitors with the identical idea, which Foroughi turned into fuel. He calls it “a failure in seeing who the founder was” (this was his third advertising business), and says “we had it as a goal to put those companies out of business.” Today the company is worth roughly $140B; the funded rivals “either sold for very little or are gone.”
• The core product thesis is “make the advertiser an arbitrager”: performance ads where the client provably makes more than they spend, so “the only constraint on them scaling in our system is the money that they have in their bank account.” Foroughi deliberately avoided brand advertising from day one — “I wanted to be in the business of no sales, high-value product” — and still runs an ad giant with no CRO, no COO, and no sales force.
• To train ML models without advertiser data that clients refused to share, AppLovin vertically integrated — buying 14-15 game studios to obtain data and understand developers’ needs, temporarily damaging developer trust, then selling the whole portfolio to Tripledot roughly a year ago once advertisers were increasingly connecting to the platform. The lesson he admits: the trust breakdown was “a communication error, not a desire to deceive” — he never explicitly told the community why he was entering gaming.
• Axon 2, the deep-learning model rolled out in April 2023, coincided with the stock moving from $9-11 to a $750 peak (~$250B valuation) and marked the inflection: any advertiser can “plug in, get going, spend money, get revenue on the other side, and make a spread.” The company now generates ~$1.3B of cash in a single quarter — $5B+ run-rate — with only 350-400 people in the core business.
• The talent doctrine is the operational moat: in a year the business grew nearly 100% while Foroughi fired ~40% of the team, concentrated equity in a little over 100 critical people, and still personally approves every single hire — killing automatic “backfills” took annual hiring from hundreds to tens. “A players don’t like to work with B’s, C’s, D’s… eventually you end up churning out your A’s.”
• AI is compounding the leverage asymmetrically: over 80% of code was LLM-written a year and a half ago (“now it’s much higher”), and “your 10X engineer might be 100X more efficient” — which is why he resists hiring ahead of the e-commerce and broader expansion opportunity. The stated ambition, wired into a comp plan with a top trigger at a $1 trillion market cap: expand beyond gaming across SMB categories, down to “the local laundromat discovered by someone playing a game.”
• The governance lessons are candid and tradeable: no board for six years let him walk from a $600M offer (right call) but also led him into a $1B, 70%-control deal with a partially state-owned Chinese PE fund that was restructured after a year-plus CFIUS ordeal (wrong call), and IPO with only a 7-8% float — “probably a mistake in hindsight” because a thin float at a pushed valuation buys volatility and worse investors instead of long-term blue chips.

Deep dive

1. The market called a growing business trash — and Foroughi diagnoses exactly why

• The setup: April 2021 COVID IPO at ~$28B, briefly $40B, then down “literally every day for all of ‘22” to a $3.8B floor — $115 to $9 a share — while EBITDA grew from $700M in ‘21 to over $1B in ‘22. “You’re running a business, and the whole world is telling you your business is trash.”
• His post-mortem is structural, not fundamental: advertising plus gaming is “a tough space to understand,” and the COVID IPO flood meant Fidelity/BlackRock-class funds “couldn’t tell the difference” between listings, so AppLovin “ended up with a cap table that didn’t have support” while shares from private holders came to market. To outsiders it became “catching a falling knife.”
• A related confession on IPO mechanics: selling only ~7-8% float was “probably a mistake in hindsight” — a lower valuation with 10-15% float attracts long-term blue-chip investors; a pushed valuation with a thin float means “you’re signing up for just a much more volatile time in the public markets, which we saw.”

2. The $6B levered buyback that generated ~$50-60B of actual proceeds

• The pivot from despair to offense: “the whole world doesn’t like our shares, so if no one’s going to buy our shares, why don’t we just start buying our own shares?” With cash flow at roughly one-fifth of market cap, the company could in theory retire 20% of shares in a year.
• The execution twist most companies miss: instead of open-market purchases where “you don’t know who’s on the other side of that trade,” AppLovin went directly to the PE investors and departed founders who held ~50% of the cap table and were expected to sell over time, agreeing to buy them out — ~$6B deployed over 18 months, partly on borrowed money.
• On levering up to buy stock everyone hated: “to lever up to buy your own shares when everyone’s telling you your company’s a piece of shit, that’s really scary to do.” His counter: “I’m a big believer in what we’re building… if I believe in the future, and we’re a really high cash generative business, we should always be buying back our shares.” Result, by his math: “somewhere in the neighborhood of 50, 60 billion dollars of actual proceeds… one of the more successful buybacks in the history of the company.”

3. Every top VC said no in 2012 — then funded the same idea without him

• The pitch: $1M over a $4M valuation, spring 2012 — a quarter of the company for a million bucks — all rejected, because Google, Facebook, and Amazon loomed and mobile gaming was nascent. Foroughi’s verdict: “it’s a failure in seeing who the founder was. So, bet on the person” — this was his third advertising business after two prior exits.
• The kicker that stayed with him: “the same VCs that rejected me put a lot of money into those companies” — competitors that independently had the exact same idea at the same time — “and that was very motivational. We had it as a goal to put those companies out of business.”

4. Two failed consumer apps and the accidental discovery inside the third

• After selling his desktop/social ad businesses, he moved to Palo Alto in 2010 knowing only that “mobile’s going to be big” — and, feeling “the grass is always greener,” swore off advertising for direct-to-consumer. A married man launched a dating app; a self-described owner of “four outfits” launched a fashion app. “Both terrible” — though pre-Tinder and around the time Pinterest launched, so the markets were real; “we were not the right teams to do it.”
• The third app was the original AppLovin: an app-discovery app that pushed recommendations like “Adam’s playing Words with Friends. You should go play with him.” “The app itself stunk, but when you got that push, the response rate was through the roof” — so they scrapped the app and shipped the recommendation algorithm as an advertising SDK in March 2012.

5. Beating Google’s AdMob by serving developers, not brands

• Bootstrapping was deliberate: he “didn’t believe in selling to investors an idea that I wasn’t confident in,” so he structured an LLC with a tax-efficient way to absorb losses and self-funded for a year and a half until product-market fit.
• The competitive opening: Google had bought AdMob in 2008 for ~$1B and there wasn’t much innovation there by 2012, and AdMob was built for brands. Foroughi spent exactly “a one-week consideration” on brand advertising after a brutal New York agency tour — “proving that their dollars were well spent is completely hand-wavy… the biggest cut goes to the folks that are wining and dining the client the most. I wanted to be in the business of no sales, high-value product.”
• Instead AppLovin went to cash-starved developers of calendar apps, fart buttons, and early games: place our ads, and run your own performance ads on our platform to acquire users — building “tools for the developer instead of tools for the brand,” making clients companies “that can arbitrage our platform.”

6. Slightly profitable at a $12M run rate in eight months — and no board for six years

• Launch March 2012 to a slightly profitable $1M/month run rate by November, when he raised $4.25M as a convertible note to common, with his own dollars among the participants — and then no board existed until KKR in 2018. The pro: “I could make all the decisions… it was literally my signature as the president, the vice president, everything.” The con, freely admitted: “I ended up making some mistakes, especially in the capital markets… that I wouldn’t have made if I had had a board.”

7. Raff, the chip-on-the-shoulder hiring filter, and the Iran-shaped drive

• Raff — a high-school dropout who started working for Foroughi at 16, tried to hire him at 19 (“I laughed in his face”), then lived on a bunk bed in the Palo Alto office for six months — embodies the filter: “find people who have a chip on their shoulder, someone who has a reason to push hard.” Now a billionaire running marketing and growth marketing, he still haggles advertisers for “an extra 5,000 bucks a day.”
• Foroughi’s own chip: his family was uprooted from Iran when he was four; his father had run one of the country’s most successful real-estate development companies, “thousands of people worked for him, and we got booted out of our country.” Watching his father diminished created a permanent obligation — “that’s never gone away,” and success remains “a moving target. Nowhere near that moving target.”
• The daily texture of that drive: “I’d wake up every morning and go, ‘I got to check stats. Are we going bankrupt today?’” — same wake time, same routine, same stats. On Senra’s half-joking autism question: “if you knew me 25 years ago, I’d be deemed a lot more antisocial” — he once left his first post-college interview “drenched” in sweat, and treats communication as a learnable job requirement founders simply have to solve.

8. Walking away from $600M — because the business outgrew the term sheet

• In 2015, ~three and a half years post-launch and heading toward ~$50M EBITDA while still growing about 100% year over year, a $600M all-cash offer arrived. His number was “around a billion dollars at that point in time” — Senra’s dry rejoinder: “Still, only 400 million” against today’s ~$140B market cap. Foroughi’s hedge on the hindsight: “most businesses, when they sell, they lose the opportunity to keep innovating… it was a little fortuitous that I had enough conviction to not take an offer in that range.” No board also meant no pressure to take it — but saying no created internal pressure to do something.

9. The Chinese deal, CFIUS, and the convertible-note escape hatch

• The “something” was a deal announced sometime in 2016 with Orient Hontai Capital — $1B at a valuation above $1.4B for 70% control, aimed at a China listing amid the H-share boom — struck by a CEO who “didn’t even know what a state-owned business was” just as Trump’s first election soured the US-China climate. Then came CFIUS: “NSA, CIA, FTC, DOJ, everyone was in the room,” and the first meeting opened with “Get the Chinese people out of the room.”
• Regulators never voiced the concern directly, but “I’m pretty good at reading a room”: a large-scale data platform sitting on millions of devices, with a partially state-owned fund taking 70% control — “can they exert pressure on an American to then go against Americans’ best interests?” After a year-plus of frozen options and little hiring, the business doubled anyway to well over $100M EBITDA.
• A board would have said “just rip up the deal” — but Foroughi felt loyalty to investors who bet on him, and worried what happens to “a group of people from China that take a billion dollars offshore, and then the deal blows up.” The fix: restructure the $1B equity-for-70% deal as a convertible note that would give the investors 10% on conversion if the money plus interest were repaid, below the control threshold he described; dividend out $1B as pre-IPO liquidity, then bring in KKR’s Herald Chen in summer 2018 to clear the Chinese convertible debt and clean the cap table — creating his first-ever board of three.

10. Buying 14 studios for data, breaking trust, and selling the lot to Tripledot

• The strategic problem circa 2018: Facebook had “turned ads into content” with ML trained using much richer data — “what you buy is very much predictive of what you’ll buy next” — and advertisers who shared data with Facebook and Google refused to share it with AppLovin. So he vertically integrated: 14-15 game studios (PeopleFun first, moving “idea to first acquisition almost right away”) to obtain data and understand the needs of developers, feeding AXON 1, launched in 2020.
• The predictable trust crisis: clients watched AppLovin’s own games grow on its own model and concluded the platform was becoming a competitor that might clone their games. His admitted mistake: “I didn’t convey to the community of game developers directly and explicitly, ‘We’re getting in the gaming business, and here’s why.’” Dinners fixed it — helped by the credible tell that Foroughi personally hates playing games (“I don’t have the patience to get good at games… I don’t like sucking at things”) — and “trust comes back really quickly when you realize it’s a communication error, not a desire to deceive.”
• Endgame: with maybe over 50% of gaming’s marketing dollars now flowing through the platform, the 1,500-plus-person studio operation became “a distraction… not in our DNA,” and he sold the whole portfolio in one piece to Tripledot about a year ago rather than profit-maximize studio by studio. Total ownership period: roughly five years.

11. Grew ~100%, fired 40%: the distillation to ~100 equity-holding A players

• The 2022 crash forced a comp reckoning: employees underwater on four-year grants asked to be trued up, and Foroughi’s retort frames his philosophy — “when the stock goes up, are you going to give your shares back to me?” The solution: cash comp for replaceable functions, equity concentrated in “a little over 100 people” in core product and engineering, because “it forced me to go, ‘Who are the people that matter?’”
• Then the cuts: “the business grew nearly 100% year over year, and we fired about 40% of the team” (roughly 2024, from near-double today’s 350-400), including tenured people and people whose roles he believed would eventually be automated by LLMs — “I wasn’t going to keep them in a dead-end job.”
• The catalyst was Giovanni, who joined November ‘22, led Axon 2, and just became CTO, while Basil Shikin shifted to a different role after replacing co-founder John Krystynak in 2016 — “a really good manager’s job is to build a really good team and eventually get replaced.” Giovanni’s relentless “Why is this person in this role?” convinced Foroughi that culture can’t be static: “At every moment in time, we’ve got to rethink our culture for what the world has today.”

12. Ruthless efficiency as retention strategy — and what happened to the funded competitors

• The mechanism, spelled out: “A players don’t like to work with B’s, C’s, D’s,” and “really smart people… don’t want to deal with other people. They just want to get shit done.” Bloat and process burn out your best, who then leave and start companies — so leanness isn’t frugality, it’s how you keep the people who matter. Hence a C-suite of exactly CEO, CFO, CTO, and general counsel: no CRO, no COO, in an advertising business.
• The hiring-approval story: on taking over HR he found every departure auto-generated a “backfill” posting, guaranteeing headcount only rises. His test — “convince me that you need the person” — took annual hiring from hundreds to tens; now anyone who comes to him is genuinely desperate, so “Okay, go make it.”
• On the better-funded rivals, echoing Senra’s Michael Dell parallel: one is public, none “remotely close to the scale that we operate at,” and the VC-funded privates “either sold for very little or are gone” — victims of the pattern he refuses: “hire into opportunity, dilute their IQ, and then you’re stuck with a mess.”

13. Axon 2, the arbitrage flywheel, and the trillion-dollar map beyond gaming

• The inflection: Axon 1 (traditional ML) required advertisers to spend heavily and labor manually before the model learned; Axon 2, a deep-learning model rolled out in April 2023, let “any advertiser just plug in, get going, spend money, get revenue on the other side, and make a spread.” Stock: $9-11 then, $750 at the peak ~six months ago — under $4B to ~$250B — with fundamentals genuinely tracking; cash generation is now ~$1.3B in a single quarter on 350-400 core employees.
• The AI compounding: 80%+ of code LLM-written eighteen months ago, “now it’s much higher,” and the leverage is nonlinear — “your 1X engineer might be 2X more efficient. But your 10X engineer might be 100X” — plus LLMs let researchers ingest the entire open-source ML literature “not believing that they’re missing anything.”
• The audience thesis behind expansion: a billion-plus daily gaming users (150M+ US adults), skewing middle-aged — not “the same 21-year-old who’s on Instagram” — willing to watch ads for a long time, with the average ad over 35 seconds. E-commerce ads are already growing on the same make-them-an-arbitrager model; the target is SMBs across categories — “the local laundromat discovered by someone playing a game,” or the Shopify lipstick seller — and eventually enterprise, without a sales force. His proof point: a 10-15-person Turkish studio that sold for $1B within a year of launch, built almost entirely on AppLovin’s platform.
• The ambition is codified, not rhetorical: at every stage he voiced a target he believed — $1B in 2014, $10B at the Chinese round, $100B after the IPO dropped 15% on day one (“this shit motivates me. I love it when people tell us we suck”) — and the new key-people comp plan’s highest trigger is a $1 trillion market cap, with no timeline attached.