Adam Foroughi on AppLovin, AI and Building a Lean Company

AppLovin co-founder and CEO Adam Foroughi speaks with Harry Stebbings about founder motivation, company culture, AI, layoffs and public markets. Foroughi has a direct financial and professional interest in AppLovin; his account of its performance and disputes should be read as the CEO’s perspective.

Asked whether fear of losing or the desire to win motivates founders, Foroughi favors the latter. He thinks fear can make people protect against downside instead of taking consequential risks. He also says personal growth, learning and intellectually stimulating work can remain motivating after money becomes less important.

Foroughi says earlier businesses had already given his family a financial cushion when he co-founded AppLovin. He recalls an all-cash acquisition approach worth hundreds of millions of dollars in 2015. Because he did not need to sell for personal security, he says, he could focus on the company’s longer-term potential and decline the opportunity.

Stebbings raises the reported $83 million figure for Foroughi’s 2023 compensation. Foroughi says the package needs to be understood in light of a performance-based equity award granted after the stock fell in 2022. In his account, vesting thresholds required substantial share-price recovery; the headline compensation figure is not the same as cash salary or necessarily the amount realized.

He recalls a share price around $9 when the award was set and says the first threshold was approximately $38–$40, with further levels reaching the $80 IPO price. He argues that founder CEOs should be able to earn additional compensation for creating value, and says the award aligned him with shareholders. These are his recollections and his defense of a package in which he has a personal interest.

Foroughi describes running a public company as lonely and stressful, with colleagues and outsiders interpreting a CEO differently as the share price changes. He says the job can distract founders from their families and argues that its demands matter when assessing compensation. Stebbings asks whether he has ever questioned the sacrifice.

At the low point in 2022, Foroughi says he saw his health deteriorating: he was sleeping poorly, drinking excessive coffee and losing fitness. He decided that sustaining the job for another decade or more required a reset. He also realized that he was often physically with his children but mentally elsewhere.

His response was to make short periods with his children fully present and take up activities such as surfing that required him to put down his phone. He says these changes improved his stability and helped him think longer term as a CEO. He returns later to the difficulty of balancing ambition with family life.

Foroughi recalls AppLovin’s stock falling approximately 92% in 2022 and says the daily decline could undermine confidence in the business. Unlike some executives, he says he does watch the share price because employees and investors care about it. His challenge was to distinguish market pressure from the work needed to improve the company.

He explains that AppLovin’s advertising business uses recommendation models to help advertisers reach customers on a performance basis. At the low point, the company decided its older machine-learning system needed replacing. In his account, it slowed work on the old system, changed some personnel and committed to a new approach despite the internal disruption.

Foroughi says retaining the core team required him to communicate conviction in the new strategy even as the share price was falling. He describes the new Axon 2 model launching in April 2023 and the company subsequently growing rapidly. This is the CEO’s account of the turnaround, not an independent causal analysis.

Asked whether current layoffs stem from AI or pandemic-era overhiring, Foroughi thinks overhiring is the larger factor at most companies today because AI has not yet taken full effect. He also says AppLovin itself cut staff by roughly 40%–50% in most departments during a period he places mainly in 2024, partly because he expected roles to be automated or adoption to be too slow. Those estimates and reasons are his account, not evidence for every company’s layoffs.

He says he first examined processes that he considered unnecessary, including in HR, then looked at areas such as creative production where he expected automation to improve. He believed strong engineers could use AI tools to increase output. His expectations about future automation were bets, not guarantees that tools were already ready for all the work.

Stebbings challenges the risk of cutting creative staff before the tools are ready. Foroughi acknowledges it was a bet and argues that he did not want people remaining in roles he believed were heading toward a dead end. He says departing employees should receive good severance; his rationale is a management position, and the interview does not assess its effects on those workers.

Stebbings cites roughly 895 employees and asks whether everyone can be a top performer. Foroughi distinguishes approximately 400 people in AppLovin’s core advertising business from separately run acquired businesses. He estimates that core-business EBITDA per employee is around or above $10 million, based on his allocation of a high proportion of company EBITDA to that smaller headcount; this is not a company-wide per-employee figure.

Foroughi says not every role is the same, but he wants individual contributors who act without excessive process. He estimates that HR fell from roughly 70–80 people to about 15 and describes the retained team as highly effective. He says AppLovin’s executive structure is lean, with a CEO, CTO, CFO and general counsel rather than several other common C-suite roles.

He credits CTO Giovani with repeatedly asking why established roles and processes existed, prompting him to revisit structures built over a decade. Foroughi says the resulting organization returned closer to its founding culture. Asked whether other CEOs can do the same, he warns that simply firing half a team does not necessarily create an effective culture.

Foroughi speculates that some organizations may find it exceptionally difficult to rebuild once their strongest people have left. He says layoffs could produce improvements when leadership understands what needs to change, but he doubts that headcount cuts alone will do it. These judgments describe his management philosophy rather than an established result across companies.

On stock-based compensation, Foroughi says AppLovin grants roughly $300 million of stock annually and considers the amount small relative to the market capitalization he cites. He argues that high issuance can create severe dilution when a company’s stock price falls. The figures are approximate and refer to his account at the time of the interview.

He says the company moved from granting equity to everyone to granting it to roughly the top 10%–15% of employees, with others paid in cash and able to buy shares through an employee program if they choose. He frames this as a way to limit dilution while reducing exposure to volatility for employees who need predictable pay.

Foroughi prefers to evaluate businesses using cash flow minus stock-based compensation. His reasoning is that equity issuance dilutes owners, and buying shares back to offset that issuance consumes cash. This is his preferred analytical lens, not the only valid way to value a company.

Asked about AI-generated code, Foroughi estimates the share at AppLovin may be 80%–90%, but does not specify a measurement method. He cautions that a high percentage or large token count can reward low-quality output. What matters to him is whether engineers use the tools to create measurable value that exceeds their costs.

He says AppLovin has no separate product organization: engineers are expected to understand the product and the business metrics it serves. He also says they must review AI-produced code for safety and quality. Foroughi argues that agents working toward unclear deliverables can increase costs without creating corresponding revenue.

Foroughi says the key measures for his advertising business are prediction accuracy and whether advertisers earn more than they spend, with the company’s own revenue expected to grow if the system works. He presents this as an advantage of a measurable performance model. The interview does not independently audit the measurements or outcomes.

He distinguishes the company’s recommendation system from large language models. AppLovin uses language models for productivity and some business functions, but he argues that ad selection benefits from a purpose-built model. He warns that companies offering only an interface over a frontier model face competition from the model providers themselves.

On headcount, Foroughi doubts the core team will either double to 800 or shrink to 50 in the near term. He says new ventures could require additional people, while the existing core may not need substantial hiring. These are judgments about possible directions rather than staffing commitments.

Foroughi says he does not hold routine one-on-one meetings or formal reviews with his direct reports; if something is wrong, he gives feedback in real time. He believes highly capable people can learn through curiosity and do not need extensive structured mentorship. This is the culture he prefers and does not establish that coaching or formal development is unnecessary for everyone.

He proposes that written chats and transcripts of work calls can help new employees ask AI systems to summarize priorities or learn from strong sales calls. Stebbings asks about losing the value of in-person interaction. Foroughi says most work communication can be documented, while personal meetings remain valuable for building customer relationships.

Stebbings questions the value of social events for a team focused on execution. Foroughi replies that intense work debates can create resentment if colleagues never reconnect as people. He says dinners or drinks can restore relationships and also produce ideas; the interaction matters even when it does not look like formal work.

Foroughi corrects the idea that he no longer attends conferences. After the 2022 decline, he says the company curtailed investor relations while buying shares and concentrating internally. As the business and stock recovered, he returned to conferences to explain the business, despite a longstanding discomfort with public speaking.

Another difficult change was delegation. Foroughi says he had been a controlling, hands-on CEO, but recognized that colleagues were stronger in some areas. CTO Giovani took over responsibilities Foroughi had previously owned, giving him more time for strategy and investor relations.

He views “founder mode” as a response to layers and process in large companies, but argues that a founder who has assembled a strong, lean team should then delegate to the people with deeper expertise. In his account, taking control and giving it away can be appropriate at different stages.

Stebbings asks about short sellers after a sharp stock decline. Foroughi says he expected skepticism after a dramatic rise in a relatively unfamiliar company’s value. He alleges that incentives around short positions and dramatic reports can encourage market manipulation; these are allegations and a critique from an interested CEO, not an established finding that particular short sellers acted unlawfully.

Foroughi argues that the company needed to explain its model more clearly to investors and clients. He recalls approximate recent growth and EBITDA margin figures, while acknowledging he does not remember exact numbers. He sees the company’s unusual financial profile as a reason people questioned it, and disputes the suggestion that strong results alone imply cheating.

He says advertisers spend at scale on the platform and cites a previously disclosed advertising-spend run rate of about $11 billion. That is advertiser spending through the platform, not AppLovin revenue. His broader argument is that he owes customers and employees a clearer explanation of the business.

Despite his confident presentation, Foroughi says he has frequently worried that the business could fail and has checked its performance early in the morning. That fear, he says, motivates him to avoid complacency. He distinguishes anxiety about the company’s durability from his conviction in the team’s future strategy.

Foroughi praises TikTok’s recommendation system for serving engaging content and says better models can make advertising more relevant with less information about users. This is his assessment of the technology. He then discusses how AppLovin might grow through gaming advertising, connected TV and other applications.

Asked what would make AppLovin a trillion-dollar company, he speculates that producing $30–$35 billion in annual cash might support such a valuation. This is a hypothetical estimate, not a forecast or valuation guarantee. He says a social network is not required, although experimenting with engagement models could attract talent and develop the team’s skills.

On whether compute attracts researchers, Foroughi says the answer depends on the field. In his view, large language model research benefits from substantial compute, while recommendation systems also depend heavily on talent, curiosity and the choice of problem. He offers his assessment of other companies’ research cultures without presenting internal evidence.

Foroughi warns that buybacks are difficult to time and often disappoint. He recalls that AppLovin repurchased shares in 2022, including from holders seeking liquidity, and used cash and some debt. He says this reduced selling pressure and estimates roughly a third of the company’s value at the time of the interview was attributable to the buyback; that counterfactual estimate is his, not a directly realized gain.

Discussing software valuations, Foroughi thinks uncertainty about future competition from AI models can make enterprise SaaS growth harder to predict. He speculates that the market adjustment may not be over, while saying established products will not necessarily disappear because customers often remain with embedded systems. This is a market view, not investment advice.

He expects more powerful agents to speed product development, although he says he does not know where that leads. He predicts that engineering and product work may converge, with engineers understanding product needs and product people becoming more technical. He still sees human engineering expertise as important for reviewing security and code quality.

On security, Foroughi acknowledges that faster AI-assisted releases can introduce vulnerabilities and says breaches may increase as teams ship faster. He also hopes models will help identify and fix flaws over time. Both the near-term risk and the longer-term improvement are predictions rather than established outcomes.

He calls for an honest discussion of what more capable AI could mean for technology employment. He speculates about more startups and output, but also expects further layoffs as firms adopt new tools. Stebbings challenges the choice to cut staff during a strong growth year; Foroughi says his company requires employees to adopt the technology and increase output.

Foroughi cautions against using token budgets or consumption leaderboards as goals in themselves. He says organizations should measure whether spending on AI contributes to their business objectives; otherwise they risk paying for output with little value. This repeats his distinction between usage volume and useful results.

Asked about parenting and ambition, Foroughi says deep focus on a company can come at the expense of family attention. He regrets moments in his children’s lives that now feel blurred because his mind was on work, and says he is still working on a balance. His experience is personal rather than a rule for every founder.

Foroughi describes planning three to five years ahead and working backward, noting that current results often reflect earlier decisions. On governance, he recalls operating without a formal board early on, then building a larger board after investment. He says he stepped aside as chair in favor of Craig Billings so he could focus on running the business.

He says he no longer makes personal investments because they would draw liquidity and attention away from AppLovin. His preference is to concentrate on making his own company better over the next several years. This is his personal allocation choice, shaped by his existing stake and role, rather than a universal rule for founders.

Looking back at the IPO, Foroughi says timing was difficult but he tries to learn and move forward rather than relitigate decisions. He views going public as a financing event and argues that management should build for long-term returns. His hope that shareholders outperform alternative investments is an aspiration, not a promised return.

Asked which advertising competitor is most at risk, Foroughi instead describes the company’s goal as generating incremental transactions for advertisers. He gives a hypothetical example of increasing advertising spending and sales by 20%. The figures illustrate his proposed model; they are not a demonstrated result for all advertisers.

In the closing exchange, Stebbings asks whether an aggressive culture can exclude people. Foroughi acknowledges that his directness can rub people the wrong way but says he values speed and candor. He also recognizes that moving quickly can cause miscommunication and mistakes, and describes a willingness to accept that risk.