Third quarter 2026 earnings call, November 4, 2026

AppLovin reported third quarter revenue of $2.18 billion, up 55% year-over-year and 14% sequentially, exceeding the high end of its guidance by roughly $100 million. The $200 million-plus sequential dollar step-up was the largest in the company's history. The result follows a second quarter that CEO Adam Foroughi described as a miss and "our own footfall." Management used this call to argue that the problem was identified and fixed, and the numbers largely support that claim. Adjusted EBITDA was $1.84 billion, up 59%, at an 84% margin, and net income rose 72% to $1.44 billion. For context, CFO Matt Stumpf noted that the largest digital advertising platforms reported growth of roughly 10% to the mid-20s this season, and that AppLovin converted about 65% of revenue into free cash flow.

A Larger Model Architecture, Only Partly Deployed

The most important new information is technical. During the quarter AppLovin partially rolled out a next generation of its AXON models, built on the architectural changes Foroughi flagged last quarter. The stated goal is to build bigger models, use more compute and train for better output. Foroughi placed the transition in the same class as AXON 2.0 in terms of technical complexity, though he declined to call it "AXON 3.0," saying the company is "not a direct-to-consumer AI company" and does not version its models for outside audiences. Asked how large the business impact could be, he offered no quantification, saying only that the new generation "gives us a much longer road map than the old models would have given us."

Two points stand out. First, this was a partial rollout, and Foroughi emphasized that the new architecture has "a lot more room to optimize" than the one it replaces. Second, the improvement is greatest where data is sparse, meaning newer and smaller advertisers and newer categories, which is precisely where the consumer vertical sits. Gaming revenue growth reaccelerated to nearly 50% year-over-year alongside the rollout. Management would not commit to a rollout timetable or to the cadence of gains, and Foroughi stressed that model improvements "won't land on a perfect quarterly schedule."

Cost discipline accompanied the technical push. Data center costs fell as a percentage of revenue because the infrastructure team improved serving efficiency, even as training compute increased. Stumpf said the company manages to EBITDA dollars and free cash flow rather than a margin target, and will keep spending where it sees a path to more revenue. Adjusted EBITDA flow-through quarter over quarter was 87%.

Q4 Guide Implies Sharp Deceleration, Built on Known Factors Only

The fourth quarter outlook calls for revenue of $2.365 billion to $2.395 billion, up 43% to 44% year-over-year and 8% to 10% sequentially. Adjusted EBITDA is guided to $1.995 billion to $2.025 billion. The sequential deceleration from 14% drew pointed questions given holiday e-commerce strength. Stumpf said the philosophy is unchanged: guidance includes normal seasonality, model updates already live and performing, and higher training compute costs. Management declined to break out gaming and consumer assumptions, saying only that both are strong in Q4 for different reasons. The implied full-year growth is above 50%. Foroughi said the company has beaten the high end of guidance in 14 of the last 16 quarters.

Consumer: Triple-Digit Growth, but a Seasonal Hangover Is Coming

Consumer advertiser spend rose double digits sequentially and triple digits year-over-year. Foroughi was unusually candid about the downside of deeper e-commerce exposure. Gaming has historically benefited from what he called "Q5," in which gaming advertisers spend competitively in Q1 because e-commerce competition recedes. That cushion has made AppLovin less seasonal than peers, but he cautioned: "as we get more deep into e-commerce, we should see bigger Q4 followed by tougher Q1." Customer acquisition should still allow growth in traditionally difficult quarters, but investors should expect like-for-like advertisers to show a large Q4 followed by a weak Q1.

Onboarding remains a drag. Customers going through AppLovin's team are far likelier to succeed than self-serve users, and Foroughi said very small businesses are "still further into the future." Advertisers also run incrementality studies before scaling, which slows the ramp. Creative is an opportunity cost rather than a barrier, he said, and he noted that Gen AI creative tools have advanced less than some expected because Western large language model companies are focused elsewhere. Ads on the platform average more than 35 seconds of watch time, and advertisers who load more creative scale better. A referral and partnership program remains a "work in progress" on a multi-year trajectory, with Pixel account additions accelerating.

Foroughi also explained why Pixel growth skews toward Asia-Pacific. Those advertisers often spin up new sites built specifically for the channel, so attribution is simple, while North American enterprises have established domains and more complex attribution. He cautioned against reading the counts as evidence of deeper penetration.

Auction Pricing, Competition and the 30% Long-Term Target

Management reiterated that the business can compound at roughly 30% annually over the long term, while saying it intends to run well ahead of that when possible. Asked whether the new model changes gaming growth expectations, Foroughi said it is hard to judge a vertical in isolation in a unified auction, but a more accurate model should support higher gaming growth. He added that 30% does not depend on it if consumer keeps growing quickly.

UBS's Stephen Ju relayed advertiser concerns that e-commerce bidders would inflate auction prices and compress returns, prompting some budget shifts to competitors. Foroughi dismissed such checks as flawed, arguing that advertisers who talk to analysts "probably aren't working that hard." He conceded CPMs rise as targeting improves, but argued that better prediction raises return on ad spend. On competition, he said other platforms are performing well and that the industry is not zero-sum, and he again attributed last quarter's weakness to company-specific execution.

On take rate, which the company does not disclose, Foroughi said it is an outcome rather than an optimized variable. He illustrated this with a conversion rate slightly above 1%, implying that on 1,000 impressions "somewhere around 987 of them we're losing money on." Take rate expands when AppLovin out-innovates the market and contracts if it falls behind. On supply, he said the MAX mediation market is growing at "strong double digits," driven by both volume and rising prices, which means AppLovin must keep improving its technology faster than publisher payouts rise.

Data Hunger and a Hint at M&A

Foroughi said the new architecture's constraint is data rather than modeling capacity: "we're going to be starved for data." The record quarter came without incremental data, he noted, but bigger models need more of it. Management is not in a rush, yet he said the company would consider acquiring "data relationships, customer relationships" that could pull forward what might otherwise be a decade of onboarding. He cited MoPub as the template, a deal that "has paid for itself many times over." No specific target was named. He also said the architecture could eventually become more generalizable across consumer and gaming, which are currently separate models, though he did not commit to that.

He declined to size self-learning versus engineer-driven improvements, saying that the two compound and cannot be cleanly separated.

CTV Is Early; Not in the Numbers

CTV appeared in Foroughi's 2027 priorities. Pilot campaigns are live and the product is "looking very promising," but he warned investors not to bank on near-term contribution. The opportunity, in his view, is a true top-to-bottom funnel performance offering on connected television, with under-monetized inventory resembling mobile 10 to 15 years ago. Beyond CTV, he said few third-party supply sources could move the needle in one step, so supply expansion would be a slower compounding effort, while diversifying ad types within existing impressions is the larger near-term lever.

Capital Return and Balance Sheet

Free cash flow was $1.4 billion, or 77% of adjusted EBITDA, up from 53% in Q2, which was depressed by cash tax timing. Full-year conversion is still expected at 70% to 75%. AppLovin repurchased and withheld 4.1 million shares for $1.4 billion, and the Board added $3.2 billion to authorization, bringing the total to $3.6 billion. Since inception, the program has retired 104 million shares for $9.4 billion at an average of about $90, returning 85% of free cash flow. Cash was $3.17 billion against $3.5 billion of long-term debt, leaving net leverage below 0.1x versus a long-term target near 1x. Stumpf confirmed the Q3 step-up was an opportunistic response to a dislocation between the share price and intrinsic value, and said the company can do so again. A BTIG analyst noted the stock remains down about 40% year-to-date.

What Management Would Not Provide

Asked by Wolfe's Shweta Khajuria for more visibility into durability, Foroughi offered no new KPIs. He said the company cannot provide predictive metrics at this stage because growth depends on model improvements and new customer additions, which are not forecastable, though he said they "can create very high upside." Investors are being asked to extrapolate from the company's record rather than from disclosed leading indicators, and the partial rollout plus the seasonal consumer dynamic leave genuine uncertainty around the first half of 2027.