---
title: "AppLovin (Minutes): Next-Gen Model Only Partially Live; ~30% Long-Term Growth Guide Held Unchanged"
type: "Topics"
locale: "en"
url: "https://longbridge.com/en/dolphin/post/44790896.md"
description: "Q3 revenue was $2.18 bn (+55% YoY). The new-gen model, even on a partial rollout, returned the games biz to nearly 50% growth.Q4 revenue guide is $2.365–$2.395 bn. Management did not commit to the timeline for a full rollout of the model, and e-comm seasonality implies a strong Q4 and a weak Q1. Buyback authorization was increased by $3.2 bn."
datetime: "2026-11-05T00:13:02.000Z"
locales:
  - [en](https://longbridge.com/en/dolphin/post/44790896.md)
  - [zh-CN](https://longbridge.com/zh-CN/dolphin/post/44790896.md)
  - [zh-HK](https://longbridge.com/zh-HK/dolphin/post/44790896.md)
author: "[Dolphin Research](https://longbridge.com/en/dolphin.md)"
generator: "portal-rs"
---

# AppLovin (Minutes): Next-Gen Model Only Partially Live; ~30% Long-Term Growth Guide Held Unchanged

**Dolphin Research compiled the Trans of** $AppLovin(APP.US) **Q3 2026 earnings call**

**I. Core takeaways**

1\. Shareholder returns

a. Repurchased and net-settled 4.10 mn shares in Q3, spending $1.4 bn, a clear step-up vs. Q2; ending basic shares outstanding at 331 mn.

b. BOD approved an incremental $3.2 bn buyback; as of end-Oct, remaining authorization totaled $3.6 bn.

c. Since launch, cumulative repurchases and net share settlements reached 104 mn shares and $9.4 bn at an Avg. price of $90, equal to ~85% of FCF over the period. Diluted share count has declined for eight straight quarters, down ~2% YoY.

2\. Q4 guide

a. Revenue of $2.365–2.395 bn (+43%–44% YoY, +8%–10% QoQ); Adj. EBITDA of $1.995–2.025 bn (+43%–45% YoY). At the midpoint, full-year growth for both revenue and Adj. EBITDA exceeds 50%.

b. The guide embeds normal seasonality, continued expansion in Commerce, gains from already-deployed model upgrades, and ongoing investment in training compute for new models. Full-year FCF conversion still expected at 70%–75% of Adj. EBITDA.

3\. Key Q3 metrics

a. Revenue and profit: revenue of $2.18 bn (+55% YoY, +14% QoQ), with QoQ dollar growth over $200 mn, a record, beating the top end of guidance by ~$100 mn. Net income of $1.44 bn (+72% YoY), with a net margin of 66%.

b. EBITDA: Adj. EBITDA of $1.84 bn (+59% YoY), margin of 84% (+190 bps YoY), with 87% incremental margin QoQ, all above the guide high end.

c. Cash: FCF of $1.4 bn, at 77% of Adj. EBITDA (vs. 53% in Q2 due to tax timing), or ~65% of revenue. End-Q cash of $3.17 bn and long-term debt of $3.5 bn, implying net leverage below 0.1x. Long-term target is ~1x.

4\. Management lens and industry context

a. The company manages to absolute Adj. EBITDA and FCF rather than percentage margins. It will keep investing where it sees incremental revenue opportunities.

b. Management cited mid-teens growth at large digital ad platforms (roughly 10%–20%), while the company is growing multiples faster.

**II. Call details**

**2.1 Management remarks**

1\. Next-gen models

a. Partially rolled out a next-gen model based on last quarter's architectural changes. Advertisers can profitably deploy more budget at target ROAS, and the Games biz grew nearly 50% YoY, re-accelerating alongside the rollout.

b. The rollout is partial, and the new architecture offers far more headroom than the old. Gains are greatest in sparse-data areas—new advertisers, small advertisers, and new categories—which is exactly where Commerce focuses. Management will not commit to a quarterly cadence.

c. While training spend is rising, the engineering team optimized inference deployment, lowering data center cost as a percent of revenue. All investments are return-driven.

2\. Commerce

a. Advertiser spend rose double digits QoQ and triple digits YoY. Q4 is peak season for e-comm, and management says they are prepared for the holidays.

3\. Long-term framework and 2027 focus

a. Single auction across multiple ad formats, with continued improvement in Games and a lengthening runway in Commerce. Long-term CAGR around 30%, with an aim to outperform where possible.

b. 2027 focus: complete the model transition and keep optimizing on top; scale spend with existing clients and accelerate customer acquisition; enhance creative tools and ad formats; begin scaling CTV investment.

c. Last quarter's execution miss has been fixed. This quarter's results validate prior guidance.

**2.2 Q&A**

**Q: What is the biggest improvement vs. Axon 2, and when will full rollout happen?**

A: **No full rollout timing provided; the new model offers far greater optimization headroom.** The goal is to train larger models with more compute for better outputs, and early results are positive. Management will not commit to a full deployment timeline or evolution pace.

**Q: Is the Q4 guide of +8%–10% QoQ conservative vs. +14% in Q3?**

A: **Guidance philosophy unchanged, only known factors included.** This includes seasonality, model updates already live and effective, and higher data center costs.

**Q: Progress on the recommended partner program?**

A: **Ongoing, a multi-quarter and multi-year effort, with onboarding speeds improving.** New clients need to complete tuning and run incremental experiments before scaling sustainably. As partnerships deepen and reputation compounds, this effect should accelerate.

**Q: Does the new model change the 20%–30% growth outlook for Games?**

A: **The commitment remains 30% long-term company growth; if the new model continues to improve, Games could sustain higher growth.** Under unified auction, it is not useful to focus on a single vertical in isolation, as a Commerce surge intensifies competition. Commerce is already at scale and growing triple digits, so the 30% target does not rely on Games even without acceleration.

**Q: How important is CTV on the product roadmap?**

A: **Pilots are underway, the opportunity is attractive, but do not expect near-term P&L lift.** This is a new market and team confidence is rising; more details will come once client outcomes are compelling.

**Q: Can the Commerce-side data flywheel hit an inflection like Games did, and what investments are needed?**

A: **Data and architecture reinforce each other; the new setup enables larger models, and the company will remain 'data-hungry.'** Every new client increases data density; sources include recommendations, self-promotion, and potentially data acquisitions. Each data addition enables more complex models and better outputs.

**Q: How are holiday budgets shaping, can Q4 strength carry into 2027, and is Q1 seasonally softer?**

A: **Normal seasonality applies, with Commerce more clearly 'Q4 strong, Q1 soft.'** Budgets form over the coming weeks. The same e-comm client typically spends more in Q4 and less in Q1; historically stronger Q1s were due to a 'Q5' effect in Games, where lower competition spurs aggressive spend. Fast customer acquisition can partially offset seasonality.

**Q: Some advertisers worry e-comm bidders raise prices and compress ROI; is there enough supply to absorb incremental budgets?**

A: **Scale has expanded dramatically over three years while client ROAS strengthened; channel checks often misread this.** Better prediction yields better pricing: higher conversion raises CPM, but downstream ROAS rises too, which underpins high growth. Management said this quarter's QoQ delta may equal a full-year revenue figure for some peers.

**Q: As models scale, can they generalize across media and client types?**

A: **Possibly, and the new architecture can be steered in that direction over time.**

**Q: With brand advertisers shifting from mobile web to in-app due to web inventory issues, is this an opportunity?**

A: **If the goal is lower-funnel performance, significant budget should migrate in-app.** Brand-only budgets focused on reach tend to stay on the open web. Management believes in-app is one of the largest, if not the largest, ad opportunities outside social and search.

**Q: Can the new model unlock more effective inventory within current supply, and how big is the supply expansion?**

A: **New third-party open supply is limited, with CTV as the exception; near-to-mid-term the primary driver remains conversion.** CTV lacks full-funnel performance solutions and is under-monetized, similar to mobile 10–15 years ago. Commerce ads broaden mix: a user can see 15 product ads and 5 game ads in a week without hurting game conversions.

**Q: Shares are still down ~40% YTD; will the new authorization be deployed more aggressively and tactically?**

A: **Yes, buybacks will be opportunistic; the Q3 step-up reflected a disconnect between price and intrinsic value.** Management believes it retains capacity to act similarly.

**Q: Dynamic product ads can outperform video in e-comm; is the long-term goal more formats now, or wait for GenAI video maturity?**

A: **Both, with the goal of more diverse user-facing content.** Large models' creative capabilities keep improving, with multiple ecosystem players contributing; ad formats will also expand. The Commerce product is still new, and formats should evolve significantly over the next two years.

**Q: Are e-comm clients' learning periods into the holidays shorter vs. last year?**

A: **Models learn faster with less data, but clients still need incremental validation; slow ramp is industry norm.** Each iteration both improves existing clients and shortens time-to-results for new ones, lowering churn.

**Q: How are Commerce lead-to-conversion and client outcomes?**

A: **Clients coached by the team succeed far more than purely self-serve; micro and small merchants remain longer-term targets.** Mature self-serve clients can also succeed. As data accumulates, predictions for low-data new clients improve; management expects broad-based improvements in five years.

**Q: Are creative assets still a bottleneck, and is GenAI tooling slower than expected?**

A: **Creative is not a bottleneck, but clients who underinvest in creative cap their scale.** Western LLM firms prioritize other areas, so R&D here is limited. Company ads are full-screen with Avg. watch time over 35 seconds, and clients consistently request more assets; both creative quantity and quality determine scale.

**Q: Is in-house data sufficient to train larger models, and which data are most valuable?**

A: **No new data were added this quarter and it was still a record quarter, but faster data acquisition extends the runway long term.** Smarter use of existing data helps, but larger models ultimately need more data; the business is strong and not in a rush, and every new client is a new data relationship.

**Q: Can you provide more metrics to bolster confidence in sustained growth?**

A: **No forward-looking KPIs; the business is not yet stable, and growth depends on tech iteration and acquisition.** Fourteen of the past 16 quarters beat the high end of guidance, often by a wide margin. Clients generate tens of billions of dollars annually on the platform; model improvements are unpredictable, with substantial upside.

**Q: Is this upgrade effectively Axon 3.0, and what is the impact at full rollout?**

A: **Technical complexity is on par with Axon 2.0, and management did not quantify business impact.** The new model processes more data and scales compute clusters, with a roadmap much longer than the prior model; rollout remains partial and requires significant R&D.

**Q: The market worried about competition last quarter; what is the competitive landscape now?**

A: **Peers are performing well and this is not zero-sum; last quarter's issue was self-inflicted.** Forecasts and pricing differ across players, and all model improvements benefit clients. Multiple peers grew while the company delivered a big quarter, showing competition was not the key driver.

**Q: With platform annualized spend around $20 bn, how much of the 30% long-term growth comes from ROAS gains vs. acquisition?**

A: **Hard to split; Games already covers nearly all publishers, while Commerce has large acquisition headroom.** More data and the new architecture lengthen the improvement runway, and Games vs. Commerce contributions may alternate by quarter.

**Q: What are the assumptions for Games in the Q4 guide?**

A: **No split between Games and Commerce; assumptions are consistent for both and include Q4 seasonality and gains from already-deployed models.**

**Q: View on data purchases or tuck-in M&A; what is an ideal target?**

A: **Ideal targets bring valuable data and customer relationships, similar to the MoPub acquisition.** MoPub was shut within 90 days with all clients migrated, enabling MAX and generating multi-fold returns. If M&A can import, in one step, the clients and data the company would otherwise add over the next decade, it would materially accelerate the plan.

**Q: E-comm pixel growth is led by APAC; why product-market fit earlier there?**

A: **Do not infer deeper penetration from pixel counts; APAC merchants often build new sites for single channels.** APAC sellers leverage Asian supply chains and create site-by-product, making P&L clear for each channel. In North America, clients are often large enterprises with single domains and complex attribution.

**Q: The old model had 3%–5% annual self-learning gains; does that persist and is it in guidance?**

A: **Self-learning always exists but is not quantifiable or separately included.** Across 16 quarters, except last quarter, there were multiple A/B lifts in almost every quarter; these are unpredictable but compound into today's business.

**Q: How fast is MAX supply growing, and are traffic acquisition costs rising?**

A: **MAX marketplace supply is still growing at a strong double-digit pace, and tech progress absorbs cost inflation.** Most ad markets see supply growth in single digits to low double digits. MAX growth comes from both supply expansion and better monetization; the company's tech has outpaced the market over the past four years, driving faster growth.

**Q: How does take rate compare vs. last quarter and last year, and is there a difference between Commerce and Games?**

A: **No discussion on take rate; it is an outcome, and pricing is similar across both.** Conversion is a bit over 1%, meaning out of 1,000 impressions purchased, about 987 lose money. When the company leads technically, take rate rises; when it lags, it compresses. Under unified auction, a $20 CPM for Games vs. $21 for Commerce is enough to win, so there is no wide gap.

**Q: Who owns data in transactions mediated by MAX?**

A: **The company does not receive special data via the mediation layer.** Axon only knows whether it won, the clearing price, or the losing bid, and cannot see others' creatives. MAX ensures identical data permissions across bidders, and this fairness underpins its top monetization.

**Q: Are onboarding challenges on the company side or client side for Commerce, and which clients are prioritized?**

A: **Mainly a learning curve; priority is mature clients, hence the focus on partner channels.** Merchants are not used to long-form ads being fully watched and must learn how to build 60-second shop content. Once agencies crystallize best practices, processes will standardize, similar to social ads a decade ago.

\<End of text>

**Risk disclosure and statements:**[**Dolphin Research Disclaimer and General Disclosure**](https://support.longbridge.global/topics/misc/dolphin-disclaimer)

### Related Stocks

- [APP.US](https://longbridge.com/en/quote/APP.US.md)
- [APPX.US](https://longbridge.com/en/quote/APPX.US.md)

---
> **Disclaimer: This article is for reference only and does not constitute any investment advice.**