AppLovin and the ‘AI crucible’

“I want to start by addressing what’s clearly on many people’s minds. While I prefer to ignore short-term fluctuations in the stock price and focus on maximizing value over the long term, the recent volatility warrants addressing. For the past few weeks, there’s been a lot of discussion about how AI and competition will challenge our business. But when I look at our internal dashboards, we’re delivering the strongest operating performance in our history. What’s fueling that growth is our own AI models, and as research in AI, both internal and external, continues to improve, our business will grow with it. There’s a real disconnect between market sentiment and the reality of our business.”

“Building leaves a lot of thinking invisible. As the scope and impact of our work has grown, we’ve accumulated a large set of lessons that are worth sharing. These lessons span engineering, product, and organizational decisions.

This blog is a place to make some of that thinking public. Over time, we’ll write about how we build, how we make decisions, and how we try to balance scale, quality, and velocity. We hope this helps others understand how we work and reflects an engineering culture we are proud of.”

“We Were Also Encouraged to See the Company Push Back on Bear Cases Around AI/World Models, Meta, and CloudX: Management also pushed back on bear concerns across Meta, CloudX, and broader AI disruption narratives, all of which have contributed to AppLovin’s underperformance in recent weeks. On Meta, the company emphasized that Meta Audience Network has been a longtime bidder within APP’s ad supply and that incremental bidding activity on IDFA traffic is not new.

Importantly, they also argued the competitive landscape today is fundamentally different than five years ago (when Meta was last dominant) given AppLovin’s performance gains and the closed-loop data advantage the company has built. On CloudX, the company highlighted the moats around its MAX mediation business, arguing it is protected by the integrated 360 solution (i.e. combining best-in-class monetization with differentiated UA performance), which drives meaningful switching costs and share-of-wallet concentration for customers.”

“The main issue with the defense of AppLovin’s business in these remarks on the competition point is that it positions competition on the SSP front and competition on the demand (e.g., Meta) front as being unrelated and distinct. But what if they aren’t?

As I point out in my 2026 predictions, it seems that Meta — based on commentary from CloudX’s CEO — has provided material support to CloudX.

While I think the odds of any upstart in-app SSP succeeding in stealing market share from MAX are daunting (I provide more color here), Meta’s patronage certainly doesn’t hurt. That said, AppLovin delivered a strong quarter, beating analyst expectations and guiding to higher-than-anticipated Q1 revenue. (…) The company is not imperiled.”

“Last week, for example, MNTN posted a QuickFrame-generated video to Instagram that features a ‘Heated Rivalry’-inspired hockey player decked out in MNTN branding (and, it should be noted, two different jersey numbers, a continuity error that would be less likely to occur during an actual film shoot).

According to Douglas, that post, which only cost roughly $500 to produce, racked up around 190,000 views and more than 2,200 likes, making it the company’s best-performing video not featuring Ryan Reynolds.”

“Grounding with web data (or web RAG) is becoming a foundational layer of AI — the bridge that connects AI models to real‑time, authoritative information so the responses from AI chatbots stay fresh and accurate. At Microsoft, we’ve spent years advancing grounding technology through Bing, and today it powers nearly every major AI assistant.

As agentic AI transforms the way we access information on the web, the emergence of Generative Engine Optimization (GEO) is also changing our understanding of how content contributes to AI-generated answers, citations, reasoning, and outcomes. To facilitate this shift, we’ve just added new insights to Bing Webmaster Tools to give publishers visibility into how their content shows up across Microsoft Copilot, AI-generated summaries in Bing, and select partner chatbots. Publishers can now see the total number of citations displayed as sources, average cited pages, relevant grounding queries, page-level citation activity, and visibility trends over time across AI chatbots and experiences.

We were the first company to ship grounding at scale with Bing Chat and are proud to also be the first that provides publisher insights from the agentic web. We look forward to hearing your feedback.”

“Starting this month, Target will be among the first companies to work with OpenAI to test contextual advertising in ChatGPT. Sponsored, contextual and clearly labeled ads from Target and from our Roundel retail media business partners will appear alongside users’ shopping conversations in ChatGPT, helping them discover products, deals and inspiration that meet what they’re seeking at that moment. (…)

As an early pilot program partner, Roundel is unlocking the opportunity for brand partners to show up in relevant moments beyond traditional media placements — and benefit from the learnings around user behavior.”

“What makes this even more challenging is that Amazon, Alphabet, and Meta can cover a significant share of this spend from the cash they generate, so they can keep this going, while OpenAI will either have to IPO or arrange for another monster funding round.

Seems like finding a niche (like Anthropic is doing) vs. trying to be everything to everyone might be the only path for OpenAI to thrive.”