Meta's hyperscaling AI spending has worried investors, but the company's ad business is already proving a clear return on that investment. In Q2 2026, Meta reported revenue up 27% and average revenue per person (ARPP) up 24%, driven by AI-powered ad recommendation and engagement models. The company guided 2026 capex between $130B and $145B, and the market has questioned whether that spending is justified for a company without a major cloud business.
Meta is an ads business. Almost all of its $201B in revenue comes from Instagram and Facebook, while messaging apps like WhatsApp remain largely unmonetizable due to encryption. Its "other revenue" line - mostly WhatsApp paid messaging and Meta Verified subscriptions - has grown 48% annually over the past three years but still accounts for just 1.2% of total revenue.
How Meta's AI models drive ad revenue
Meta's recommendation models are the clearest beneficiaries of its compute spending. The Andromeda model narrows tens of millions of ads down to thousands worth showing a user, driving an 8% increase in ad quality. Its Lattice model, which powers user feeds, drove a 12% increase in ad quality. The company's GEM model is a generative ads recommendation model trained on Meta's largest GPU cluster.
Meta's CFO said in Q4 2025: "This is the first time we have found a recommendation model architecture that can scale with similar efficiency as LLMs. And we're hoping that this will unlock the ability for us to significantly scale up the size of our ranking models while preserving an attractive ROI."
The compute spend also improves engagement. In Q1 2026, Meta's CFO said improvements "drove a 10% lift in Reels time spent… On Facebook, total video time increased more than 8% globally in Q1, the largest quarter-over-quarter gain in 4 years." In Q2 2026, she said Instagram's global time spent grew double digits year-over-year, "largely driven by improvements to our Feed and Reels recommendations."
For marketers, these improvements mean ad budgets spent on Meta are reaching more engaged audiences with better targeting. The AI for Marketing angle here is direct: Meta's models are converting engagement into ad impressions more efficiently, which supports higher CPMs and better campaign performance.
Meta's position in the LLM race
Meta lags in the LLM and cloud race. After buying 49% of Scale AI and assigning Alexandr Wang as Chief AI Officer, Meta released Muse Spark, which reached 3rd place in the LM arena overall rankings before newer models pushed it down to 27th place. In Q2 2026, Zuckerberg said: "Since we rebuilt Meta AI and integrated Muse Spark, we have seen a 60% increase in the number of people interacting with the assistant each day, and that continues to grow quickly week over week."
Meta is still supply constrained on compute, and its CFO says she sees many internal positive ROI projects where compute would go if available. The company hasn't confirmed how it will monetize its LLM through API services, productivity tools, or business agents, and the author of the original analysis says it's too early to project performance based on an undefined strategy.
Regulatory risk from child safety settlement
The biggest risk is Meta's recent $17.7B child safety settlement from a 2023 lawsuit. Meta will pay 70% of that amount unconditionally and pay the remaining $5.3B - plus set a default one-hour daily limit for underage users - IF YouTube and TikTok also add a default limit and pay $5.3B each. If that limit happens, engagement from underage users would drop sharply, which could hurt ARPP and CPM.
Valuation suggests market discount
A discounted cash flow model with a WACC of 10.65% and revenue growth declining from 26% in 2026 to 6% in 2035 shows an intrinsic price of $682, an 18% upside from Meta's current price. The reverse DCF solves for a revenue CAGR of 12.3%, significantly lower than Meta's historic 10-year CAGR of 26.3%. The market is not pricing in a massive ROI from Meta's hyperscaling ventures.
For marketing managers, the practical takeaway is that Meta's AI investments are improving ad targeting and measurement in measurable ways - engagement is growing double digits and ad quality scores are up. The AI for Marketing Managers playbook applies here: understanding how recommendation models affect ad performance helps you allocate budgets and set expectations with stakeholders.
Why this matters for marketers
Meta's AI-driven ad improvements mean marketers get better conversion from their ad spend, but the regulatory risk could change the platform's user base and ad inventory. If the one-hour limit for underage users takes effect, advertisers targeting younger demographics will see reduced reach. The core lesson: Meta's ad platform is becoming more efficient, but its legal exposure remains a factor worth monitoring when planning campaigns.
Your membership also unlocks:
