Feb 23, 2026 · 24-Hour AI Briefing: Anthropic’s $80B Cloud Revenue-Share Bet, Meta Pushes GenAI Video Ads Into Ads Manager
Over the past day, two updates point to the same shift: generative AI competition is moving from “model capability” toward the unit economics of compute and distribution. Anthropic is scaling through multi-cloud delivery plus revenue-share agreements with hyperscalers, while Meta is embedding generative video ad creation directly inside Ads Manager to turn creative production into an automated loop.
1. Anthropic plans to pay at least $80B to Amazon, Google, and Microsoft for cloud services by 2029, with most of it tied to revenue sharing
Commentary:
This isn’t just “buying cloud.” It’s binding go-to-market to hyperscaler channels. If cumulative spend reaches $80B by 2029, Anthropic’s defining challenge becomes margin durability under platform take-rates—not just training better models.
Payments to the three cloud providers were only about $1.3M in 2024, but are projected to jump to $360M in 2025, $1.9B in 2026, and $6.4B in 2027. Within the $80B cumulative total by 2029, the revenue-share component could represent 28% or more.
Going multi-cloud helps with data residency and compliance, and reduces single-vendor lock-in risk. But the structural tension is obvious: all three partners run their own model ecosystems, making them collaborators and potential competitors. If hyperscalers change revenue-share terms, raise compute prices, or prioritize their in-house models, Anthropic’s cost structure and go-to-market path face real uncertainty.
At a macro level, even massive AI companies aren’t guaranteed to be profitable—cloud providers, selling the “picks and shovels,” sit in a very different position.
2. Meta targets mid-2026 for a full rollout of generative AI video ad tools, likely integrated into Ads Manager; beta access is already available to some large advertisers
Commentary:
If genAI video creation lives inside Ads Manager, “creative” becomes workflow-native: brief → generation → variants → A/B → scaling → attribution, all in one place. Meta’s edge isn’t only the model—it’s distribution, targeting, and conversion signals that can turn video into a continuously-optimized parameter set.
Meta’s current AI ad stack is largely centered around Advantage and related tooling, so genAI video ads represent a clear capability expansion. The staged rollout makes sense: large advertisers have the budgets and asset libraries to run controlled lift tests, and they care about incremental ROAS/CPA validation. Once it works, Meta can productize “best-practice templates” and push one-click workflows to SMBs.
If the tool ships at scale, traditional agencies will feel more pressure. But real constraints remain: inconsistent output quality, privacy/compliance risk, and inference compute costs will determine whether this becomes a durable production system rather than a novelty.