Google has capped Meta's access to its Gemini AI model capacity since around March 2026, disrupting and delaying some...

Google has capped Meta's access to its Gemini AI model capacity since around March 2026, disrupting and delaying some of Meta's internal AI projects, according to a Financial Times report (also covered by Bloomberg). Surging demand for advanced AI models has made computing power the scarcest commodity in tech. This constraint signals that Google's AI cloud infrastructure is operating near capacity, which could strengthen its pricing leverage and accelerate its own capital expenditure plans. Separately, Google DeepMind acquired Windsurf, an AI coding agent startup, as noted in a Latent Space podcast, underscoring Google's commitment to expanding its AI capabilities. The capacity squeeze is bullish for Google's cloud business and for chip suppliers like NVIDIA and Broadcom, who may see increased orders, while Meta faces delays in its AI roadmap. The window for financial impact (capex ramp and potential revenue upside) is estimated at 45 days.