Google reportedly rationed Gemini capacity for Meta

The Times of India reports that Google declined to sell Meta all the Gemini AI computing capacity it requested, leaving some of Meta's internal AI work disrupted and delayed. The article says Meta had been using Gemini for coding, customer-service tools, advertising support, and safety workflows because it outperformed Meta's own models in some tasks. The capacity cap has pushed Meta toward lower token use and more reliance on internal tools.

Key details: Google reportedly told Meta around March 2026 that it could not meet all requested Gemini capacity, Meta had used Gemini for coding, ads, customer support, and safety automation, The restriction coincided with Meta trying to improve its own AI models and internal assistants.

Why it matters: Even the largest AI builders can become dependent on rival model capacity, making inference access a strategic bottleneck.