Google Limits Meta's AI Capacity, Disrupting Projects

Google refused to sell Meta all the Gemini AI computing capacity it requested around March 2026, stating it could not meet the demand. This restriction has delayed some of Meta's internal AI projects and forced the company to advise employees to limit their use of AI tokens.

The timing coincided with Alexandr Wang, Meta’s chief AI officer, announcing an upcoming Muse Spark update. Meta had chosen Gemini due to its superior performance compared to its own Llama models.

The reliance on Google’s AI for essential functions like automating safety processes and customer service has exposed a capability gap for Meta. Earlier in 2026, Meta encouraged employees to maximize AI token usage, but after a surge in costs, the company shifted to limiting AI use and directing engineers to its internal coding assistant.

Meta's spending on AI tokens reached approximately $50,000 per employee annually, with over 60 trillion tokens consumed in a single month. The company is now prioritizing its own models, particularly Muse Spark, to reduce dependency on external resources.