Google restricts Meta’s Gemini use as AI demand strains computing capacity

Google restricted Meta’s use of Gemini after Meta’s computing demand exceeded available capacity, underscoring how scarce AI infrastructure has become even for the largest technology companies. The Financial Times said Google told Meta around March that it could not meet all of Meta’s requested Gemini capacity, disrupting and delaying some internal AI projects. Meta had relied on Gemini for tasks where it outperformed its own Llama models, including harmful-content removal, scam detection, customer-service automation, advertiser chatbots and coding. In response, Meta has told employees to use AI tokens more efficiently, accelerated development of its internal Muse Spark model under Superintelligence Labs and begun shifting workloads away from Gemini. The strain extends beyond Meta: Google Cloud revenue reached $20 billion in the first quarter, but Sundar Pichai said compute constraints limited faster growth and helped double the cloud unit’s backlog over the quarter, highlighting a broader industry bottleneck in AI infrastructure.