Google is telling some of its biggest AI customers they can't have as much Gemini as they want.
Google has placed limits on how much access Meta and other clients have to its Gemini AI models, according to a Financial Times report. The restrictions stem from a compute shortage — Google simply cannot supply the processing capacity these customers are asking for. Meta has been among the hardest hit, and the rationing has begun to ripple into the company's internal AI projects.
The significance here cuts two ways. First, it's a reminder that selling AI-as-a-service is still a hardware problem at its core — no matter how capable your models are, you can't deliver them without the chips and data centers to back them up. Second, it puts a dent in Google's pitch as a reliable cloud AI partner, particularly when a customer as large as Meta is visibly affected.
Google is not alone in facing this kind of pressure — the entire industry has been racing to expand data center capacity — but getting caught short with a marquee client is the sort of story that rivals like Amazon and Microsoft will be quietly happy to let circulate.