What happens when AI compute demand outstrips supply? Google just gave Meta the answer: rationing.
What happened
According to the Financial Times, Google told Meta back in March that it couldn’t deliver the full Gemini capacity Meta wanted to purchase. The fallout: Meta has told staff to use AI tokens more efficiently, and some internal projects have been delayed.
The context makes it explosive
Google itself doesn’t have enough compute. In June, Alphabet signed a deal with SpaceX worth $920 million per month — for roughly 110,000 Nvidia GPUs housed in xAI’s data centers. Google openly called it a “bridge” to meet Gemini Enterprise demand. When the provider needs to build itself a bridge, you know things are tight.
What this means for the industry
The AI compute crisis is no longer theoretical. If Google — one of the companies with the largest infrastructure on the planet — has to ration its own customers, this affects the entire industry.
For Meta, this is a multi-layered problem. Its own AI development gets slowed down. It exposes dependence on Google as an infrastructure provider. And it makes Meta’s all-in AI strategy more expensive than planned.
June 2026 will be remembered as the month it became clear: AI compute is the new scarce resource. Not models, not talent — raw processing power.
Sources: Bloomberg, CNBC, The Next Web