AI Bubble: Google Ran Out of Compute for Meta in 2026

AI Bubble: Google Told Meta It Ran Out of Compute, Then Signed a $920M Monthly Deal With SpaceX

Google could not give Meta the Gemini compute it paid for, then signed a $920 million monthly deal with SpaceX to fix its own shortfall. That is what the AI crunch actually looks like.

Jun 29, 2026

by David Adubiina

The AI bubble debate has mostly been about valuations and hype. The real stress is showing up somewhere else entirely, in the servers. The race to build smarter AI models has created an unexpected problem for even the world's biggest technology companies, where there isn't enough computing power to go around.

According to a Financial Times report, Google has restricted the amount of Gemini AI computing capacity available to Meta after the Facebook parent requested more resources than Google could supply. The decision reportedly delayed some of Meta's internal AI projects and underscores how demand for AI infrastructure is beginning to outpace available capacity.

The shortage has also forced Meta to rethink how employees use AI internally. Sources familiar with the matter told the FT that staff have been encouraged to use AI tokens more efficiently as the company works to reduce costs while dealing with limited access to computing resources.

During Google's first-quarter earnings, CEO Sundar Pichai acknowledged that demand for cloud computing has exceeded what the company can currently deliver.

"We are compute-constrained in the near term," Pichai said, adding that Google Cloud revenue would have been higher if the company had enough infrastructure to meet customer demand.