
Google capped Meta's Gemini access in March and is paying SpaceX $920M/month to fill its own compute gap
After Meta's engineers burned 60 trillion tokens in 30 days, Google told Meta in March it couldn't meet the demand — while separately signing a $920M/month deal with SpaceX for 110,000 Nvidia GPUs to bridge Google's own capacity shortfall. With over 60% of 2027-scheduled data centers still unbroken ground as of May, the crunch is structural: Google can't guarantee AI access even to its largest customers while its own pipeline stalls on permitting and supply chains.
Source: techbrew.com ↗
Google has reportedly capped Meta's Gemini use because it can't provide the compute needed to fuel the demand
Financial Times
Why this matters
- → Even trillion-dollar tech firms can't guarantee AI access when compute runs short
- → Data center buildout delays expose structural capacity crisis through 2027
- → Forces hyperscalers into blank-check spending arms race with uncertain ROI
Compute crunch hits the giants