Google Cuts Off Meta’s Excessive Gemini Token Usage, Disrupting AI Projects
Meta’s AI token addiction forced Google to impose usage caps, delaying internal projects.
According to a Financial Times report citing three anonymous sources, Google was forced to cut off Meta's access to its Gemini AI models earlier this year after Meta's token consumption became unsustainable. Meta had been one of the largest participants in a so-called "tokenmaxxing" trend—evaluating employees by how many AI tokens they used—and its engineers were heavily deploying token-hungry agentic platforms like OpenClaw. Google informed Meta that it could no longer keep up with demand and imposed usage limits that remain in place today.
These restrictions disrupted and delayed several of Meta's internal AI projects. In response, Meta shifted its internal culture from maximizing token usage to encouraging more efficient token consumption. The strain on Google's own AI infrastructure may have contributed to its decision in early June to rent compute capacity from SpaceX, parent company of xAI, for $920 million per month. While other large companies also faced caps, Meta was described as exceptional even among AI high-rollers. Both Meta and Google declined to comment.
- Meta's excessive Gemini token usage led Google to impose limits that disrupted internal AI projects.
- The restrictions forced Meta to replace 'tokenmaxxing' culture with efficiency-focused token counting.
- Google later committed $920M/month to rent compute from SpaceX (xAI), partly due to capacity strain from heavy users like Meta.
Why It Matters
AI token management becomes a strategic bottleneck as hyperscalers cap usage by major clients, reshaping corporate AI strategies.