Google limits Meta’s access to Gemini AI models, report says

Google told Meta back in March that it couldn’t deliver the full computing capacity Meta wanted for its Gemini AI models. The Financial Times reported the restriction on June 28, noting that the shortfall has caused real disruptions and delays across several of Meta’s internal AI initiatives.

What happened and why it matters

Meta had been purchasing access to Google’s Gemini models through cloud and API services, seeking significantly more capacity than what Google ultimately proved able to supply.

Meta responded by directing employees to optimize their usage of AI tokens, the units that measure compute consumption for AI projects.

Other Google clients also faced limitations, but Meta’s outsized demand put it in a uniquely difficult position. As of late June 2026, the restrictions remain in place.

The situation is particularly ironic given that Meta is simultaneously a competitor to Google in the AI space. Meta develops its own Llama family of open-source large language models, yet was still leaning heavily on Google’s Gemini infrastructure for internal projects.

What this means for crypto and decentralized compute

For the crypto market, this story lands squarely in the narrative that has been building around decentralized compute networks. Projects like Render Network, Akash Network, and io.net have been positioning themselves as alternatives to centralized cloud providers, offering distributed GPU power sourced from a global network of node operators.

Decentralized compute networks are not about to replace Google Cloud for training frontier AI models. But for inference workloads, fine-tuning, and less compute-intensive AI tasks, distributed networks could absorb some of the overflow demand that centralized providers are struggling to meet.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.