Google caps Meta's Gemini access over capacity limits

Key Takeaways

• Google informed Meta around March 2025 it couldn't fulfill the social media giant's full Gemini capacity requests
• The shortfall has disrupted and delayed some of Meta's internal AI projects
• Meta has told staff to be more efficient with AI tokens as a result of the restrictions

Google has restricted Meta's access to its Gemini AI models after Meta requested more computing capacity than Google could provide, according to a Financial Times report. Google informed Meta around March that it couldn't meet the full Gemini capacity Meta wanted to purchase. The shortfall has disrupted and delayed some of Meta's internal AI projects.

The restriction is notable because it involves two of the world's largest tech companies. Meta, despite spending over $65 billion on its own AI infrastructure this year and developing its open-source Llama models, has been quietly purchasing access to a direct competitor's AI models for internal work. That Google couldn't meet the demand says something about the state of AI compute supply across the industry.

Why is Meta buying Gemini access from Google?

Meta runs massive AI workloads. The company uses AI for content recommendation, advertising optimization, and safety systems across Facebook, Instagram, WhatsApp, and its Reality Labs division. While Meta has its own Llama models, Gemini apparently serves specific internal needs that Meta's own infrastructure couldn't cover.

The FT report notes that several other Google clients have also been affected by capacity constraints, though to a lesser extent. Meta's impact has been larger because of its "exceptionally high demand" for Google's models.

How is Meta responding to the restrictions?

Meta has told staff to be more efficient with AI tokens, the units that measure AI usage. This is a practical response: if you can't get more compute, you squeeze more value out of what you have. But it also signals that the restrictions are biting hard enough to require operational changes.

Neither Google nor Meta responded to requests for comment outside business hours. Reuters could not independently verify the Financial Times report, which cited people familiar with the matter.

Google Cloud's capacity crunch is real

Google CEO Sundar Pichai has acknowledged the problem publicly. During the company's Q1 earnings call, Pichai said computing power constraints prevented even higher growth at Google Cloud. The unit posted $20 billion in revenue for the quarter, but its backlog nearly doubled quarter on quarter. Customers want more AI compute than Google can deliver.

This isn't unique to Google. The entire AI industry is scrambling for compute. Nvidia's data center GPUs remain supply-constrained. Cloud providers are racing to build new data centers, but construction and chip production can't keep pace with demand. Alphabet plans to spend $75 billion on AI infrastructure in 2025. Meta is spending $65 billion. Neither company can build fast enough.

What this means for the AI compute market

The Meta-Google situation reveals the uncomfortable reality of AI scaling. Even companies with tens of billions to spend can't always get the compute they need. Smaller companies face even tougher constraints.

It also shows that Big Tech's "frenemy" relationships run deep. Google and Meta compete fiercely in advertising and AI model development. Yet Meta still buys Gemini access from Google when it makes operational sense. Commercial necessity overrides competitive posturing.

For enterprise buyers, the lesson is clear: AI compute is a seller's market. Planning lead times, diversifying across multiple cloud providers, and optimizing token efficiency aren't just good practices. They're survival tactics.

Frequently Asked Questions

Why is Google limiting Meta's access to Gemini AI models?

Google told Meta it simply couldn't meet the full computing capacity Meta requested. The demand exceeded what Google could provide, likely due to infrastructure and GPU constraints affecting the entire AI industry.

How has the Gemini capacity shortage affected Meta?

The shortfall has disrupted and delayed some of Meta's internal AI projects. Meta has responded by telling staff to be more efficient with AI tokens, the units that measure AI usage.

Are other Google Cloud customers affected by AI capacity limits?

Yes. The Financial Times reports that several other Google clients have been affected, though to a lesser extent than Meta due to Meta's exceptionally high demand.

Why is there an AI compute shortage across the industry?

Demand for AI compute has exploded faster than companies can build data centers and acquire chips. Even with Google and Meta each spending $65-75 billion on infrastructure in 2025, supply can't match demand.

Source: Tech-Economic Times / ET