Google Rations AI Compute to Meta as Capacity Tightens

Google has restricted Meta’s access to its Gemini AI model, the Financial Times reported, citing people familiar with the discussions. The move reflects a deepening infrastructure crunch that is forcing even the largest cloud providers to ration compute capacity.

The restriction came after Google determined it could not meet Meta’s request for additional GPU clusters, the report said. Google sells access to its custom tensor processing units and Nvidia H100 chips through its cloud unit. Internal demand for AI training and inference has surged. The company now faces a choice between serving external customers and reserving capacity for its own products, including Gemini itself.

For Meta, the limitation could slow its AI development pipeline. Meta has been integrating generative AI features across Facebook and Instagram. It relies on third‑party cloud providers to supplement its own GPU fleet. Any delay in accessing compute could push back product launches or force Meta to pay higher prices for alternative capacity from Microsoft’s Azure or Amazon Web Services.

The FT report did not specify how long the rationing would last or whether it applies to other Google Cloud customers. Google has been expanding its data centre footprint. Lead times for new facilities and GPU deliveries remain long. Capital expenditure is expected to rise sharply this year, though Google has not disclosed a specific target.

Both companies are racing to deploy large language models that require massive clusters of accelerators. Meta’s open‑source Llama models have gained traction. Training them at scale demands tens of thousands of GPUs. Google’s Gemini is being rolled out across consumer and enterprise products, putting further strain on its own compute supply.

The broader AI infrastructure market shows signs of stress. Cloud providers limit access to high‑end GPUs. Some startups report waiting months for capacity. Google’s rationing is one example of these constraints.

META carries an Alpha Score of 65, reflecting moderate positioning. GOOG scores 66. Stock reaction has been muted. META is up 1.36% on the day at $550.25. GOOG trades near its recent range. Investors will watch for further disclosures on capacity allocation and capital spending when both companies report earnings next month.

The FT cited people familiar with the discussions who spoke on condition of anonymity. Google declined to comment. Meta did not respond to a request for comment.