Google restricts Meta's Gemini AI access due to compute capacity limits

Google has placed limits on Meta's use of its Gemini AI models because the company cannot provide as much computing capacity as the social media giant requested, the Financial Times reported on Sunday. The restrictions have affected several Google clients, but Meta has been hit particularly hard.

According to the report, the capacity shortfall stems from Google's inability to scale its infrastructure fast enough to meet surging demand for AI compute resources. The constraints have had a direct knock-on effect on Meta's internal AI projects, though specific details of which projects were impacted were not disclosed.

Compute rationing hits internal projects

The restrictions are not unique to Meta; multiple Google Cloud customers using Gemini models have faced similar allocation limits. However, Meta's reliance on Google's infrastructure for training and inference has made it especially vulnerable. The social media company uses Gemini for a range of tasks, including content recommendation algorithms and AI-powered tools for its platforms.

Key facts from the report include:

• Google has limited compute capacity for Gemini customers due to supply constraints
• Meta is the most significantly affected client among those rationed
• The shortage has disrupted work on Meta's internal AI projects
• Google is unable to provide the full compute resources Meta requested

The move highlights a broader industry challenge: hyperscalers are struggling to build out enough AI infrastructure to keep pace with demand. Google's own investments in data centers and custom TPUs have not closed the gap, forcing allocation decisions that prioritize certain customers over others.

Implications for AI compute supply chain

The rationing is a red flag for the entire AI ecosystem, where demand for computing power has outstripped supply. While Google, Microsoft, and Amazon have all announced massive data center expansions, the construction and hardware lead times mean bottlenecks persist. For companies like Meta that rely on third-party cloud providers for AI workloads, the risk of capacity shortfalls is now a tangible operational concern.

What comes next is uncertain. Meta may accelerate its own GPU procurement or shift more workloads to its internal hardware. Meanwhile, Google is likely to continue prioritizing its own AI products and high-value enterprise customers over adjacent companies like Meta. The episode underscores the increasing strategic importance of compute access, a resource that companies can no longer take for granted.