Google Reportedly Rations Gemini AI Access as Meta Faces Compute Limits

Google has limited Meta’s access to its Gemini AI models because it cannot provide the level of computing capacity Meta requested, according to a Financial Times report. The restrictions have affected multiple Google customers, with Meta among the companies most impacted.

The reported limits have prompted changes inside Meta. According to three people familiar with the matter cited by the Financial Times, employees have been instructed to use AI tokens more efficiently as the company adjusts to reduced access. Google and Meta both declined to comment.

Meta had been using Gemini for internal tasks, including automating safety operations such as detecting harmful content and removing scams. The company chose Gemini because it outperformed its own Llama open-source models for those workloads. More recently, however, Meta has been shifting those tasks to Muse Spark, an internally developed model, as it works to reduce its reliance on outside AI providers.

The reported restrictions come as Google faces its own infrastructure constraints. The company recently agreed to pay SpaceX $920 million per month for access to 110,000 Nvidia GPUs, describing the arrangement as “bridge capacity” to support growing demand for Gemini Enterprise.

The situation highlights the pressure AI infrastructure demand is placing on even the industry’s largest providers. Despite operating one of the world’s biggest AI computing networks and planning more than $180 billion in capital expenditures this year, Google is reportedly unable to meet all customer demand for its AI services.

For Meta, depending on a competing company’s AI models was already viewed as a temporary solution. The company cut 8,000 jobs in May while increasing investment in AI infrastructure. It also issued capital expenditure guidance of $115 billion to $135 billion for 2026, reassigned 7,000 employees to AI-focused roles and introduced Muse Spark through its Superintelligence Labs division.

The limits on Gemini access could accelerate Meta’s transition toward internally developed AI systems for core operations, including large-scale content moderation.

The reported developments also underscore a broader challenge facing the AI industry. Demand for computing capacity continues to outpace available infrastructure, forcing companies to secure additional GPU resources while balancing growing customer workloads. Google is adding external compute capacity through SpaceX, while Meta is reportedly reducing AI token usage as it expands its own AI infrastructure.

This analysis is based on reporting from TNW.

Image courtesy of Google.

This article was generated with AI assistance and reviewed for accuracy and quality.