Google reportedly limits Meta’s Gemini AI usage amid computing capacity pressures

The growing demand for artificial intelligence infrastructure has reportedly led Google to restrict Meta’s use of its Gemini AI models after the social media company exceeded the computing capacity it had available.

According to a Financial Times report, the move highlights the growing strain on AI infrastructure, even among the world’s largest technology companies. Despite investing billions of dollars in data centres and advanced hardware, major firms are struggling to secure sufficient computing resources to support the rapid growth of AI services. Reuters also reported that Google informed Meta earlier this year that it could not meet the company’s full demand for Gemini computing capacity.

Meta’s growing demand for AI infrastructure

Meta has been expanding its use of artificial intelligence across a wide range of products and internal operations. Although the company develops its own large language models through the Llama family, it also relies on external AI providers for specific workloads where their performance is stronger.

According to people familiar with the matter cited by the Financial Times, Meta uses Google’s Gemini models to power customer support systems, advertiser chatbots and software development tools. Gemini is also reportedly used in processes such as harmful content moderation and scam detection. The report said Meta initially selected Gemini for several of these tasks because it outperformed the company’s own open-source Llama models in those areas. Meta is also understood to use AI models from Anthropic, including Claude, for similar workloads.

The report stated that Google warned Meta about capacity constraints in March after demand exceeded the available computing resources allocated to the company. As a result, Meta reportedly instructed employees to use AI tokens more efficiently to reduce consumption and manage limited resources. The restrictions are also said to have delayed some of Meta’s internal AI projects. Neither Google nor Meta publicly commented on the report.

Billions invested, but AI capacity remains limited

The reported restrictions come despite continued investment across the technology industry in AI infrastructure. Meta has committed hundreds of billions of dollars to expanding its data centre network over the coming years as it seeks greater control over its computing resources, rather than relying heavily on external providers.

Google has also acknowledged that demand for AI services is outpacing available infrastructure. During its recent financial reporting, the company said computing constraints had limited cloud growth and contributed to a growing backlog of customer demand. According to the Financial Times, Google has also signed a major agreement to lease additional AI computing capacity from SpaceX to meet rising demand for Gemini Enterprise.

The reported agreement is said to be worth around US$920 million per month and would give Google access to additional graphics processing units needed to run large AI models. The deal illustrates how even leading cloud providers are turning to external infrastructure partners as demand continues to accelerate.

AI boom increases pressure on providers

The reported limitations Meta faces reflect a broader challenge across the AI industry. As organisations deploy more generative AI tools for software development, customer support and business automation, demand for specialised chips and data centre capacity continues to rise faster than new infrastructure can be built.

While enterprise customers continue to increase spending on AI services, analysts have noted that infrastructure providers are still facing enormous operating costs. Revenue generated from AI products remains relatively small compared with the expense of building and operating large-scale computing facilities. At the same time, increasing demand has also pushed token prices higher for some users. These higher costs have reportedly prompted some companies to reduce AI usage or improve efficiency, including organisations that are developing AI models in-house.

The reported restrictions on Meta’s use of Gemini suggest that access to computing power has become one of the industry’s most valuable resources. As AI adoption expands across businesses and consumer services, securing sufficient processing capacity may prove just as important as developing increasingly capable language models.