Google has placed limits on Meta’s use of its Gemini AI models after the social media company asked for more computing capacity than Google could provide, according to a Financial Times report cited by Reuters on June 28, 2026.
The report said Google, which is owned by Alphabet, told Meta around March that it could not supply the full amount of Gemini capacity Meta wanted to buy. The shortage reportedly disrupted and delayed some of Meta’s internal artificial intelligence projects.
Reuters said it could not immediately verify the Financial Times report, which cited people familiar with the matter. Google and Meta did not immediately respond to Reuters requests for comment outside business hours.
Meta Faces Limits on Gemini AI Use
Meta, the parent company of Facebook, Instagram, WhatsApp, and Threads, has been investing heavily in artificial intelligence. The company has been using AI models and computing systems to support internal projects and improve its products.
According to the Financial Times report, Meta was especially affected because its demand for Google’s Gemini AI models was unusually high. The report said several other Google clients were also affected by capacity limits, though not as much as Meta.
Gemini is Google’s family of artificial intelligence models. These models can help process and generate text, code, images, and other types of information. Companies often use such models through cloud services, which require large amounts of computing power.
AI Tokens Become a Focus
The report said Meta has encouraged staff to use AI tokens more efficiently because of the restrictions. AI tokens are units used to measure how much text or data an AI model processes.
For example, when a user enters a prompt into an AI system, that input is broken into tokens. The system’s response also uses tokens. More tokens usually mean more computing power is needed.
By asking staff to be more careful with tokens, Meta may be trying to reduce pressure on the limited AI capacity it can access. The report did not say which specific Meta projects were delayed.
Growing Demand for AI Computing Power
The reported limits show a broader problem in the AI industry. Many companies are spending billions of dollars on chips, servers, and data centers. Still, demand for AI computing power continues to grow faster than supply in some areas.
AI models need powerful chips and large data centers to run. This is especially true for major companies that use AI across many teams, products, and services.
Google Cloud has been one of the major providers of AI and cloud computing services. In the first quarter ended in March, Google Cloud revenue grew to $20 billion. Alphabet CEO Sundar Pichai said computing power constraints stopped the cloud unit from growing even faster.
Pichai also said the cloud unit’s backlog nearly doubled quarter on quarter. A backlog can show that customers want more services than a company can immediately provide.
Competition and Dependence in AI
The report also highlights a complex relationship between major technology companies. Google and Meta are rivals in online advertising, social media, and artificial intelligence. At the same time, large tech companies may still rely on each other for cloud services, AI tools, or computing resources.
Meta has been building its own AI systems, including its Llama model family. But the reported use of Google’s Gemini models suggests that even major AI companies may seek outside capacity when demand is high.
Google, meanwhile, is working to expand its AI infrastructure while serving many customers through Google Cloud. Capacity limits could affect how quickly customers can develop and test AI projects.
What’s Next
The Financial Times report suggests that AI computing power remains a major challenge for the technology industry. Meta’s reported delays show how capacity shortages can affect even the largest companies.
Google, Meta, and other AI firms are expected to keep investing in chips, cloud systems, and data centers. For now, the race to build and run advanced AI models continues to depend not only on software, but also on access to enough computing power.
