Google has limited Meta's access to its Gemini artificial intelligence models after the Facebook parent sought more computing capacity than the search major could provide, underlining how infrastructure has become a central pressure point in the global AI race.
According to a Financial Times report cited by Reuters, Google imposed limits on Meta's use of Gemini as demand for advanced model access and compute resources exceeded available capacity. The restrictions have reportedly affected several Google customers, with Meta among the most impacted because of the scale of its internal AI requirements.
The development comes as large technology companies increasingly depend on external AI models even while building their own systems. Meta has been using Gemini for a range of internal functions, including scam detection, customer support, advertising-related workflows and coding support, the report said. The company had chosen Google's model for some tasks because it performed better than its own Llama models in those areas.
The limits have reportedly delayed some internal Meta AI projects and prompted the company to ask employees to use AI tokens more efficiently. Meta is also said to be moving some workloads to its own newer model, Muse Spark, as it looks to reduce dependence on external providers for critical AI functions.
The reported restriction is significant because it shows that AI competition is no longer only about model quality, talent or product launches. Access to enough computing power, including chips, data centres and energy, is becoming equally important. Even companies with large AI investments are facing constraints as demand for generative AI tools continues to rise across consumer products, enterprise platforms and internal engineering teams.
Google has also faced broader pressure on AI capacity. The company has told investors that it remains compute constrained in the near term, even as Google Cloud continues to benefit from strong demand for AI infrastructure and services. Like other major cloud providers, Google is investing heavily in data centres and custom AI chips to support the growing use of its models.
For Meta, the issue highlights the challenge of scaling AI ambitions while relying partly on third-party infrastructure and models. The company has been investing aggressively in AI talent, data centres and model development as Chief Executive Mark Zuckerberg pushes toward what he has described as personal superintelligence. Meta has also committed hundreds of billions of dollars in long-term infrastructure spending to support its AI roadmap.
The reported Gemini limits come at a time when AI companies are competing to secure scarce resources needed to train and run advanced models. Nvidia chips, cloud capacity, energy supply and data centre availability have become strategic assets as companies race to add AI features to search, social media, productivity tools, coding platforms and advertising systems.
The development may also have implications for marketers and enterprises that are rapidly adopting AI tools. As AI usage becomes more compute intensive, companies may face changing pricing models, usage caps and availability limits across platforms. For businesses using AI for customer service, content production, campaign optimisation or software development, reliable access to models could become an operational planning issue.
Neither Google nor Meta has publicly detailed the commercial terms of their reported arrangement. However, the episode reflects a broader shift in the AI economy, where model access is increasingly tied to infrastructure availability rather than software capability alone.
As AI adoption deepens, the ability to supply compute at scale could determine which companies can serve large customers consistently. The reported limits on Meta's Gemini use suggest that even the biggest technology players are now navigating an AI market where demand is rising faster than supply.