Google has reportedly limited Meta's use of the Gemini AI models due to compute capacity constraints for the cloud provider.According to a report from the Financial Times, three people familiar with the matter said that Google told Meta in March that it could not purchase all the capacity it wanted.Those sources said that this has limited some of Meta's internal AI projects.Other customers have also been limited in access to compute, but to a lesser extent.Google and Meta declined to comment to the FT.Reports previously emerged that Meta was limiting employee AI token usage in an effort to cut costs.Meta and Google reportedly signed a six-year cloud computing agreement in August 2025, said to be worth more than $10 billion.That the hyperscalers are operating in capacity-constrained environments is no surprise, with the issue frequently being mentioned during the quarterly earnings calls for Google, Amazon Web Services, and Microsoft.Amin Vahdat, Google's cloud VP, said in November 2025 that the company would need to double AI capacity every six months to meet demand.To meet this demand, all three companies are rapidly investing in expanding their compute capacity, with Google increasing its capex expectations for 2026 earlier this year to between $180 and $190 billion, noting that it expects this to "significantly increase" in 2027.With this in mind, just earlier this month Google was looking to raise $84.75bn in equity to fund its "AI compute infrastructure to meet its unprecedented customer demand."The company has also notably signed a major $920 million-per-month lease with SpaceX-xAI for access to its AI data center capacity.