Google has restricted Meta's access to its Gemini AI models after the social media company requested more computing capacity than Google could supply, disrupting internal AI projects at Meta, the Financial Times reported. The move highlights how even the largest tech firms are hitting compute limits as AI demand outpaces infrastructure.
The report said around March, Google informed Meta it could not fulfill the full Gemini order. The shortfall delayed some of Meta's AI work. Other Google clients were also affected, though less severely, due to Meta's exceptionally high demand. Reuters could not independently verify the report, and Google and Meta did not comment outside business hours.
Capacity limits trigger efficiency push
Meta has since told employees to be more efficient with AI tokens, the units that measure AI usage, according to the FT. AI for Developers Courses often address token optimization and resource management, skills that are becoming vital as companies confront similar constraints.
The restrictions come as Google Cloud's revenue hit $20 billion in the first quarter, a figure that CEO Sundar Pichai said could have been higher if not for computing power shortages. He noted those constraints caused the cloud unit's backlog to nearly double quarter over quarter.
Growing AI demand strains infrastructure
Even as companies invest billions in chips and data centers, they still struggle to secure enough computing power for AI services. Meta's heavy reliance on Google's Gemini models, which are covered in Google Gemini Training resources, illustrates the pressure on cloud providers. The capacity shortfall forced Meta to adjust its internal timelines and reflects a broader industry challenge.
Why this matters for IT and development professionals
For technical teams, the episode underscores that AI compute is not an infinite resource. Capacity planning, token budgeting, and efficient model usage are shifting from nice-to-have skills to operational necessities. When even a company with Meta's resources gets throttled, developers at every level need to anticipate similar bottlenecks in their own projects.
Your membership also unlocks:
