Google has reduced the AI compute available to some of its customers, and the restrictions have affected Facebook and Instagram parent company Meta. According to the Financial Times, Google notified Meta this spring that it could not provide the full Gemini capacity the company requested.
What happened and when?
Rising demand for complex AI systems has increased pressure on infrastructure. In response, Google has scaled back access for certain customers. The Financial Times reports that these measures began in the spring and specifically limited the Gemini compute resources that Meta had sought.
Why this matters
Running advanced AI models requires enormous compute power, which depends not only on specialized chips but also on large-scale data centers and substantial energy supplies. When major users such as Meta do not receive requested capacity, internal AI projects can slow down.
Other customers and the context
Google has applied similar restrictions to other clients, though the extent varies by customer. Faced with growing demand, Google is searching for additional capacity and reportedly plans to expand by involving external service providers.
Meta's position
Meta develops its own AI models while also relying on external solutions. The company is making significant investments in building out its own AI infrastructure to reduce long-term dependence on third-party providers. This situation highlights that the competition in AI now includes not only model quality but also who can secure sufficient compute capacity to run those models.
Conclusion
The Google restrictions — reported by the Financial Times in spring — illustrate that infrastructure and energy constraints are currently a bottleneck for AI development. Companies are investing billions, but demand still outpaces available resources, reshaping supply chains and strategic investments across the industry.
