In March of this year, Google sharply restricted Meta's access to its Gemini AI models. The reason was a simple shortage of computing resources to handle the entire volume of requests from Meta. This dealt a serious blow to the company's internal projects in the field of artificial intelligence: some of them were disrupted or significantly slowed down. The restrictions have not yet been lifted, creating additional pressure on Meta.
Against the backdrop of this disruption, and as part of a broader cost-cutting policy, Meta urged its employees to use tokens—a key resource for interacting with Gemini—more sparingly. This is a direct indication that even for a giant like Meta, access to cutting-edge AI models is not unlimited and can be restricted at the most inopportune moment.
Other Google clients also faced limits, but to a much lesser extent. This highlights the uniqueness of the situation: Meta, being one of the largest consumers of AI services, found itself hostage to its own ambitions and Google's limited infrastructure.
Situation Analysis
This incident is a clear signal that the AI computing market is overheated. Google, like other providers, is struggling to keep up with demand. For Meta, which is actively investing in AI to improve its social platforms and advertising algorithms, such restrictions pose a serious threat. In my professional opinion, this situation will force major technology companies to invest more aggressively in their own computing capabilities to avoid dependence on external providers. Over the next 12-18 months, we will see an increase in the number of proprietary data centers and specialized chips among giants like Meta, Amazon, and Microsoft.