Google has restricted Meta’s access to its Gemini AI models after the Facebook and Instagram owner requested more computing capacity than Google could provide, according to a report by the Financial Times.
The report said Google informed Meta around March that it could not supply the full Gemini capacity the company wanted to buy. The shortfall reportedly delayed some of Meta’s internal AI projects and forced teams to adjust their work.
Meta was not the only customer affected. However, the report said the impact on the social media company was greater because of its unusually high demand for Google’s AI models.
Neither Google nor Meta responded to requests for comment outside normal business hours.
As a result of the restrictions, Meta reportedly asked employees to use AI tokens more carefully. AI tokens are the units used to measure activity on AI models and play a key role in managing computing resources.
The reported limits highlight the problem facing the technology industry. Companies invest billions of dollars in chips and data centres, yet demand for computing power still exceeds available supply as businesses expand their AI services.
Google has already acknowledged those capacity challenges. During the company’s first-quarter earnings report, Chief Executive Sundar Pichai said computing power limitations prevented Google Cloud from growing even faster.
He added that the cloud division’s backlog almost doubled from the previous quarter, despite generating $20 billion in revenue during the period.
Meta has been expanding its AI efforts across Facebook, Instagram and WhatsApp, increasing demand for the computing power needed to train and run large AI models.
The company is also developing new AI agents and expanding its Llama family of models, making access to reliable computing infrastructure increasingly important.
The reported restrictions also reveal that even the world’s biggest technology companies are competing for limited AI resources.
Demand for advanced chips and computing capacity overtakes supply, slowing some projects across the industry despite record levels of investment.
