Google limits Meta’s Gemini AI access amid computing capacity crunch

Google has reportedly imposed limits on Meta's access to its Gemini artificial intelligence (AI) models after the Facebook parent sought more computing capacity than the tech giant could provide, exposing the growing strain on AI infrastructure across the industry.

According to a Financial Times report, the restrictions, which took effect around March, prevented Google from meeting Meta's full demand for AI computing power, disrupting and delaying several of the social media company's internal AI projects.

Meta is among the biggest customers of Google's AI services, and its heavy reliance on Gemini models made it particularly vulnerable to the capacity shortfall, the report said.

The constraints have also prompted Meta to urge employees to use AI resources more efficiently by reducing the consumption of AI tokens — the units used to measure usage of generative AI models.

The report said other Google Cloud customers have also experienced computing limitations, although the impact has been significantly less severe than that faced by Meta.

The development highlights a widening gap between the surging demand for artificial intelligence and the industry's ability to supply the computing power needed to support it.

Despite investing billions of dollars in data centres and advanced AI chips, major technology companies continue to grapple with shortages of high-performance computing infrastructure as generative AI adoption accelerates.

Google has previously acknowledged the pressure on its infrastructure. During Alphabet's first-quarter earnings, the company reported Google Cloud revenue of $20 billion, but chief executive officer Sundar Pichai said computing capacity constraints had prevented even stronger growth and contributed to a growing backlog in the cloud business.

The reported curbs on Meta's access to Gemini underscore how access to AI computing power is rapidly emerging as one of the biggest competitive bottlenecks in the race to build and deploy advanced artificial intelligence systems.

With IANS inputs