Google has placed limits on Meta Platforms Inc’s use of its Gemini artificial intelligence (AI) models because it could not provide as much computing capacity as the social media company wanted, according to the Financial Times.
In the latest sign of AI infrastructure constraints, Alphabet Inc’s search giant has enforced restrictions on several clients, with Meta particularly affected, the FT reported on June 28. The move has had a knock-on effect on Meta’s internal projects, and meant the company has told staff to make more efficient use of AI tokens, the newspaper reported, citing three unidentified people familiar with the matter.
Google and Meta both declined to comment to the FT.
Meta had initially relied on Gemini, which had proven better than its own Llama open-source models, to automate safety processes like removing harmful content and wiping out scams, the newspaper said. It has, though, been increasingly using its new Muse Spark model as it looks to lessen reliance on external models.
The AI boom is testing the limits of compute and the energy required to power the profusion of data centres cropping up. Earlier in June, Google agreed to pay Elon Musk’s SpaceX US$920mil (RM3.74bil) a month for computing power as part of a US$30bil (RM121.83bil) cloud services deal running through mid-2029 as it races to meet the ravenous demand for capacity.
Meta, which doesn’t sell cloud computing services, is dialling up spending on AI as the top priority of Chief Executive Officer Mark Zuckerberg, with the technology now key to his corporate vision.
Earlier this year, the social media company told staff of plans to cut 10% of workers, or 8,000 roles, in an effort to offset its heavy spending. It has also reassigned 7,000 workers to new jobs related to AI as part of the wider corporate restructuring. – Bloomberg
