Google has imposed limits on Meta's consumption of its Gemini AI models after the social media company requested more model capacity than Google could provide, according to a Financial Times report. The restriction, which Google communicated to Meta around March, left some of Meta's internal AI projects disrupted or delayed.
The report says several other Google clients experienced similar restrictions, though the impact on those customers was smaller. Meta's situation was singled out because of its exceptionally high demand for access to Google’s Gemini models.
As a result of the limits, Meta has told employees to be more conservative with "AI tokens" - the units used to measure and bill AI usage - in an effort to stretch available capacity while internal projects proceed.
The Financial Times account also notes wider industry strain: even as firms invest billions in chips and data centres, demand for computing power to run AI workloads is outpacing supply. Google Cloud reported $20 billion in revenue in the first quarter ended March, but Chief Executive Officer Sundar Pichai said that constraints on computing capacity held back even stronger growth and helped push the cloud unit's backlog to nearly double quarter on quarter.
The report could not be immediately verified, and both companies did not immediately respond to requests for comment outside business hours, the report said.
Context and implications
The limits placed on Meta's usage of Gemini models reflect a wider bottleneck in AI infrastructure capacity. The details in the report emphasize that allocation decisions by major cloud providers can have disruptive effects on large-scale internal AI development programs. The situation also illustrates how cloud capacity constraints can shape growth outcomes for cloud vendors themselves, as cited by Google Cloud's backlog and CEO commentary.
What was reported
- Google informed Meta around March that it could not fulfil the full Gemini capacity Meta wanted to buy.
- The shortfall disrupted and delayed some of Meta's internal AI work.
- Other Google customers were affected to a lesser extent.
- Meta has encouraged staff to conserve AI tokens to manage the restriction.
- Google Cloud posted $20 billion in revenue for the quarter ended March, with capacity limits cited by the CEO as a factor in both growth moderation and a near doubling of the cloud backlog quarter on quarter.
Data notes
The article presents reported developments about access limits, internal effects at Meta and related comments about Google Cloud's revenue and backlog. The limitations on verification and the companies' lack of immediate comment are noted in the report.