Pakistan, Iran Move to Strengthen Scientific and Higher Education Cooperation
June 29, 2026
Google capped Meta’s Gemini AI access after it requested more capacity than available, disrupting Meta projects and forcing token cuts.
Google has placed limits on Meta’s access to its Gemini AI models after the social media giant sought more computing capacity than Alphabet’s cloud could provide, the Financial Times reported. The restriction, communicated around March, has disrupted some of Meta’s internal AI projects and delayed timelines, sources told the FT.
Several other Google Cloud customers have been affected to a lesser degree, though Meta’s exceptionally high demand made it particularly vulnerable to the shortfall. The constraints prompted Meta to encourage staff to economize on AI tokens, the units that measure usage of model compute, the FT said.
The curbs come amid an industry-wide scramble for compute power as companies pour billions into chips and data centres. Google Cloud reported $20 billion in revenue for the quarter ended March, but CEO Sundar Pichai has acknowledged that computing-power constraints “prevented even higher growth,” contributing to a near doubling of the cloud unit’s backlog, Reuters reported.
Meta initially relied on Gemini for automation tasks including safety moderation, but the company has been shifting toward its own models such as Muse Spark and the open-source Llama variants to reduce external dependence. Bloomberg noted that Meta’s AI push has reshaped workforce plans: earlier this year it reassigned 7,000 employees to AI roles and announced cuts affecting roughly 10% of staff.
Industry observers say the episode highlights how limited supply of large-scale compute can influence which companies control critical AI services. “We’re seeing real-world limits on capacity that are reshaping product roadmaps,” one person familiar with the matter told the Financial Times.
Google has also pursued other routes to secure capacity; in June it agreed to pay SpaceX about $920 million a month under a multi-year cloud deal to expand compute options. Meanwhile, neither Google nor Meta provided immediate comment to the FT or Reuters outside business hours, according to reports.
As demand for generative AI soars, companies face trade-offs between buying external model access and investing in proprietary infrastructure. For Meta, the Google limits represent both a bottleneck and an incentive to accelerate in-house model development.
