Google Rations Gemini To Meta As Demand Finally Outruns Supply

Google restricted Meta's access to its Gemini AI models around March, unable to supply the compute Meta wanted even as its cloud backlog swelled toward $460 billion.

Key Points:

Google Rations Gemini Supply

Google informed Meta around March that it could not fully meet the Gemini capacity the company hoped to buy, three people familiar with the talks reported. The limits disrupted several of Meta's internal AI projects, some tied to coding, advertising tools, and content moderation across Facebook and Instagram. Managers told engineers to use AI tokens, the units that measure model usage, more sparingly.

Other Google clients felt the squeeze, though Meta absorbed the hardest hit because its demand ran far above most customers. Both Google and Meta declined to comment on the terms.

In May the search giant made the caps formal, imposing usage limits across its Gemini apps. Access now scales with available capacity, not with how much a customer is willing to spend. That single change rewired a basic assumption about cloud AI.

Also Read: Bitcoin Holds $62K As HYPE And ADA Ignite Weekend Crypto Rebound

Pichai Warns On Compute

Google is not short on demand, it is drowning in it, which is what makes the cap so counterintuitive. Cloud revenue cleared $20 billion in a single quarter for the first time, up roughly 63% from a year earlier.

Chief executive Sundar Pichai acknowledged the strain on the earnings call, where he said the company was "compute-constrained in the near term." For any firm building on outside AI platforms, the message lands hard. A signed enterprise contract no longer guarantees the compute a company plans around, no matter the price.

Meta had leaned on Gemini because it beat the firm's own Llama models at scrubbing scams and harmful posts. The cap sped up its pivot to a first in-house model, Muse Spark, as it cuts thousands of jobs and steers up to $135 billion into AI this year.

Google's own bind runs deep.

It agreed to pay SpaceX roughly $920 million a month for about 110,000 Nvidia GPUs, a stopgap several outlets confirmed. For every dollar of committed demand, the company spends only about 40 cents on new capacity, so the gap keeps widening rather than closing.

Read Next: Claude Fable 5 Coding Drop Reveals A Router Problem, Not Model Decay