Google has placed limits on Meta’s access to its Gemini AI models because Meta’s demand for compute capacity exceeded what Google Cloud could supply. This development reported by the Financial Times on June 28, 2026 is the clearest signal yet that the AI boom has hit a hard physical ceiling. Even the world’s most powerful AI companies cannot build data centers fast enough to keep up with what they are selling.
The situation is striking for a simple reason: these two companies are fierce rivals in digital advertising. Yet Meta had become one of Google’s most dependent AI customers quietly relying on Gemini models to power everything from ad targeting tools to platform safety systems. Now, that dependence is catching up with them.
What Happened: Google Tells Meta It Cannot Meet Gemini Demand
Around March 2026, Google Cloud informed Meta that it was unable to fulfil the full volume of Gemini computing capacity that Meta had sought to purchase. The shortfall was not minor. Meta’s usage of Google’s AI models had grown so large relative to other customers that it became one of the hardest-hit clients when Google began rationing access.
Multiple internal AI projects at Meta were disrupted and delayed as a result. Neither Google nor Meta responded to media requests for comment. The report, originally from the Financial Times and confirmed by additional sources, cited people familiar with the matter.
Why Google Is Running Out of Compute and What It Is Doing About It
Google is not struggling because of poor planning. It is struggling because demand for AI compute has grown faster than anyone in the industry modelled. Gemini API requests more than doubled between March and August 2025 alone. That kind of growth rate breaks even the best capacity forecasts.
The SpaceX Deal: $920 Million Per Month for Bridge Capacity
Google’s most dramatic response came on June 5, 2026, when it signed a cloud service agreement with SpaceX to lease approximately 110,000 Nvidia GPUs alongside CPUs, memory, and related systems from October 2026 through June 2029. The monthly cost is $920 million, making it a deal worth roughly $30 billion over its lifetime.
This is a short-term, timely agreement to ensure we have bridge capacity to meet surging customer demand for our agent platform, Gemini Enterprise, which has been even higher than we expected.Google Cloud spokesperson, June 2026
Google has framed the agreement explicitly as bridge capacity a temporary solution while its own infrastructure programme continues to scale. SpaceX must deliver the committed GPU access by September 30, 2026, or face termination clauses. The deal also gives either party an exit option with 90 days’ notice after December 31, 2026.
Alphabet has already committed to between $180 billion and $190 billion in capital expenditure in 2026, up from earlier estimates. It also announced an $80 billion equity sale, including a $10 billion investment from Berkshire Hathaway, specifically to fund this infrastructure expansion. These are not small numbers. They are a sign of how seriously Google views the compute gap as an existential competitive risk.
Frequently Asked Questions
1. Why did Google cap Meta’s access to Gemini AI?
Google capped Meta’s Gemini AI access because Meta’s demand for compute capacity exceeded what Google Cloud could supply. Around March 2026, Google informed Meta it could not fulfil its full Gemini computing request, disrupting several internal AI projects at Meta as a result.
2. How does the Gemini capacity cap affect Meta’s AI operations?
The restrictions forced Meta to instruct employees to use AI tokens more efficiently. Several internal AI projects including platform safety work, ad targeting, and coding tools were reportedly slowed or delayed. Meta is now accelerating its migration toward proprietary, in-house AI models to reduce future exposure.
White House Tells OpenAI to Slow-Roll GPT-5.6 Release Over Safety Fears