Google rents SpaceX compute to bridge surging Gemini demand

Google has agreed to pay SpaceX about $920 million a month for compute capacity from October 2026 through June 2029, a 32-month contract worth roughly $29 billion. The deal gives Google access to around 110,000 NVIDIA GPUs plus supporting CPUs and memory at a SpaceX facility, with either side able to walk away on 90 days notice after the end of this year. SpaceX announced it a week before its IPO, which targets a $1.75 trillion valuation.

Google framed the contract as a stopgap. "This is a short-term, timely agreement to ensure we have bridge capacity to meet surging customer demand for our agent platform, Gemini Enterprise," the company said. That line is the interesting part. Google is one of the largest owners of AI compute on earth, and it is still buying capacity from a third party because Gemini Enterprise growth has run ahead of what its own buildout can deliver right now.

The shape of the contract echoes a bigger one. Anthropic locked in $1.25 billion a month from SpaceX through 2029 in May, drawing on the Colossus 1 facility near Memphis. Google's contract appears to draw on a different facility, possibly Colossus 2. Either way, the pattern is now clear: the frontier labs are renting compute from each other's neighbors on multi-year terms because waiting for their own next datacenter is no longer an option.

Why it matters

If you work on AI agents at the frontier, the supply side has changed. The labs are no longer waiting for their own builds. They are signing nine and ten figure monthly contracts to bridge the gap, and that pricing eventually feeds into what your API quota and rate limits look like.