SpaceX has signed a major artificial intelligence infrastructure agreement with Google ahead of its planned initial public offering, according to a regulatory filing. The deal will give Google access to about 110,000 Nvidia graphics processing units, along with CPUs, memory and related components housed in SpaceX data centers. The agreement is scheduled to run from October through June 2029, with payments of $920 million per month once capacity is fully available. A reduced fee will apply during the ramp-up period.
The filing says Google may terminate the deal if SpaceX does not provide the committed GPU capacity by Sept. 30, 2026, subject to a one-month grace period. After this year, either company may end the agreement with 90 days’ notice. Google Cloud said the arrangement is intended to provide temporary capacity for rising demand tied to Gemini Enterprise, its AI agent platform for business customers.
The agreement is the latest effort by SpaceX to generate revenue from AI infrastructure following its merger with xAI, Elon Musk’s artificial intelligence company. SpaceX recently announced another large compute arrangement with Anthropic involving its Colossus 1 data center in Memphis. The company has reported sharply higher capital spending, much of it directed toward AI, while its AI business posted an operating loss in the latest quarter.
SpaceX is preparing for an IPO that could value the company above $1.75 trillion. Its filings describe Google as both a customer and a competitor, including in connectivity and artificial intelligence. Google, meanwhile, is expanding AI spending and plans to raise additional capital to meet customer demand.
The deal also places SpaceX more directly in competition with infrastructure providers known as neoclouds, including CoreWeave and Nebius. Google and SpaceX previously partnered in 2021, when Google supplied cloud and networking services for Starlink. The new arrangement reverses those earlier roles.