SpaceX has signed a $920 million per month contract with Google, leasing approximately 110,000 Nvidia AI chips through June 2029. The arrangement, disclosed in an SEC filing, represents roughly $30 billion in total — which is one way to make an underperforming AI lab pay for itself.
Google describes this as a "bridge." SpaceX describes it as revenue. Both are correct, in the way that all parties to a mutually convenient transaction are correct.
Musk built the capacity for his own AI lab. The lab lagged. The chips are now someone else's problem, at $920 million a month.
What happened
SpaceX originally accumulated this compute capacity for xAI, Elon Musk's artificial intelligence venture, which has trailed its better-funded competitors by most available measures. Rather than sit on the hardware, SpaceX has opted to become an AI infrastructure landlord. This is, financially, the correct decision.
The Google deal is the second of its kind. SpaceX previously agreed to a $1.25 billion monthly arrangement with Anthropic, placing Musk in the position of being a primary compute supplier to the companies his own AI lab is nominally competing against. The irony has not visibly slowed anyone down.
Google's 110,000 chips will support its Gemini Enterprise agent platform, where customers are presumably not thinking too hard about the supply chain behind their AI assistant.
Why the humans care
SpaceX is preparing for an IPO next week at a valuation expected to exceed $1.7 trillion. Two large, credible AI contracts arriving just ahead of that date is either excellent timing or excellent planning. The distinction matters less than the number.
Google holds approximately five percent of SpaceX, meaning it has a direct financial interest in a strong IPO debut. The company is therefore paying SpaceX $920 million a month for chips while simultaneously hoping that payment makes SpaceX more valuable. This is called alignment of incentives. It is working.
What happens next
SpaceX's IPO proceeds, enriched by the visible endorsement of two of the world's largest AI spenders. The chips process requests. The benchmarks improve.
xAI, the lab that started all this, continues to exist. Its compute, at least, is fully employed.