SpaceX Bets 100% on Nvidia Vera Rubin With $37B in 2026 AI Capex

For AI infrastructure and model teams, this is a landmark single-customer validation of Nvidia's next-generation Vera Rubin platform. SpaceX grew AI compute capex from $749 million a year ago to $15.8 billion in Q2 and now directs 86% of capex to Nvidia, with CFO Bret Johnsen signaling roughly $37 billion more in AI capex this year. The decision effectively removes the nearest AMD alternative from one of the most aggressive compute build-outs on record.

Elon Musk used SpaceX's first earnings call as a publicly traded company on August 4, 2026, to announce that the company would build its AI computing infrastructure exclusively on Nvidia hardware, specifically citing the upcoming Vera Rubin architecture as the best AI computer. The decision formalizes a single-supplier relationship for what has become SpaceX's largest capital outlay. It is not a marginal procurement choice: Musk is committing the most expensive, fastest-growing line item on SpaceX's balance sheet to one vendor, while Nvidia is simultaneously a major SpaceX shareholder with roughly $21 billion in shares at the end of June. The move locks in both companies at a moment when SpaceX's compute spending is growing at an extraordinary rate.

“

SpaceX grew AI compute capex from $749 million a year ago to $15.8 billion in Q2 and now directs 86% of capex to Nvidia, with CFO Bret Johnsen signaling roughly $37 billion more in AI capex this year.

The numbers behind the commitment show how quickly AI computing has become central to SpaceX. In the second quarter of 2026, SpaceX reported total capital expenditures of $18.4 billion, of which $15.8 billion went to AI computing infrastructure. That means 86 cents of every capital dollar is now being spent on AI compute. A year earlier, the comparable line item was only $749 million. In the first quarter of 2026 it was $7.7 billion. That is a more than 20-fold year-over-year increase in the computing line item. CFO Bret Johnsen told analysts that the next two quarters would be very similar to the second quarter, implying about $37 billion in additional capital spending later in the year. SpaceX also disclosed that it ended June with 1.4 gigawatts of installed computing capacity.

By going exclusive with Nvidia, SpaceX is locking in several things at once. It secures access to Nvidia's roadmap and the Vera Rubin architecture, which SpaceX leadership views as the strongest available AI compute platform. It also deepens an existing partnership that includes co-engineering and close coordination. But exclusivity also removes competitive leverage. SpaceX will no longer be able to play Nvidia against AMD or other GPU suppliers to negotiate price or guarantee supply. It concentrates operational risk in Nvidia's production and delivery schedule, which matters when SpaceX is planning to spend tens of billions of dollars in a matter of quarters. For shareholders, this cuts both ways: if Nvidia maintains architectural leadership, SpaceX gets best-in-class performance and likely priority treatment; if Nvidia hits supply constraints or execution problems, SpaceX has no easy substitute.