Vera Rubin Wins: Musk Picks Nvidia for SpaceX's 10 GW AI Buildout

For AI researchers and infrastructure engineers, SpaceX's exclusivity provides a real-world validation of Nvidia's Vera Rubin architecture at unprecedented scale. A 10 GW compute fleet would create one of the world's largest non-hyperscaler training and inference clusters, with implications for model capabilities, distributed training, and edge-AI deployment patterns.

SpaceX's first earnings call as a publicly traded company turned into a defining moment for the AI accelerator market: CEO Elon Musk said the company will build its AI computing infrastructure "exclusively on Nvidia," citing the superiority of the Vera Rubin architecture. The exclusivity is not a signed supply contract SpaceX has disclosed. It is a stated intention and an "understanding" regarding supply, as The Motley Fool emphasized. But the financial scale is enormous. In Q2 2026, SpaceX reported capital expenditures of $18.4 billion, $15.8 billion of which — 86 cents of every capital dollar — went to AI computing infrastructure. That line item was $7.7 billion in Q1 2026 and just $749 million a year earlier, meaning it grew more than 20 times year over year.

“

In Q2 2026, SpaceX reported capital expenditures of $18.4 billion, $15.8 billion of which — 86 cents of every capital dollar — went to AI computing infrastructure.

The company ended June with 1.4 gigawatts of installed computing capacity, up from 1.0 GW in March and 0.4 GW a year earlier. Musk expects SpaceX to finish 2026 with more than 2 GW, and to be "closer to 10 gigawatts of compute than 5 gigawatts" by the end of 2027. CFO Bret Johnsen indicated the next two quarters will be "very similar" to Q2, implying about $37 billion more in capital spending this year. Under the new commitment, every gigawatt added from now on would use Nvidia hardware, making SpaceX one of Nvidia's most important AI infrastructure customers.

The market reaction underscored the stakes. Nvidia shares closed up 3.4% the day after the call. Advanced Micro Devices, the nearest alternative GPU supplier, closed down 7% after reporting its own quarterly results that same afternoon. The exclusivity announcement likely compounded investor concerns about AMD's competitive position, though the decline was not solely attributable to Musk's statement. Nvidia also held about $21 billion in SpaceX shares at the end of June, aligning the supplier's interests with SpaceX's growth. That dual role as supplier and shareholder can soften concerns about extraction, but it does not eliminate single-vendor risk.