FULL STORY

Musk's Space Compute Ambition with Nvidia

Elon Musk unveils an aggressive compute expansion, partnering with Nvidia to build space-based data centers using next-gen Rubin GPUs, while committing SpaceX and xAI exclusively to Nvidia hardware.

2026-08-05 ~ 2026-08-06 · 2 episodes · 11 posts

Episode 1 · SpaceX and NVIDIA Partner to Send Vera Rubin AI Computing into Space (2026-08-05, 4 posts)

SpaceX is partnering with NVIDIA to design the Starmind AI1 satellite computing payload. The initiative aims to launch satellites equipped with next-generation Rubin GPUs and Vera CPUs, bringing AI factory-level computing power into orbit.

Episode 2 · Musk Says SpaceX and xAI to Use Nvidia Exclusively, Tripling Compute by 2027 (2026-08-05, 7 posts)

At SpaceX's first earnings call, Musk revealed an aggressive AI compute expansion plan, stating that SpaceX and xAI will exclusively use Nvidia hardware, with compute capacity tripling by the end of 2027. This clarifies his deployment roadmap and directly benefits Nvidia.

Confirmed

  • Compute expansion targets: Over 2 GW online by end of this year; cumulative compute "closer to 10 GW than 5 GW" by end of next year; more than triple current capacity by end of 2027. Musk also said not to limit to terawatt-scale superfactories, aiming for petawatt-scale.
  • Exclusive Nvidia adoption: Musk confirmed exclusive use of Nvidia hardware, locking in Nvidia's future Vera Rubin chips for SpaceX. He also discussed with Nvidia and expects a significant portion of GPU capacity next year.
  • Infrastructure investment: SpaceX's AI infrastructure spending in Q2 was about $200 million. Musk plans to apply rocket and satellite manufacturing tech to ground data centers to meet future power and cooling demands.

Why it matters

  • Revenue potential: Melius Research estimates that if SpaceX purchases 6 GW of GPU compute in 2027, it could add over $200 billion in revenue for Nvidia that year; even a conservative 2-3 GW would add about $100 billion.
  • AI arms race: Musk claims his AI team's deployment speed and efficiency are industry-leading, intensifying the compute arms race in large models.