SpaceX commits exclusively to Nvidia GPUs for Starmind orbital AI satellites
SpaceX · Nvidia
On its first earnings call as a public company, SpaceX said it will use only Nvidia GPUs for its Starmind AI1 orbital compute satellites.
updates
- 2026-10-18Musk says the Starmind computer will reach volume production in 2027, after Nvidia's Jensen Huang detailed the space-data-center partnership publicly.
- 2026-10-04Musk later specified per-satellite power (250kW) and connectivity (10Tbps, with a path to 100+Tbps) figures for Starmind on X.
what happened
On SpaceX's first earnings call as a public company, held August 4, Elon Musk said the company is "exclusive to Nvidia" for its orbital AI compute hardware, per Forbes and Teslarati. Each satellite in the planned Starmind AI1 constellation will carry Nvidia Rubin GPUs and Vera CPUs, using Nvidia's Vera Rubin NVL72 rackscale architecture, codenamed Kyber, both on the ground and in orbit. SpaceX said it is targeting 2 gigawatts of compute capacity online by the end of 2026 and about 10 gigawatts by the end of 2027. The company's existing AI business, which leases Colossus compute capacity to customers including Anthropic and Google, generated $2.6 billion in revenue in the second quarter, up roughly 250% year over year, per Teslarati. Musk later gave per-satellite figures on X: each Starmind spacecraft will carry over 250kW of solar power, about 20% more than the International Space Station, with 10 terabits per second of bidirectional connectivity and "a path to 100+Tb," per posts on September 20 and October 1. On October 15, after Nvidia CEO Jensen Huang discussed the partnership on the BG2 Pod, Musk wrote on X that "the Star Mind computer being co-engineered by SpaceX and Nvidia is far better than almost anyone expects and will reach volume production next year," calling Huang's engineering team "incredibly talented"; Huang said space-based data centers are "one particular new market" the companies are exploring together, per Yahoo Finance.
why it matters
On SpaceX's first earnings call as a public company, Musk gave investors concrete, binding figures for the first time: exclusivity with one GPU supplier, a named architecture, and a compute-capacity ramp that ties Starmind's build-out directly to Nvidia's hardware roadmap rather than a vaguer "AI in space" ambition. For data center operators, cloud providers, and satellite manufacturers watching the orbital-compute market emerge, this sets the technical and timeline benchmark competitors (Google, AWS, and other proposed space-data-center efforts) will be measured against, and the 2 GW/10 GW targets give suppliers of launch, power, and thermal-management hardware for space-based AI a demand signal to plan against. Musk's October 15 production-timeline update, tying volume production to 2027, sharpens that demand signal for suppliers further up the Starmind chain.
for who
Cloud/AI compute buyers and satellite operators evaluating orbital data centers
signal-to-noise
quick facts
- Companies
- SpaceX, Nvidia
- Category
- partnership
- Impact
- major
- SNR
- 4 / 5
- Event date
- 2026-08-04
- Published
- 2026-08-07 17:35 UTC