Particle.news
Download on the App Store

SpaceX Will Build Its AI Infrastructure Exclusively on Nvidia's Vera Rubin

The pledge concentrates massive GPU demand with one supplier, strains high‑bandwidth memory supplies, and makes upcoming company reports the key test of whether orders turn into delivered hardware.

Overview

  • SpaceX publicly committed to buy only Nvidia's Vera Rubin platform and has begun placing Rubin GPUs and Vera CPUs in its Starmind AI1 satellite for orbital AI workloads.
  • The company recorded $18.37 billion in quarterly capital spending with $15.83 billion earmarked for AI infrastructure and it is targeting roughly 2 gigawatts of compute by late 2026 and about 10 gigawatts by the end of 2027.
  • Industry research firms report shortages and yield issues for high‑bandwidth memory and DRAM that have prompted Nvidia to consider lower‑memory Rubin Ultra configurations, a constraint that could slow deliveries or change product specs.
  • Markets reacted immediately: Nvidia shares rose roughly 4% on the buying pledge while SpaceX stock dropped double digits as investors focused on the size of the spending and the lack of forward capex guidance.
  • The next inflection points are Nvidia's late‑August results and any future SpaceX capex guidance, which together will show whether the announced exclusive deal becomes confirmed backlog and deliverable hardware; SpaceX's own Terafab chip‑fab plans add a longer‑term variable to supply dynamics.