SpaceXAI Plans NVIDIA Vera Rollout for Grok and Orbit, Targeting Several Gigawatts
Updated
Updated · igorslab.de · Aug 24
SpaceXAI Plans NVIDIA Vera Rollout for Grok and Orbit, Targeting Several Gigawatts
3 articles · Updated · igorslab.de · Aug 24
Summary
NVIDIA said on Aug. 24 that SpaceXAI plans to use Vera processors for CPU-heavy agentic AI work, expand Grok infrastructure with Vera Rubin, and eventually adapt the platform for orbital computing.
88 Olympus cores, LPDDR5X memory and up to 1.2 TB/s bandwidth are the main disclosed Vera specs; NVIDIA also cited up to 1.8x faster task execution than x86, but gave no comparison systems or test methods.
Several gigawatts is the stated scaling goal for SpaceXAI’s compute buildout, though the announcement did not say how much capacity is installed now, which systems will be deployed, or on what timetable.
Starmind is the first named orbital project, built around an NVL72 rack optimized for space, where power, heat dissipation, bandwidth, radiation protection and ground links remain unresolved engineering hurdles.
No independent benchmarks, binding schedule or validated space deployment were provided, leaving the announcement as a strategic roadmap rather than proof of completed product integration.