Elon Musk is bringing forward SpaceX’s timeline for putting AI computing infrastructure into orbit, saying the company has designed a space-optimized NVIDIA Vera Rubin NVL72 system targeted for launch in the fourth quarter of 2027. The move places SpaceX closer to testing a radically different model for AI infrastructure at a time when the industry is facing intense demand for computing power and growing constraints around terrestrial data centers.

SpaceX Targets First Orbital AI Systems in 2027

Musk said SpaceX, working with NVIDIA, has designed a space-optimized version of the Vera Rubin NVL72 system for orbital deployment in Q4 2027, with significant scale expected in 2028. The system is intended to bring the same accelerated computing architecture used in next-generation AI factories on Earth into space.

NVIDIA has confirmed that SpaceXAI is expanding its use of the Vera Rubin platform for AI workloads and plans to extend an optimized Vera Rubin NVL72 system into orbit through its first-generation Starmind AI satellite. The companies are adapting the architecture to the very different requirements of orbital computing, including power management, thermal control, bandwidth, reliability and physical integration.

Ahead of SpaceX’s Earlier 2028 Timeline

The updated target represents an acceleration from SpaceX’s earlier public timeline. In regulatory filings submitted as part of its IPO process, the company said it expected to begin deploying orbital AI compute satellites as early as 2028 and acknowledged that the program relies on technologies that remain unproven or do not yet exist at the required scale.

That gap between the earlier filing and Musk’s latest target highlights both the speed of development and the uncertainty surrounding the project. Reaching orbit in 2027 would require progress not only in computing hardware but also in launch capacity, satellite design, thermal management and the economics of operating high-performance computing systems in space.

Why NVIDIA and SpaceX See an Opportunity in Orbit

The strategy reflects a broader effort to solve the infrastructure bottlenecks created by the AI boom. SpaceX is already expanding its terrestrial computing capacity rapidly and has said it is building exclusively on NVIDIA hardware, citing the Vera Rubin architecture as its preferred platform. The company expects to have more than 2 gigawatts of compute online by the end of 2026 and has indicated that cumulative capacity could approach 10 gigawatts by the end of 2027.

Space-based computing could eventually offer a new way to add capacity without relying entirely on land, electricity grids and conventional data-center infrastructure. The economics remain unproven, however, and launch costs, radiation exposure, maintenance, networking and thermal constraints could determine whether the concept can move beyond demonstration projects.

Investors will be watching SpaceX’s execution against the 2027 target, the development of Starship and the rollout of NVIDIA-powered systems. The first orbital deployment would be an important technical milestone, but achieving meaningful scale in 2028 will matter more for determining whether space-based AI can become a commercially viable extension of the rapidly expanding global compute market.