SpaceX has partnered with Nvidia to develop the compute payload for its planned Starmind AI1 satellite constellation, a project aimed at creating orbital data centers capable of running artificial intelligence workloads in space. Under the partnership, each Starmind AI1 satellite will be equipped with Nvidia Rubin GPUs and Vera CPUs, the chipmaker’s latest AI computing architecture, delivering what the companies describe as “datacenter-class space compute.” The announcement represents one of Nvidia’s most significant entries into the emerging market for orbital AI infrastructure and positions SpaceX as one of the largest prospective customers for the company’s new space computing platform.
The agreement builds on SpaceX’s broader vision of relocating portions of AI infrastructure from Earth into orbit, where continuous solar power, the vacuum of space and radiative cooling could potentially reduce some of the energy and thermal constraints facing terrestrial data centers. Earlier this year, SpaceX filed plans with the US Federal Communications Commission (FCC) seeking approval for a constellation of up to one million AI computing satellites operating between 500 and 2,000 kilometers above Earth. The proposed network would be interconnected through high-speed laser links, allowing distributed computing workloads to be processed in orbit before transmitting results back to Earth through Starlink infrastructure.
AI Infrastructure Moves Beyond Earth
The Starmind initiative reflects the growing convergence between artificial intelligence and the commercial space industry. According to SpaceX, each AI1 satellite is designed to provide approximately 120 kilowatts of sustained AI computing capacity, peaking at around 150 kilowatts. The satellites will use large deployable solar arrays for power generation and liquid radiator systems that dissipate heat directly into the vacuum of space—one of the project’s principal engineering advantages over conventional terrestrial data centers.
Nvidia’s Rubin GPUs and Vera CPUs will form the core compute platform for the satellites, extending the company’s AI hardware ecosystem beyond traditional cloud data centers and edge computing into orbital infrastructure. The collaboration also strengthens Nvidia’s position in what it has begun describing as the “space computing” market, following earlier partnerships with aerospace and satellite companies developing in-orbit processing capabilities.
Regulatory and Technical Challenges Remain
Despite the announcement, significant hurdles remain before Starmind becomes operational. SpaceX’s proposed satellite constellation still requires regulatory approval from the FCC, and deploying even a fraction of the planned one million satellites would represent one of the largest commercial space infrastructure projects ever attempted.
The economics also remain uncertain. Building hyperscale AI infrastructure in orbit will require substantial investment in launch capacity, satellite manufacturing and long-term operations. While advocates argue that abundant solar energy and passive cooling could eventually offset those costs, critics question whether orbital computing can compete economically with increasingly efficient ground-based AI data centers. Nevertheless, investor interest in space-based computing has accelerated as artificial intelligence dramatically increases global demand for electricity and data center capacity. Companies across the aerospace and semiconductor industries are exploring whether orbital infrastructure can alleviate some of those constraints.
For SpaceX, the Nvidia partnership provides a crucial technology component for its ambitious Starmind vision. For Nvidia, it extends the company’s AI platform into an entirely new computing environment. Together, the collaboration signals that the race to build the next generation of AI infrastructure may extend well beyond Earth, positioning orbital data centers as a potential long-term complement to conventional cloud computing rather than a replacement for it.
