Elon Musk plans first Nvidia-powered AI satellite launch in 2027
Elon Musk's SpaceX and Nvidia are preparing to test orbital AI computing with a first launch targeted for late 2027.
SpaceX and Nvidia are working on a plan to send powerful AI computing hardware into orbit, with Elon Musk saying the first launch is expected in the fourth quarter of 2027.
The project centres on a space-optimised version of Nvidia’s Vera Rubin NVL72 system, which is being prepared for use in a first-generation Starmind AI satellite.
Today, most advanced AI systems rely on huge data centres on Earth, packed with specialised chips, cooling systems and high-power networks. SpaceX and Nvidia are now exploring whether some of that computing could move into orbit.
What SpaceX and Nvidia are building
According to Nvidia’s announcement, SpaceXAI is expanding the infrastructure behind Grok, Elon Musk’s AI chatbot, using Nvidia’s Vera Rubin platform. The technology is also being adapted for space through the planned Starmind AI satellite.
The Vera Rubin NVL72 is a rack-scale AI computing system that combines processors, graphics chips, memory and networking for demanding AI workloads. For space, the system needs to be lighter, denser and more efficient than a conventional data centre rack.
Elon Musk has said the design is intended to be simpler, lower-cost, and lighter than conventional systems. That matters because every kilogram launched into orbit adds complexity and expense. The hardware must also deal with radiation, limited power, heat management and the need for high reliability.
Why AI computing in orbit matters
Orbital AI systems could process data closer to where it is generated, including from satellites, Earth observation tools and space-based communication networks. For instance, a satellite with onboard AI could analyse images, signals or sensor data before sending only the most important results back to Earth.
This could save bandwidth and support faster decision-making in areas such as climate monitoring, disaster response, navigation and communications. Nvidia has also linked the project to agentic AI, which refers to systems that can carry out tasks, use tools, process code, analyse data and support simulations.
Such systems require significant computing power, increasing demand for more efficient AI infrastructure.
The link to Grok and what comes next
SpaceXAI’s planned use of Nvidia’s Vera CPUs is expected to support the next generation of AI applications behind Grok. Nvidia says these systems need strong CPU performance and high memory bandwidth so AI agents can manage orchestration, code execution and data processing while graphics chips focus on core AI workloads.
The first launch is planned for Q4 2027, with Musk suggesting significant scaling could follow in 2028. However, the project remains technically challenging. Space-based AI computing will need to operate reliably within power and cooling limits while justifying the cost of launching advanced hardware into orbit.
If successful, the Starmind AI satellite could become an early step towards a new layer of computing infrastructure in space. It would not replace Earth-based data centres soon, but could eventually see AI workloads shared between the ground and orbit.