NVIDIA has just unveiled a new frontier for artificial intelligence with its 'Space Computing' initiative.
At the heart of this new vision is a very down-to-earth problem: AI data centers are becoming incredibly power-hungry and hot. NVIDIA's latest systems, like the GB200 NVL72, can draw over 130 kilowatts per cabinet—that's enough energy to power dozens of homes. This immense heat forces data centers to rely on expensive and complex liquid cooling systems. In fact, the demand for liquid cooling in AI data centers is expected to more than double in just one year. This is why NVIDIA's vision of a future without bulky 'chillers' is so compelling.
So, where can you find a place with abundant solar power and a natural, perfect vacuum for cooling? Space, of course. The core idea of 'Space Computing' is to move some AI tasks, especially AI inference, into orbit. Satellites can be powered by the sun and cool themselves naturally through radiative cooling, shedding heat directly into space. This bypasses the energy grid and physical space constraints we face on Earth.
This isn't science fiction anymore, thanks to two key developments. First, optical backhaul—using lasers to send data between satellites and back to Earth—has matured significantly. Companies like SpaceX are already commercializing these high-speed links, and recent tests have shown gigabit-speed connections are possible. This solves the critical problem of getting data to and from an orbital data center. Second, NVIDIA’s own technology stack is prepared. Their next-generation Rubin architecture is entering production, and they've been building a powerful ecosystem of open models for 'physical AI' in robotics (like the GR00T family) and healthcare (BioNeMo). It's a complete package of hardware and software ready for this new environment.
We've already seen the first steps. A company called Starcloud successfully launched an NVIDIA H100 GPU into orbit late last year, proving the hardware can operate in space. For investors, this announcement introduces a new, long-term growth story. While the immediate revenue will be small, it creates 'option value'—the potential for a massive new market in orbital computing. NVIDIA is positioning itself not just as a chipmaker, but as the foundational architect for AI everywhere, including the final frontier.
- AI Inference: The process of using a trained AI model to make predictions or decisions on new, live data.
- Optical Backhaul: Using laser-based communication links to transmit large amounts of data, in this case, between satellites and Earth.
- Radiative Cooling: The process of dissipating heat by emitting thermal radiation, which is highly effective in the vacuum of space.