New orbital data center solutions emerging to overcome terrestrial power and space constraints.

Elon Musk, Jensen Huang, Jeff Bezos, and Sam Altman have all begun openly discussing orbital data centers as a response to mounting energy availability and infrastructure constraints on land that increasingly threaten to bottleneck data center buildouts through the end of this decade and well into the 2030s. Nvidia-backed startup Starcloud has released a white paper arguing that a constellation of artificial intelligence data centers in low Earth orbit offers a practical solution to Earth's looming power crunch, cooling challenges, and permitting constraints.

As terrestrial data center projects reach capacity limits due to AI workloads scaling to multi-gigawatt levels, orbital data centers aim to bypass these constraints by leveraging near-continuous, high-intensity solar power, passive radiative cooling to deep space, and modular designs that enable rapid scaling and launch via SpaceX rockets. According to Starcloud's report, "Orbital data centers can leverage lower cooling costs using passive radiative cooling in space to directly achieve low coolant temperatures. Perhaps most importantly, they can be scaled almost indefinitely without the physical or permitting constraints faced on Earth, using modularity to deploy them rapidly." The startup added that "With new, reusable, cost-effective heavy-lift launch vehicles set to enter service, combined with the proliferation of in-orbit networking, the timing for this opportunity is ideal."

Starcloud has already demonstrated this concept by launching its Starcloud-1 satellite carrying an Nvidia H100 GPU, the most powerful compute chip ever sent into space. Using the H100, the company successfully trained NanoGPT, a lightweight language model, on the complete works of Shakespeare, marking the first AI model trained in space. Starcloud is also running Google's open-source LLM Gemma in orbit, representing the first time a high-powered Nvidia GPU has been used to operate a large language model in space.

As one potential interim solution before nuclear power generation scales up to meet demand, shifting some data center operations to low Earth orbit could help prevent bottlenecks while satisfying the rapid advances and ever-increasing power requirements of AI. This opportunity is spurring a space race-themed investment theme, with SpaceX planning to go public next year at a valuation of $800 billion, with Starlink positioned to power these space-based data centers.