SpaceX and NVIDIA are partnering to bring AI data centers into space using the Starmind AI1 satellite, but cooling a data center in space is not straightforward. The vacuum of space is not cold, and objects can heat up or cool down depending on their exposure to the sun.
The heat generated by AI hardware must be radiated away as infrared radiation, a slow process that creates a trade-off between computing power, radiator area, spacecraft mass, and operating temperature. A system running hundreds of kilowatts of AI hardware must reject nearly all of that power as waste heat.
Liquid cooling can carry heat away from chips, but it does not eliminate the requirement for extensive radiator surfaces. The Starmind satellites will have a deployable liquid radiator system measuring 160 square meters, but the choice of liquid is still unknown.
Hugh Lewis, a professor of astronautics at the University of Birmingham, expects the system to use ammonia, which is already used on the International Space Station. However, it remains to be seen whether this will reliably scale to data-center-class AI deployments with their enormous heat.
They will need to find a solution to this problem, as the success of the project depends on it.
The architecture depends heavily on Starlink’s optical inter-satellite links, which have published technology specifications describing mini laser terminals operating at up to 25 Gbps across distances as long as 4,000 kilometers. However, this does not establish that a satellite constellation can function like the tightly coupled networking fabric of a terrestrial AI supercomputer.
Related: New AI Tools Learn Your Company Data
Large-scale machine learning and training in space require huge, predictable bandwidth and very low latency for GPU-to-GPU communications, which is not currently possible with orbital networks. Physical propagation delays, laser-link acquisition and handoffs, routing across a moving constellation, and limits on available capacity per spacecraft all pose significant challenges.
Doug Mohney, a long-time space influencer, also raises concerns about debris and space warfare, which could lead to a Kessler event that turns the selective orbit into a roaming cloud of debris. A Kessler event could be triggered by a piece of random junk hitting one satellite, which fragments into multiple pieces of shrapnel, hitting another satellite and so on.
SpaceX has offered a broad technical vision and a hardware partnership with NVIDIA, but few of the operational metrics that would establish commercial viability. The real test will be whether SpaceX Starship becomes a practical launch vehicle and can overcome its cooling and safety issues.
In practice, the development of space-based AI data centers will likely have significant implications for the people involved in the project, from the engineers designing the satellites to the researchers relying on the data. As the project moves forward, it will be important to consider the potential consequences of such a large-scale undertaking, including the potential risks and benefits to the environment and society, and the use of AI database systems.
The success of the Starmind AI1 satellite will depend on its ability to overcome the numerous technical challenges it faces, from cooling and networking to debris and space warfare. If SpaceX can address these issues, the project could potentially pave the way for a new era of space-based computing, but for now, it remains a highly ambitious and uncertain endeavor.
It is a complex problem.
