In a new report, “Space-Based Data Centers Roadmap: The Case for Caution,” the Innovation for Cool Earth Forum warns that space-based data centers are likely to become a reality within the next few years—well before the risks of large-scale deployment can be fully understood or governed.
Led by David Sandalow, inaugural fellow at Columbia’s Center on Global Energy Policy, the report urges companies and policymakers to proceed with great caution.
Some space-based data centers are already operating today, with proposals to launch more than 1 million data centers into orbit. According to the roadmap, neither technical challenges nor high costs are likely to stop initial deployment, since leading companies will be willing to absorb high costs for years to gain technical expertise and secure their market position.
One central concern is the scale of rocket launches required to meet the projected global demand for these centers. To put even 10% of the artificial intelligence data centers anticipated by 2030 into orbit would require more than 700 launches of large rockets every year, about two per day, in perpetuity.
The report also emphasizes the need to heed the lessons of accidents such as Deepwater Horizon and Fukushima. Policymakers should not allow the deployment of large constellations of space-based data centers until there is more operational experience and the risks and impacts are better understood and well managed, the authors warn.
Alexis Abramson, dean of the Columbia Climate School and a principal co-author, also contributed two chapters to the roadmap: “Cooling” and “Night Sky and Astronomy,” which she wrote with Sandalow.
In “Cooling,” Abramson notes that thermal management is a key issue with space-based data centers, which can shed their waste heat only by radiating it into space. The best system design, while seemingly counterintuitive, would be to run the AI-grade chips as hot as they can safely tolerate, since a hotter radiator sheds far more heat and keeps panels small and launch mass low, she writes.
However, the degradation of radiator coatings and the satellite’s debris-avoidance maneuvers could undermine thermal performance over a satellite’s lifetime—factors that might be not captured in initial analyses, Abramson points out.
The report urges developers to treat cooling as a critical design constraint and calls for standard thermal performance metrics and disclosures so that claims can be independently verified.
In “Night Sky and Astronomy,” Abramson and Sandalow warn that these large constellations of space-based data centers could cause a dramatic transformation of night skies around the world. For example, satellites could outnumber visible stars in many places. Observational astronomy and related research would be especially affected, they write. While technical measures could reduce these effects, there would be operational trade-offs and the need to radiate waste heat would persist. The report outlines a range of policy responses, including brightness limits, pre-launch impact assessments, required mitigation measures, caps on satellite numbers and international standards, that could help mitigate some of these effects.
According to the paper, there are other significant risks and impacts from space-based data centers, including large debris fields in low Earth orbit, loss of life and property damage from de-orbiting satellites, depletion of the stratospheric ozone layer and increased global warming. The current governance regimes will not be sufficient to manage these risks, especially as addressing many of the challenges will need to be coordinated on a global scale.
Other principal co-authors include Colin McCormick, chief innovation officer of Relae; Roger Aines, former energy program chief scientist and distinguished member of the technical staff at Lawrence Livermore National Laboratory; Kathleen Curlee, senior research analyst at the Center for Security and Emerging Technology at Georgetown University; Julio Friedmann, chief scientist at Relae; and Hari Osofsky, Myra and James Bradwell professor and former dean of the Northwestern Pritzker School of Law. Contributing co-authors: Theiija Bala, J.D. candidate at Columbia Law School; Anne-Amélie Campant, Schwarzman scholar at Tsinghua University; and Alp Kucukelbir, director of artificial intelligence and supply chain optimization at Amazon.
