Elon Musk and other AI leaders have repeatedly emphasized that the solution to the industry’s extremely expensive and energy-intensive data centers is to launch them into space, taking advantage of uninterrupted access to solar power and virtually unlimited real estate.
Late last month, SpaceX – which has now merged with xAI – filed a patent orbital data center constellation With the Federal Communications Commission. The idea is to place one million satellites in a sun-synchronous orbit circling the Earth at an altitude of between 310 and 1,200 miles to maximize the amount of solar energy captured.
The application did not elaborate on any specifics, suggesting SpaceX was only beginning to consider the idea. That’s despite Musk promise of Space-based data centers could overtake their Earth-based counterparts as the most affordable way to power AI within just three years – potentially another of his typically ultraambitious timelines.
Many experts remain highly skeptical, questioning the financial feasibility and technical limitations of running a data center in space.
And as Rebekah Reed – former Associate Director of NASA and Associate Director of Harvard University’s Program on Emerging Technology, Scientific Advancement, and Global Policy – argued in a Essay For financial TimesThis may be an even more doomed idea than we initially thought, pointing to significant environmental concerns, in addition to the obvious questions regarding cost and operation.
“Conceiving of Orbit as a workaround for AI’s current energy-hungry training requirements is, as OpenAI co-founder Sam Altman recently put it, ‘ridiculous’,” he wrote. “Orbital data centers are many years, perhaps decades, away.”
Google CEO Sundar Pichai has predicted we’re only a decade away From orbital data centers, Altman scoffed at the idea, While debating during a recent conference That we’re just “not there yet.”
Reed made the case that launching all that mass into orbit would be prohibitively expensive. To become “economically viable” the cost would need to be brought down to below $200 per kilogram, which would be “a sevenfold reduction from the current level”.
“That limit is not expected until the mid-2030s,” he wrote.
Then there is the issue of maintenance. If a chip malfunctions – or essentially becomes obsolete – there’s a solution as simple as sending out an IT technician to fix the problem.
“In orbit, that task requires either sophisticated in-space servicing or acceptance of degraded performance and stranded capital that becomes orbital debris as components age and fail,” Reed wrote.
Worse, falling satellites can deliver harmful pollutants, including metals, into the upper atmosphere, which environmental scientists are still struggling to understand.
Reed pointed to recent findings from researchers at Saarland University in Germany, who found that the carbon footprint of space data centers could exceed that of terrestrial data centers when taking into account manufacturing, launch and disposal.
“The results show that, even under optimistic assumptions, in-orbit systems have significantly higher carbon costs – up to an order of magnitude higher than terrestrial counterparts – primarily due to embodied emissions from launch and re-entry,” he says. Written in a yet to be peer-reviewed paper Last year.
Finally, a massive orbital data center constellation of thousands of satellites could further clutter Earth’s orbit, increasing “the risk of collisions and debris, threatening communications, weather and navigation services,” Reid concluded. “Scaling data centers to match terrestrial demand will accelerate congestion and spoil the night sky.”
More information on orbital data centers: Irregular Elon Musk asked employees to build a giant catapult to the moon
