AMSTERDAM — One of the big topics, especially in light of the SpaceX initial IPO documents, is the future of orbital data centers and the possibilities of moving compute to space. Experts at SmallSat Europe tried to decipher between the hype and reality of orbital data centers — in terms of both technology and economics.
Dr. Paul Struhsaker, CEO of engineering and technology consulting firm Arrasar Partners, who formerly worked as a corporate vice president at semiconductor company AMD. He described the buzz as a reaction to terrestrial challenges.
“There are a lot of constraints [to putting data centers on Earth],” Struhsaker said. “Then you have the question — Why don’t we put them all in space? Here, there is unlimited energy, for example. However, you need custom silicon to put data centers in space.”
There a number of players beginning to emerge in this space with the likes of SpaceX, Starcloud, Blue Origin’s Project Sunrise, Google with Planet Lab, in addition to Sophia Space, Axiom Space, Aetherflux, and others. While Struhsaker said he understands the logic behind these moves, there are a number of issues that will need to be addressed. He spoke of the level of compute that players want to put in orbit will be “orders of magnitude more than the constellations we have now.”
In addition, the way processors and chips go through iteration will create a new set of problems. “An AI data center blade lifespan is five years. AI has made this even worse. Instead of a five-year span, the next generation of processors come out every two years … we’re now shortening the lives of the hardware in the data center. That creates a lifespan issue for satellites in orbit.”
Compute lifespan and replacement is a significant logistics issue, which will also bring up issues around decommissioning satellites, he said.
Struhsaker spoke of an “AI ASIC problem” – and this could impact the likes of chipmakers like Nvidia and AMD. “They burn energy like nobody’s business. Every generation of chip doubles in energy,” said Struhsaker. Struhsaker sees companies like SpaceX and Amazon with an advantage, as they can leverage their full stack capabilities — including launch — to deliver their vision of space-based data centers.
Overall, Struhsaker sees opportunities for more chip companies to gain a foothold in the space industry, for edge computing to bring opportunities to the Earth Observation sector, and for orbital data centers to lead to full stack supply chain growth in areas such as optical terminals and thrusters, for example.
Dr. Oguz Karasu, postdoctoral research fellow at the University of Oxford, pointed to launch availability as a key issue to realizing orbital data centers. He thinks SpaceX’s Starship launcher will be necessary to accommodate the size of orbital data center spacecraft, when taking into account design constraints like power and cooling in orbit.
“We are all are waiting for Starship’s success — but without that, data centers on orbit won’t be feasible either,” he said. “There is an economic danger.”
Karasu also pointed to data security concerns, saying that people who build data centers on Earth have raised this question to him, of how to protect a data center in orbit from being targeted.
Karasu cautioned against the industry overly investing in this concept at the expense of other areas of space. He believes there’s been “misleading” information about the concept.
“Investing into this idea will limit our capabilities of investing in other, real problems of space industry — the cost efficiency is a major problem here,” Karasu said. “We have other current real problems on Earth that we need to solve via space constellations.”
