Sam Altman fires back at Elon Musk's proposal for space-based data centers, says orbiting data centers 'ridiculous' for now — cites high failure rates and cost as primary limiters
The world is not there yet.
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
You are now subscribed
Your newsletter sign-up was successful
Numerous visionaries, including Elon Musk and Jeff Bezos, are thinking about putting AI data centers into orbit to tap into unlimited amounts of power, fewer physical constraints, and the lack of regulations. But while the idea may be worthy in the long-term future, it is "ridiculous" for now, believes Sam Altman, chief executive and co-founder of OpenAI.
"I honestly think the idea with the current landscape of putting data centers in space is ridiculous," said Sam Altman at a press conference hosted by The Indian Express. "It will make sense someday, but if you just do the very rough math of launch costs relative to the cost of power we can do on Earth, to say nothing of how you are going to fix a broken GPU in space, and they do break a lot still, unfortunately. We are not there yet."
Indeed, it costs $5.6 million to launch 1,764 pound (800kg) into low Earth orbit (LEO) using a SpaceX rocket, though for those who plan to launch tens of tons, the price per kilogram will probably come down. Still, one of Nvidia’s NVL72 GB200 rack-scale solutions weighs from 3,000 to 3,245 pounds (1,360 to 1,472kg), depending on the exact configuration, without data center-scale connectivity, cooling, and power infrastructure. Even with discounts, launching data centers into space is still extremely expensive today, so it is unclear whether it can make economic sense any time soon.
Article continues below"There will come a time — space is great for a lot of things," Altman added. "Orbital data centers are not something that is going to matter at scale this decade."
Radiation-hardened components required
While launching hardware into space is expensive, there must first be hardware to launch. Leading-edge process technologies — such as TSMC’s N4 (4nm-class) — used to build leading-edge AI accelerators like Nvidia’s B200/B300, advanced CPUs, sophisticated DPUs, and network processors are not radiation-hardened, meaning that they cannot survive in space. Yet radiation-hardened fabrication technologies tend to be rather outdated (think 90nm), so before space-worthy computational hardware emerges, new process technologies must be developed.
In addition to space-worthy microelectronics, the industry must also develop space-worthy cooling systems and power generation technologies that are capable of powering millions of AI accelerators. Companies like Elon Musk’s SpaceX and Jeff Bezos’s Blue Origin are probably closer to developing such infrastructure than traditional companies specializing in terrestrial data centers, which is perhaps why Musk and Bezos are so vocal about orbital data centers today, even if they are not going to become viable for at least a decade from now.
Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.

Anton Shilov is a contributing writer at Tom’s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.
-
Arkitekt78 Every time one of these AI sycophants start talking about the technology as if it isnt the thing that is supposed to be making them money, they sound RIDICULOUS.Reply
And Altman is the absolute worst at this. I dont know how anybody takes anything this guy says seriously. -
Notton Reply
No, it's just common sense if you'd use your brain.Arkitekt78 said:Cause Musk bad, right?
GPUs in these data centers break frequently and need to be replaced constantly. This is a well known fact.
How are you going to replace a broken GPU orbiting earth?
Sending a repair crew into space isn't exactly a cheap endeavor. -
Sam Hobbs This is where both success and failure are potentially huge. As in, it never rains in California but it pours. Anyone with the vision to make it successful can be very successful. Any fool that thinks it can work and is wrong will loose big time.Reply
One problem only slightly mentioned in the article is power. In addition to the GPUs and related hardware there must be a way to get the power (electricity). Whoever is able to use raw materials already in space to manufacture solar panels or some other power source will potentially be excessively more successful. -
ngrok2b Sam Altman fears that Elon Musk will run josephson junctions and quantum computers in the cold vacuum of space efficiently and literally to dominate AI and all computing. I would be worried too if I were Sam Altman.Reply -
gschoen Reply
It's not good or bad, he has a long term habit of saying outrageous, ridiculous predictions that never come to pass.Arkitekt78 said:Cause Musk bad, right?
In this particular case, besides the still massive costs of launch and inability to service, AI GPUs make a ton of heat.
Physics of heat transfer - radiation, conduction, convection. In the vacuum of space only radiation is available. The ISS has a complicated liquid ammonia cooling system to transfer heat from the solar panels, equipment, and crew environment to large radiators. A high power orbital data center would need elaborate systems to keep the GPUs cool. -
Eximo Yep, the launch mass of cooling plus the solar panels really eats into the budget of launch mass for compute. I honestly think it is to fool people into investing. Space is cold, computers are hot, perfect!Reply
Moon based might be a better idea for quantum if it also wasn't cost prohibitive and long term reliability a complete unknown. Could put a quantum computer on the 'dark' side and have minimal radio interference from Earth. And put the compute in a deep crater so that it never gets sunlight/radiation from the biggest nearby sources. -
alan.campbell99 Keep with increasingly clogging up orbit with satellites that mess up radio and optical telescope work and it might very well lead to installations on the dark side of the moon, assuming any states would be willing to fund such an enterprise.Reply