Astronauts are bringing computers to space—and IT troubles, too
The Artemis II mission now underway aims to send three Americans and one Canadian back to and around the moon, a journey that will mark humans’ farthest trek into space in decades. The program is one of a series of endeavors, including next-generation space stations, a lunar habitat, and even a manned Mars mission, that will vastly expand human presence in outer space.
Critically, these missions are all certain to involve spacey versions of consumer technologies, which have now become a fixture of life in space. (Astronauts aboard the International Space Station use laptops and smartphones.)
Yet such modern conveniences also create a host of IT issues. On the latest Artemis II mission, astronauts—again, also with smartphones—had to reach out to mission controllers experiencing issues on their systems. The culprit: Microsoft Outlook.
Indeed, making computers work in space can be difficult. This includes issues related to the microgravity environment, as well as challenges related to communication networks, since signals back to Earth can be limited—and slow. There is, after all, no live tech support when you’re on the moon, or even Mars.
“When we’re colonizing something like the moon or eventually going to Mars, I think it brings in lots of complexities,” Manoj Leelanivas, who serves as president of HP Solutions, which frequently works with NASA on in-space hardware, explains. “In the computer environment, typically you assume liquid cooling or air cooling. You look at the standard principles of convection, which don’t really apply in space. You have to deal with the radiation.”
Fast Company chatted with Leelanivas about what it takes to build computers for space, and how that work might change as manned missions take on even more ambitious efforts.
This interview has been edited for length and clarity.
How is designing a workstation for space, or microgravity, different from designing a workstation for Earth?
Let me start with, first of all, addressing unreliability and connectivity. In space, connectivity is not super reliable, which means compute has to really be close to data in supporting real-time decisions. You’ve got to add the compute horsepower and the GPU horsepower to make AI-based decisions locally, even without connectivity. That’s one of the capabilities that our flow workstations provide. They have huge elements of GPU capabilities, and you can even run 200-billion-parameter models. Some of them can even go higher than that.
So that allows you to work in a non-connectivity environment. That’s important for flagging anomalies—quicker and faster—without waiting for the delay all the way to come back to Earth and whatnot. It’s not just about space, but down in the control room and beyond, you need to have data gathered in real time and have quick decisions made.
What about computers on the moon?
One of the biggest challenges the moon poses to us is that you can’t just quickly go and solve a problem. . . . You have to really plan, which means that you have more self-reliant systems—meaning systems that can break and recover. We think about: How can you actually process things as locally as possible? That’s slightly different from the Earth world, where we assume a lot of things in terms of redundancy and capabilities.
How does HP see its relationship with private space companies?
We are engaged with most of these companies without actually going into a specific name, because you’re not allowed to do that. Our workstations are the No. 1 workstations in these activities. The deep compute that is needed, and the modeling that is needed, and the history of looking at large amounts of telemetry data—those all put HP in a great position to work with many of these companies. . . . The only thing we are publicly talking about is this NASA relationship we have.
What about orbital data centers?
It’s always a good idea to think of new ways to solve problems in a business. . . . Energy is a big requirement for data centers. . . . If you go to space with the independent power of the sun, that can be harnessed. The energy may not be that difficult a problem in terms of energy-generation. But then it creates other problems, like: How do you cool your systems in an environment where convection doesn’t really work? It is definitely an aspirational idea, but at the same time, it’s a very meaningful one, because if we can solve data centers in space, it’ll solve a huge energy crisis, which is going to come in our future on Earth.
How do you design AI for space?
AI actually creates the opportunity for self-sustaining systems in space. . . . Machines are able to perform a lot of functions autonomously. We sometimes are injecting the human in the middle to make sure that certain things are done the right way, ethically. Sometimes, too much human intervention actually can slow down the AI process a little bit. . . . Sometimes, a high-powered, autonomous system within the guardrails is the right way to go. . . . You don’t have time for the back and forth.
I do believe the work we’re doing, and the work a lot of other companies are doing, is enabling speed for the agents, with the models below [them]. . . . We are not only building great hardware and compute systems. We are also building technologies to bring them together quickly.