Latest / Technology Now / How do you put a supercomputer on the moon?

Transcript

How do you put a supercomputer on the moon? · Technology Now

• MICHAEL BIRD Good morning Sam. Good morning. Hello from the future. SAM JARRELL Good morning Michael. MICHAEL BIRD No, I mean, not really the future. I'm just. I'm just eight hours ahead of you anyway. I got a quick. I got a quick question for you. Um, we've discussed it before on this podcast. I'm pretty sure I asked Aubrey this question, but I want to get your answer now. You're officially a member of the technology in our family. SAM JARRELL Okay. What's your question? MICHAEL BIRD So, Sam, if you could send anything to space, what would it be? It doesn't need to be anything like the good of humanity or anything like that. Just something, you know, something cool that you could send to space. SAM JARRELL Well, I think if I'm honest, the thing that I would send to space already has been sent probably by someone way smarter than me. Um, have you ever heard of this thing called the Lunar Library? It's like, basically this giant backup of the Earth's knowledge. And it all exists with, like, archives on the moon and orbiting our planet. As a writer, comms person. You know, I like the idea of not losing everything that we've worked on as a civilization man. MICHAEL BIRD You can tell you what, for a tech company, because, like, what you're basically talking about is like the ultimate disaster recovery. Like the ultimate backup. So well done, sir. That's a good answer. That's a great answer. SAM JARRELL So, Michael. Then what are we sending to Space on this week's episode? MICHAEL BIRD Well, a couple of weeks ago, we explored the spaceborne computer program and what we've achieved so far. But this week, for our final episode of 2025, we are looking to the future and exploring how we could quite literally, I kid you not, put supercomputers on the moon. How cool is that SAM JARRELL That is very cool. Well then, should we get on with the show? MICHAEL BIRD Yep. Let's do it. I'm Michael bird. SAM JARRELL I'm Sam Jarrell MICHAEL BIRD And welcome to Technology Now from HPE. MICHAEL BIRD Super computers in space. Once upon a time this was the stuff of science fiction films but like so many things that started off on film and TV it has become a reality. SAM JARRELL (leading question) Ok. I get that, but generally when we discuss supercomputers on this podcast, they’re huge. I mean, we’re talking the size of entire rooms so I’m assuming for a supercomputer to make it to space, it’s got to be a little smaller right? MICHAEL BIRD Yeah. Well, just really, really big rockets. Sam. No no no it is actually it is actually smaller. But that being said, even though it's small, it is still pretty powerful. Now, according to the HP newsroom, the first spaceborne computer was able to achieve 1 trillion calculations per second back in 2017. So while it may not be on the same scale as some of the other computers we’ve discussed previously, it’s still pretty powerful, and now we’ve put one in space, we’re sending a computer to the moon which we are going to be talking about a little bit later with the Senior Director, HPE Global Technical Marketing, Space Technologies & Solutions · Hewlett Packard Enterprise, Norm Follett. SAM JARRELL But we’re going to have to wait just a few minutes to get to that conversation because first, I thought we should look back at the computers which helped man reach the moon in the first place, which means, and I’m excited because this is my first time saying this. It must be time for… Technology Then. SAM JARRELL We’ve all heard of the Apollo missions right? Well, have you heard of the Apollo Guidance Computer – or AGC? MICHAEL BIRD Response Oh, is this the one that's sort of woven like it's sort of woven memory, like knitted memory or something like that? Oh, it's something like that. . SAM JARRELL Something like that. Each Apollo mission used three AGCs which went into both the command module, and also the Lunar Excursion Module – which most people just call the lunar lander. These computers were pretty solid – weighing in at 70 pounds (that’s just under 32 kilograms for you, Bird) and measured 2 by 1 by half a foot (that’s 61 by 31 by 15cm). It’s about the same size as your old tower PC would have been back in the day. Ok so I hope you’ve got an image in your head because I’ve got a question for you, Michael. How much RAM do you think the AGC had? MICHAEL BIRD Are we talking kilobytes or are we talking bytes? I'm going to say one kilobyte of memory. This is what I'm going to go with. SAM JARRELL Not a terrible guess. It only had four kilobytes. It only had 4 kilobytes – to put that in perspective, laptops today will often have more than 16 gigabytes of RAM. This was a computer people were trusting with their lives and was good enough to land in a precise location. According to a 2022 paper by Charles Averill, analysing the Apollo Guidance Computer, the AGC could perform 14 thousand two hundred and forty-five calculations per second – or 14 kiloFLOP s. This is positively primitive when you compare it to today’s supercomputers which perform in the exaflop range. That’s 15 orders of magnitudes bigger in today’s computers than the ones which put man on the moon! MICHAEL BIRD [[reaction to Tech Then]] The speed of progress is really quite amazing and so to find out more about the next steps of the spaceborne programme – a lunar programme - I spoke with Norm Follett, Senior Director, HPE Global Technical Marketing, Space Technologies & Solutions at Hewlett Packard Enterprise. MICHAEL BIRD Norm. Thank you. So much for joining us on Technology Now again. NORM FOLLETT I'm a repeat offender. MICHAEL BIRD I know you are, you are, you are getting dangerously close to a friend of the show, if you keep coming on the show… right now, norm, we're standing in front of what looks like, uh, some sort of rover, this incredible thing here with this lovely Gold Solar panel, what we're looking at here? NORM FOLLETT That is a solar panel. This is, this is, well this is the Flip rover and that acronym stands for the Flex Lunar Innovation Platform. But this is really kind of the size of a, uh, a bit of a golf cart, if you will. Uh, with four very large wheels that kind of raise up to a waste level with an incredibly unique design, uh, that is really meant to allow it to navigate and roll over some, uh. Unforeseen and unknown types of terrain and in a very, uh, adaptable way. It has a, a, pretty much a flat chassis. And on top of that chassis is a solar panel, which it can rotate in a variety of directions in order to absorb the ever necessary radiation in order to actually power, uh, this, this, uh, assembly. MICHAEL BIRD stunning. NORM FOLLETT Absolutely. It is very nice. It is very nice. Uh, and it's, you know, there's a little bit of bling associated with this particular one. Uh, this is the, uh, an exact replica of the one that will be flying on the Griffin one mission in July of 2026. MICHAEL BIRD so this is part of the Space Born computer program? NORM FOLLETT Yeah, it's a progression of the Space Born program. And just to kind of recap where we are on that. Uh, Hewlett Packard Enterprise has the most powerful computer to ever go to space. Uh, we're flying our third iteration currently right now in the International Space Station. Uh, it is flying on the Columbus module. Our first one went up in 2017. And again, we've had a couple of different versions and we've evolved that version. And what we're really delivering is a scientific platform, a high performance compute, edge compute platform. Uh, and it's a scientific platform that we're allowing the international scientific. Community to utilize. We have nothing to do with the avionics on the space station. Again, it is a pure science platform providing, you know, computational services, uh, at the edge. MICHAEL BIRD So we're gonna have a computer on this. NORM FOLLETT we're actually, we are gonna have a computer. In fact, what we've done here is we've designed a space borne blade. So this is a, this is about 10 inches long by about four inches wide. Uh, it has a really lovely chassis. Uh, now we might think that that's kind of bling, but in fact what it really is, is this chassis is the cooling system for this particular blade. So a couple of different microprocessors in there. Uh, we also have, uh, one terabyte of storage provided by our partners Kochia memory. Uh, and this chassis actually radiates the heat or lets the heat. Dispensed through the chassis. This blade is, it will be inside the flip rover connected to the radiation plate, and that's how it distributes the in space. And it's powered and connected to this solar panel, uh, in the power system. That's all part of the flip rover. And when the flip rover goes operational, uh, we're gonna be performing the first, uh, edge computing and AI experimentation on the surface of the moon. MICHAEL BIRD Wow. And so this. Is the, this is the flip over NORM FOLLETT This is the Flip Rover. And as a prototype, it has, again, you know, some, some overstated, uh, wheels, if you will. But these are actually the wheels of the size and the same solar panel configuration that will be on the next version of the Rover, which is the Flex Rover. And the Flex Rover will be deployed on the surface of the moon as part of our return to the moon and the Artemis program. MICHAEL BIRD Can you just sort of show us whereabouts on the rover that the, um, this, this space form computer NORM FOLLETT Well, it'll be, it'll be, it'll be internal. It'll be internal and right about here MICHAEL BIRD Okay NORM FOLLETT Uh, it is, again, a scientific platform to us. Do experimentation on there, and we're gonna do a couple of different things. One is actually gonna be a lidar experiment, which people might be familiar with. MICHAEL BIRD So what, what are the sort of specs on the, the space born lunar? NORM FOLLETT Well, one, one terabyte of memory is probably the most significant one, you know, and, um, uh, a fairly performance CPU and processor. Uh, and we're also gonna have, uh, a small GPU in there as well. So, um, it, it's a hearty little beast. It looks like a, a, a gold brick, you know? MICHAEL BIRD I described it as a gold graphics card. NORM FOLLETT A gold graphics card? Well, there's actually, it, it is about the size of a graphics card. It's really, um. It's really, there's actually two different, uh, two different cards in there, which we actually separated and we actually built and rewired those cards. Yeah. Uh, in fact to support the kind of configuration, both a high-end processing, uh, memory chip that we put in there. And also to distribute the heat internally and radiate the heat out. So some mo some graphics cards are stacked on top of each other. Again, kind of a wafer configuration. This is actually separated and split with multiple layers internally, none of which you see. MICHAEL BIRD How does the design of this differ from, uh, a normal computer? NORM FOLLETT So first of all, this, this has to operate in the vacuum of space, MICHAEL BIRD Right. Okay. So that will bring some challenges NORM FOLLETT that will bring some challenge. I'll go back to the power and cooling comment earlier. So that's kind of the first challenge is to make sure you're doing that. And then also, you know, the connectivity is, you know, a dare I say, it's traditional ethernet of how we're actually. Into the networking of the rover and the communication system of the rover. Now, one of the challenges that we have again, is, uh, you know, in this environment is connectivity. Uh, we have a very limited bandwidth of communication. We are dependent upon the flip communication system, but we only have small slices, time slices, to be able to do that. So a lot of our experimentation, the, I say, is gonna go back to the notion of batch processing. You know, execute autonomously, save the results. Then when we have our command window for communication, you know, send the results back to earth. Okay. MICHAEL BIRD You talked about the vacuum space, but can you sort of go into the particulars of why, why these sort of changes need to be made? Like what, what in particular about the vacuum space means that, you know, I couldn't just put a, the laptop in my bag and take it up with me and. NORM FOLLETT Lemme back up a little bit. So, traditionally computers are operating in an atmosphere, right? So an atmosphere has pressure. Uh, you have the ability to radiate heat out your PC at home. You got a fan in it, it's air cooled. It's dissipating the heat, uh, by blowing that heat away from it up into the atmosphere. Uh, of course with large compute systems. Then that heat becomes more pronounced, if I described earlier about how the, the, the chassis itself is designed to heat up and then dissipate the heat and radiate the heat through the radiation plate within the flip rover. So that's one challenge about what we need to do there. And then of course, you know, of course the power. I mean, this is completely solar powered in every sense of the way. Uh, and that's a, MICHAEL BIRD there's no, no batteries in it. it relies solely on this, it relies… NORM FOLLETT No batteries. No. Solar MICHAEL BIRD solely on solar. NORM FOLLETT Well, I mean, there are batteries on flip, you know, and the, and the, the solar power is gonna charge those batteries to the extent that it goes. But flip is a, a really unique little beast. Um, just a, a couple of comments on, on that. Um, and well, going back to your notion about, about what, what else in space. So first of all, flip, as you see it here, is actually going to be exposed to space. Right for the whole duration. So the mission itself, it's a Falcon Nine heavy, uh, flip writes on top of a lunar lander. So the rover writes on top of that lander through the vacuum of space traveling to the moon. Therefore, our compute system is actually, will be exposed and traveling through the vacuum of space between earth and the moon. MICHAEL BIRD Does this have any developments sort of made from what you've learned from the, the space ball one and space born two? I guess in terms of maybe the sort of hardening element of it? Yeah NORM FOLLETT Yeah. So, um, you should think of everything on the International Space Station as being, uh, a test for deeper space travel. Right. You know, we have concepts, you know, across the spectrum, this across the scientific community, uh, concepts about what we think you'll need to know. Do. Uh, and what you'll need to be able to travel back to the moon or, or, you know, deeper, uh, you know, along the lines of Mars. And so those concepts and notions, uh, including our software is tested in that environment in preparation for going further into space. MICHAEL BIRD does this communicate back to back to back to earth? I mean, it doesn't look like it's got an ethernet airport on NORM FOLLETT No, it actually really does. I mean that, that is really those ports and that, but it, but it, it's wired into, uh, the subsystems of the flip rover and it's dependent upon the communication apparatus on the top of the solar panel. Right for going back through the different cycles. There's actually, there's gonna be a couple of different, uh, satellite dishes on a global perspective. And, and when this mission goes live in July, uh, I, it's our intention to stream so people will be able to experience the mission for themselves and see what's really transpiring and, and see that communication back and forth to the rover. MICHAEL BIRD so this connects into the, the rover using, you know, standard ethernet, maybe a i an IP addressed switch. NORM FOLLETT Actually, correct. It's very traditional protocols and communications. It's really based a little bit on the, the, the TRDS system, which is, uh, which is how, uh, satellites communicate Currently, uh, the TRDS system is the satellite array that's orbiting the earth in that communication. MICHAEL BIRD So, um, what are the biggest challenges of, of putting a, a computer on the moon versus the challenges of putting a computer on the International Space Station? NORM FOLLETT Well, I mean, in this case, we had to adapt to the size and what we're doing for the mission, right? And so we had to really design a whole new compute, uh, apparatus, what we've done with space born on the International Space Station. Is take, uh, uh, off the shelf compute systems and kind of adapt the configuration. But we're really starting with, you know, products that people can purchase today, uh, and put in their data centers. Uh, and, but we had to kind of evolve the container, uh, for it to be operational on the space station. In this case, we had to really purpose design. That challenge is gonna change a little bit as the rovers get bigger. So in this case, this guy is about the size of a golf cart, but the next version scales up. And again, the size of a, you know, a small utility truck and a utility vehicle, which is really what it is, it's a utility truck on the surface of the moon that's gonna be key to building the infrastructure that allows us to have a continuous human presence, uh, on the moon. MICHAEL BIRD So what, what is the point of, of this? Like, why couldn't you just put this level of compute on the surf and just on on, on earth and just have a, you know, have a radio connection, sending information back and forth? Like, why do we, why do you need to have compute on the moon or even on the International Space Station? NORM FOLLETT we need to have compute on the moon? Well, um, it's kind of the classic edge use case, right? And, you know, an edge use case, it's do the computational work. At the point of data acquisition, right? Send back answers in kilobytes rather than data in gigabytes. Data transport is expensive in terrestrial context. If the, the TRDS system, which I mentioned the expense to move data from a satellite perspective is 10 times that it's the estimates, they vary, but it's gonna be a hundred times more expensive to move data. Back and forth from earth to the moon. So when you're, when you're, when you're doing experimentation and you're collecting mineral samples and data, you don't want to be sending all of that information, uh, raw back to planet Earth at that extreme cost. Uh, not only is it costly, but the, the window of opportunity for communication is delayed. So what we're, what we're offering is the opportunity to do that at a fraction of the cost, the analysis, uh, and then send it back. You know, dare, I dare I say almost instantaneously, but you know, quite a bit quicker because you're moving kilobytes of data rather than gigabytes of data. Uh, when we first go back to the moon, the number one export off the surface of the moon. Is going to be data as we learn, as we search for water, as we analyze and as we plan the continuous human presence on the surface. But the number one expense outside of rocketry is gonna be the communication of that data. MICHAEL BIRD So rather than sending photos of the surface of the moon back to earth, every single photo, every single photo, we are gonna use that to say, analyze load of photos or some lidar information or some set. Information and basically figure out either, you know, yes there's water, or no there's not water, or, yes, that's that sort of rock, or no, it's not that sort of rock. So you're sending sort of the results rather than the raw data. NORM FOLLETT Yeah, I mean, that's a good example. It's like, and, and really what it's gonna be doing is that this, the, the flip over has a variety of cameras, variety of angles, a lot of analysis going on, but it's gonna be taking pictures continuously, boom, boom, boom, boom. And what it's gonna send back, and what we're gonna do from a computational perspective is we'll identify a delta in that picture. Oh, I took this picture, then I took this picture. There's something different between these two pictures. What is that? MICHAEL BIRD So, as far as you're aware, when this lands on the moon, will it be the fastest, NORM FOLLETT It will be the fastest computer at the edge MICHAEL BIRD Yeah. Wow NORM FOLLETT You know, and, uh, and it'll have, you know, the longest, uh, connection, so to speak. You know, so it's, it's, it's pretty exciting. And again, I mentioned, you know, I mentioned, um. The federated learning, uh, let me talk about that for just a moment. So we did this on the space station, uh, and, uh, a federated learning. It's a, it's an AI concept, a machine learning type of concept. And the notion is that, let's say you have a triangle of, of three points talking to each other. Each point in the triangle has a large data set rather than. That triangle, exchanging those large data sets. You, you have a model and you're sharing information about that data with each other. So if you are one of the, you're my partner in that triangle, you are benefiting from my data and my knowledge. But what we're gonna do on the moon is we're gonna expand that into a full triangle and we're gonna have the space station talking to the moon. Talking to our lab and we're gonna be exchanging information and building that model and prototyping that out. As the rover moves and as our technologies move across unknown areas, we will continuously benefit from all partners in that federation and the information that we're gonna be sharing. So that's one thing that we're gonna proof again and make sure that works. Again, it's something we've experimented with, learned with, implemented in orbit. We'll apply to the moon as well. MICHAEL BIRD What are you actually hoping to discover on the moon with, with this, um, you know, eventually. NORM FOLLETT I mean, it's a race, you know, I mean, we're back in a, the space race is real again, the reality is the moon will be a stepping stone, uh, to further exploration into the solar system, specifically Mars. That is the plan, right? It is a way station, if you will, on that journey. Uh, there'll come a time. When, you know, spacecraft for going deeper into space will be manufactured on the moon. I mean, why build it on earth and try to lift it out of earth's gravity when you can actually build it in a, a relatively low gravity environment? 'cause that's where it's gonna be operating anyways, that changes all the costs and all the dynamic. So what do we intend to discover in our lifetimes is, uh, you know, the, the, the tools necessary to have a sustainable. Uh, presence. The Flip Rover is only designed to live and operate for one lunar day. Now one lunar day is 14 earth days. Now we're hopeful that it's gonna find a place to hibernate, and hibernation is kind of a strange word because it's really gonna try to get as much sun as it can for as long as it can. Uh, so it doesn't, you know, get too terribly cold. Uh, and hopefully it will survive a lunar night. And if it survives a lunar night, that's just gonna be, you know, an exceptional result. MICHAEL BIRD Alright, norm, thank you so much again for your time. NORM FOLLETT Oh, it's been a pleasure. SAM JARRELL Wow. That was pretty fascinating. I. I also kind of appreciated that you guys talked a little bit about, like, the business side to this entire thing that I didn't realize it would be a hundred times more expensive to move data back and forth from the Earth to the moon, rather than what we're doing right now, which probably is still quite expensive, but it just seems like this is the direction everything's going, especially if eventually we're going to try manufacturing on the moon, too. MICHAEL BIRD Yeah. One of the things that I thought was really fascinating from the interview, I think sort of to your point, like the business side of things was like the the fact that the moon will sort of be this launch pad for stuff in the future. I found that fascinating. The listeners may not know this, but we recorded this at Discover Barcelona a few weeks ago. You were you and I were both there. Did you have a chance to go and see the rover in person? SAM JARRELL I did see the rover in person. I will say one of the things that was quite interesting to me about it was how blingy it is. MICHAEL BIRD Yeah, like, yeah, it's quite bright. I guess it's the solar aspect of it. All right. Yeah. It was like. It was like a bright gold solar panel. I thought, yeah, I could I could have one of these in my house. This would look quite cool. I imagine it's quite a lot more expensive than the solar panels I have on my house. It was so cool. And it was so. It was so amazing to see it in person. Yeah, and it didn't. He say it's going to be on top of the lunar lander? So that thing is just out there and kind of exposed along with the computer itself. Right. And I'm fascinating to sort of see where it's going to go. Um, I love the fact that he said, this is going to be the fastest computer at the furthest edge or worst to that effect. It's like it's going to be the the furthest and fastest thing we've ever made. SAM JARRELL Yeah. And do the first edge computing and AI experimentation on the surface of the moon. It's a lot of firsts. Yeah. I thought it was quite interesting to, you know, it's going to be the fastest computer at the edge. It's doing all of this, you know, brand new experimentation, but it connects with Ethernet, which was so shocking to me. MICHAEL BIRD When we were at discover, I heard Ethernet more on those those two days of discover than I've heard it in the last like two years. Um, and the answer for a lot of this sort of conversation around Ethernet is Ethernet is is an open standard, but it's also like really reliable. And it's been around for years. And why wouldn't you use it? Why wouldn't you? Something that's proven to work. Yeah. At the end of the day, it is still Ethernet, even if they're like a gold plated Ethernet cables. MICHAEL BIRD Yeah, yeah. Very blingy Ethernet cables Now, you know that phrase “objects in the mirror are closer than they appear”? SAM JARRELL I do… why? MICHAEL BIRD Well, we’ve been discussing lunar supercomputing like it’s really far away, but when I asked Norm about it – his answer was a bit sooner than I think most of us would expect… MICHAEL BIRD So when is this? What is this due to be going up… NORM FOLLETT July 2026 MICHAEL BIRD July 2026? So it should be on all of our news feeds. We’ll be shouting about it NORM FOLLETT We are operating from Mission control, Hawthorne, California. And ideally, we're going to be able to stream that and share and all experience that mission together. And I, you know, think about it. This will be the first wheels in the dirt, so to say, on the moon since 1972. Right. Since 1972, and it will also be the, we believe, well, may not be flipped, but certainly flex will be the fastest rover on the moon. Set some speed records up there as well. SAM JARRELL Okay that brings us to the end of Technology Now for this week. Thank you to our guest, Norm Follett And of course, to our listeners, thank you so much for joining us if you want to explore the spaceborne computer programme further, make sure to check out the show notes for more information including links to our other epsiodes covering it. MICHAEL BIRD If you’ve enjoyed this episode, please do let us know – rate and review us wherever you listen to episodes and if you want to get in contact with us, send us an email to technology now AT hpe.com and don’t forget to subscribe so you can listen first every week. Technology Now is hosted by Sam Jarrell and myself, Michael Bird This episode was produced by Harry Lampert and Izzie Clarke with production support from Alysha Kempson-Taylor, Beckie Bird, Spencer Trinwith, Karl Norberg, Branson Tarr, Allison Gaito, Alissa Mitry and Renee Edwards. Our video editor was Leon Radschinski-Gorman and our music was, as always, composed by Greg Hooper SAM JARRELL Our social editorial team is Rebecca Wissinger, Judy-Anne Goldman and Jacqueline Green and our social media designers are Alejandra Garcia, and Ambar Maldonado. MICHAEL BIRD Technology Now is a Fresh Air Production for Hewlett Packard Enterprise. (and) and we’ll see you at the same time, in the same place, on January 4th . Cheers!