How to install a quantum computer

Welcome to season two of the Coherence Times, where we bring coherence to the entangled world of quantum computing. I'm your host, Ryan Mandelbaum. Every other week, I'm bringing you stories about how scientists, developers, and businesses are making quantum computing a reality. We'll explore the latest research and development, highlight the exciting advances in the field, and see how users are trying to extend quantum to real-world use cases. >> [music] >> One thing that amazes me about quantum computers is the fact that the chips that we used to all do all of this mind-bending mathematics, well, they kind of just look like computer chips. You know, they're around the same size, and they're fabricated using similar semiconductor processes, just like regular computer chips are. Uh but that doesn't mean that installing quantum computers is easy. Uh they still require cooling and shielding and control electronics, and it takes a lot of time and effort to do all that. >> [music] >> Um that makes it even cooler, haha, that folks at RPI and IBM were able to install a quantum computer in an old chapel. So, today I'll be sitting down with two people who can talk about what it's like to install a quantum computer, and how they think about quantum integrated into superconducting hardware overall. First, we have John Kolb, vice president and head of the Future of Computing Institute at Rensselaer Polytechnic Institute, who has a long history installing quantum at RPI's Computational Center for Innovations, and led the installation of RPI's Quantum System One. Also joining me is Rajiv Malik, program director for deployment and systems engineering at IBM Quantum, who's in charge of system deployments on the IBM side. Uh let's just get started. So, John, I just wanted to ask, why were you excited to install a quantum computer in the first place? >> Well, maybe uh twofold. Uh one is uh it pretty it's pretty clear to me that this is the wave of the future. Uh no pun intended on that either, but um Uh, quantum is going to be incredibly important uh, over the next few decades and century. And getting started early in that and exposing our students and community to that is is very important. I mean, I think workforce and we'll come back to that. But developing the workforce of uh uh, for this century's computing, I think is incredibly important. And RPI, as you may know, just celebrated last year, a couple years ago actually, our 200th anniversary. So, we've always been at the forefront of technology and how do you bring technology to bear? Having said that, on a personal level, it's just great to have the challenge of doing something that nobody has done before, uh, at least in higher ed and outside of some of the IBM sites. Uh, trying to figure out how to do that, trying to figure out how to do that in a 100-year-old chapel uh, that was once a chapel for a seminary. Uh, and all the intricacies of the old and the new and the uh, the theology and technology pieces was really a lot of fun. And working with Rajiv's team and our team, uh, we were able to put together a world-class group of problem-solvers and that's just a blast watching world-class people interact with each other. >> Totally. So, why a chapel? I'd love to learn more about the location. >> Uh, why not? No, just kidding. Uh, um, it turns out that uh to the south of our main campus in the 19 uh, '50s was a Catholic seminary. And that seminary was actually sold to RPI, the property and the buildings, uh, in the 1950s. And we tore down the residence hall uh, building and we're about to actually tear down the chapel until uh, our 14th president, George Low, um, uh, got the idea of restoring it and turning it into our computing center. So, it's been a computing center ever since the late 1970s. But, uh putting a quantum system in there was yet another challenge for that building. It wasn't quite ready for a quantum system compared to if you will run-of-the-mill servers and uh switches and so on. And there's still some stained glass windows in the building, so it's just a wonderful setting for kind of the latest technology in quantum computing uh just opposed with uh a 100-year-old building with stained glass windows. >> Yeah, I mean Rajiv, what did you uh think when you first heard about this installation that you were going to uh help them install a quantum computer in a building like this? >> The speed at which this all happened was unprecedented in terms of when um uh you know, the senior leaders started to talk and saying we need to do this to the day when it actually happened. As as John mentioned, the 200th anniversary was a goal, and so the end date wasn't moving. >> [laughter] >> We had to start when we had to start, but uh so we were suddenly worried about the time it would take. Um suddenly the location as we found out more about it as we visited and found out more about it, but exactly as John said, it was I think there was confidence that the commitment that the RPI team was already showing and when we met the team for the first time and the excitement that our own team had that we were going to figure out a way to make it happen. >> So, I mean what does a building in a building like this uh what challenges does it bring? I mean, is the the floors or the walls, the vibrations, you know, what's it like? >> Yeah, no, that's that's a good question, right? So, uh uh I I would say when we did the started the RPI related installation we were still learning in terms of what specifications are needed. We had a lot of them laid out, but we already knew, for example, we already knew exactly, like you mentioned, the floor has to be vibration-free. The There is cryogenics involved in a quantum system. The processors themselves are inside a fridge that's at pretty low temperatures, and so they can't take too much vibration. Similar to some electron microscopes or so on. We knew it had to be able to take a certain amount of weight. Many of the equipment, you know, it needed a certain amount of power, it needed cooling water. So, we had a set of specifications, and really the goal was what we had done to date so far of our quantum system one installations, we knew that had to be modified a little bit to fit the environment that RPI had. And so, we knew there were some changes to be made. For example, one of them was uh what we called a mechanical room couldn't be next to it, but had to be underneath it. Uh so, that was one. So, it was really looking at And at that point, we were in the mode saying, "We'll go survey it, we'll figure out what has to be changed, we'll evaluate and we'll figure out a solution." But it was We knew there were going to be challenges, uh but we also knew there was something, you know, the team was up to figuring out. >> So, John, I mean, Rajeev has to figure out how to build a quantum computer in a church, and then you have to figure out how to build a quantum computer in a church. So, how does your team sort of kind of step in and get that expertise kind of from the other way? >> Yeah, it's interesting and spot-on with Rajeev. Um uh We We had this 100-year-old building. It was built by the Catholic Church. We had no uh blueprints, no plans for the building. We didn't know about floor loading, we didn't know about damping, we didn't know any of those things uh from that you might get off of a normal set of blueprints. Uh so we had to figure that out as we went. We signed a contract the end of June in 2023. That summer, uh we understood temperature as uh as Rajiv said, the cryostat is self-contained, so we understood how to get to the temperature. Um but we weren't sure about vibration, and vibration can certainly disrupt the the uh superpositioning and so on. Um so we actually put some um monitoring equipment in the space we wanted to use. And of all things, we had our mission stores going up and down the hallways outside of this to create some uh traffic and some vibration. It turns out the vibration that building was uh that or the vibration cancellation might be a better way to say it was excellent. That was great. That was really great news. But then getting back to the weight issue that Rajiv mentioned, the the the piece as he said it's a vertical uh install, the piece that's on the main floor above the piece that's on the bottom floor is about 5 tons worth of equipment. And we sort of wanted it to stay on the main floor and not go down to the bottom floor by itself. So we uh we had to actually create a room within a room in our machine room. We have an active computer machine room downstairs, so we actually walled off the the room in plastic. Had its separate ventilation system. We brought in jackhammers, actually took the false floor out, jackhammered down 2 ft, and put in what's called a mat slab. Uh and that rests on the ground that goes right up to through four pillars right up to the mat slab that's on the main level where the the system sits on top. We built all that from scratch in an active working machine room. We're literally about 9 in away from the main switch bringing internet service into the uh campus. Uh so this this was a little bit of a trick. We couldn't get the concrete trucks uh close enough to the building, so we had to use motorized wheelbarrows to bring in the concrete. And it was about 35 tons of concrete by the time we were done with all this. But by hand, I've I've joked that I wouldn't want to arm wrestle any of those guys that brought that concrete in because they worked hard. Um the other the other joke was when we were uh jackhammering out the concrete to get ready for all this. Uh the one thought I had was please don't find any bones down there cuz the last thing is I don't want to make this an archaeological dig. So, uh um but all that worked well and then uh and then we started building on top of that and you know, Regi's team led by Jim Spidel and Mike Burns did a great job. Uh they blended well with our project manager Jeff Minor. Uh and they they sort of found things together as a team that uh like Jim and Mike were very interested in the fact that we have an experimental machine shop that was two buildings away from the chapel. So, when a piece came in that wasn't drilled right or needed to be milled a different way, they you know, the team put the piece of aluminum or whatever on the shoulders, went over to our experimental machine shop, zip zip, and they were back in and gear. They didn't have to send it out and wait a week for a part to come back. Um that helped a lot. And we can probably come back to this, but uh uh the one probably place where I didn't think about it as closely as I should have up front was the 10-ft by 10-ft piece of glass uh on the uh the enclosure for the uh Quantum System One. And I thought that that would fit right through some door in the Voorhees Computing Center in this chapel. It turns out there's no opening that would accommodate that. So, that's that was a story unto itself that we had five 10-ft by 10-ft pieces of glass that had to get into the building uh to build a vitrine and uh uh we ended up uh finding one way of doing it, which was about 30 ft up. We took some windows out and we had about uh uh 5 in of spare. It was uh 120 in, 10 ft, plus 4 in of packing top and bottom, 128 in, and we had 133 in by the time we were done taking some of the windows out. >> So, um I think we have to take a step back and ask, you know, So, now we know how to build a quantum computer in a chapel, but how do we build a quantum computer at all? >> Oh, good question. A good question, Ryan. So, so look for a sup- for a quantum computer with superconducting qubits, it's really I'll call it three main elements, right? One of them is the call it the cryogenic or the cold environment itself, uh which is really needed to reduce the the external interference to the the the quantum processor or the quantum chip itself. And that's kind of a cylindrical fridge, if you will, that's about, I don't know, 4 ft tall, about a foot and a half wide, right? So, that's the cryogenics part of it. And then it has some equipment that's used next to it to keep it cold, right? Similar to compressors inside a regular fridge. The next is and that fridge itself has a bunch of wiring that goes inside it. And so, cables coming from inside from the top to the bottom, some of them are super- conducting cables, which have very low resistance, so that very low signals that are coming from the processor itself can can go out of the fridge. Then you need a bunch of electronics. So, there's a rack couple of racks of electronics that can send signals to this fridge. And one way that goes to the processor, and then the return signals from the processor go back to the electronics, right? And so, that's what converts the classical signals to quantum computation inside the processor, reverses back to classical signals that the electronics reinterprets. So, those are the two pieces. And then, the third piece is some level of classical uh servers, right? Typical servers that go could be similar to your home computer or similar to a very small portion of today's supercomputers. So, those are really the three key elements that put together a quantum computer that can make uh that can run workloads, run what the users want to do with a combination of quantum computing as well as classical computing. >> Got it. I mean, it's mainly like there's the processor, there's the fridge, the box the fridge goes in, there's the classical computing that controls it, and then there's a whole bunch of stuff that has to connect them and kind of deal with the day-to-day and all that, right? >> You got it. You got it, right? And then, exactly to just to kind of connect to what John was saying, part of the effort not only is to make this computer functional, but also to make it look good, make it look iconic, right? That was kind of part of it, and that was really the intent of the design aspect of this quantum computer, which ended up this one ended up being really surrounded by a glass cube, which ended up using these 10-ft by 10-ft glass pieces. >> John, how did this differ? I mean, you've already seen a lot of uh you know, large-scale computer installations separate from the chapel. I mean, how did the superconducting quantum computing part of it um you know, challenge and surprise your team? >> Well, certainly, when you you talk about cooling, you're you're not just trying to get down to, you know, uh 65, 70, 75° Fahrenheit, you're trying to get down to close to 0° Kelvin. That's a little different the cooling going on in the dilution refrigerator and before that. Um and it's also very sensitive to all those phases before it. I guess I've always looked at it in kind of three stages of cooling to do different things, but we have a chilled water loop that goes through the building into the system. We have nitrogen pumps that take it down to on the order of about 80° Kelvin and then you have a helium isotope bath of HE3 and HE4 to take it down to almost 0° Kelvin. That's a little different than most of the supercomputers I've put in place where you're mostly worried about um either forced air cooling hot and cold corridors or or even water cooling, but the water cooling chill chilling level is um in the you know Fahrenheit range that's reasonable, right? This is this is when you're at 0.015 degrees Kelvin. Um you know, and we calculated at one point that outer space is about 200 times warmer than that. Um you know, you're dealing with something entirely different in terms of and maintaining that. Um there's a interesting story when we first opened the system up on our grand opening day April 5th 2024 um we had of all things during the the event and it's a longer story, but I'll keep it short. There was an earthquake in northern New Jersey on the order of about a 4.6 degree earthquake that we actually felt the aftershocks and tremors in this region with our with our supercomputer and our quantum computer. The doors inside that chapel fluttered. I was in another building cuz we were we had a big auditorium where we had a bunch of people. But when those doors fluttered, you know, one of the first questions a lot of us asked is how did the quantum system do? And um I'll let Rajeev tell the rest of that story because it's pretty interesting what the quantum did, but it actually it stayed intact and made its way through an earthquake and on the stage I called it spooky science at a distance, but Rajeev was actually monitoring it much more closely where he was. >> Rajeev, how did the quantum computer do in an earthquake? It's a topic I've always wanted to talk about. >> What was interesting is to the temperatures that John was mentioning, you know, the baseline is about 15 millikelvin, right? And we saw a very clear spike that went up to about 80 millikelvin and then slowly, you know, came back down to base temperature and and that spike of course corresponded with exactly the time that uh the earthquake waves reached uh uh reached uh through New York uh and I thought that was pretty cool, right? And and I honestly use that example everywhere going forward when there are clients who ask us, "Hey, how should we design for earthquakes, right? How you know, about the quantum systems?" And that has happened certainly in Japan and and so on. Uh but really the reality is in mechanically, I think these are pretty sound and we follow what the rules are for that particular location. And you know, we have a very sensitive earthquake monitor itself has in the dilution refrigerator and >> It's probably a more expensive than something that monitors the Richter scale. >> [laughter] >> But but a cooler one nevertheless. >> Very expensive quantum sensing device which is put in place. >> [laughter] >> Right, you're actually building something that's like detecting the thermal properties of the system that you built, right? Because the whole idea is how strong this is and how good it is at isolating it from the rest of the world. So, it's actually amazing that it only went up by 0.003 K or something like that. It's amazing. >> It is and I I think part of it is and to be very honest, right? John's mentioned vibration, we've mentioned it is not something that's a perfectly known science even, right? We know vibration affects it. We know it affects the temperature. But, how much does it affect some of the key properties is something that we actually continue to study, right? Uh on how it affects you know, it Hey, we are part of the coherence times, the coherence of the device, right? Intermediate and and how it cycles and and and and and it uh evolves as a result of vibration. That will be a kind of continued study, but I thought that the impact is fairly sudden and and and you know, we know that within that range there's probably some temperature impact. If people were actually running jobs, they would see uh some issues related to the errors within the cubits that would very quickly died out. >> John, I mean, earthquakes aside, did you see any I mean, that's about as big a surprise as you could come across. But, I mean, what other surprises I I mean, I'd love to know kind of There were so many learning moments on this install, which I think is part of what made it amazing. It's stuff that was just like problems that were solved that have kind of helped us understand how to install things afterwards. Can you talk about that from your team's side? Like, what this means to maybe even building other classical computers, things like that? >> Well, I mentioned it earlier and I I I'll continue to reinforce that one of the fun things for me is when you get really good teams of people together and some of them were in my organization, some of them were in the IBM. They work off of each other and solve problems and uh the rest of us can just smile and say, "Well, it's good we have really good people working on this." Uh one of them that we hadn't mentioned yet is those 10-ft by 10-ft pieces of glass. Unfortunately, Rajiv and I had agreed to send them over by boat from Italy to the US and then by truck up from Italy to Troy. I'm sorry, from New York City to Troy. And then we had this whole mechanism with some wonderful outfitters that lifted up the piece of glass, slid it through that window that I mentioned before onto another platform, and then had another crane inside to lower it in and so on. So, we had this whole thing put together. We're doing this in the middle of January after a snowstorm just for, you know, raise the difficulty level of everything. And the first piece went in. Everybody's holding their breath and it went in just fine. It took a while, but it went in just fine. And I I used to tell people it's sort of like a NASCAR race where you may not care who wins, but everybody's waiting for the first crash, you know, around the first bend type of thing. And we didn't have that until we went to do the second piece. And it turns out the second piece of glass was damaged somewhere in transit, unfortunately. And we're all just heartbroken on this thing. It turns out the third, fourth, and fifth pieces of glass were fine, so we got them in. And I I went away just really angry because we'd spent all this time getting this right. The team had worked really hard to get this going and so on. And we had one piece that was broken. And And it turns out that the the glass that was made by Goppion in Milan is milled on the edges and there's three different patterns. There's a pattern for the front and the back, they're made the same. Pattern for both sides, they're made the same. And a pattern for the top. So, three different patterns, five pieces of glass. Well, one of the pieces the piece that broke was either the front or the back. So, the team came back to me later on in the day and probably to Rajiv also. And the guy who came to me said, "John, I think we have a solution." And I said, "I really don't want a solution right now. I want to be just angry at somebody. I'm not really looking for a solution. And he said, "No, no, no, hear me out." And I did. And he said, "What we think we can do is since it was either the front or the back that broke, we'll put the full piece of glass on the front. Still be a an integral 10-ft piece of glass. But on the back, we thought we would put two 5-ft pieces of glass. And instead of on one axis opening up like that, it would be on two hinges opening up from the middle out like barn doors. And and he said, "The advantage to that is Gopian actually can make that faster. They think they already have some in stock and mill it correctly. They think they can air ship it to us at a reasonable price rather than having to go by boat. And it'll now fit through the front door cuz it's only 5-ft high rather than a 10-ft high. And and as he sort of laid out every piece of this, I was like, "Okay, but okay, but okay, all right. So let's go with that." So by the end of the day, we had a solution to what could have been a major catastrophe and instead was a minor setback and we were off to the races. And it once again is a real credit to Rajiv's team and our team working together so closely. >> Yeah, that's awesome. I mean, and Rajiv, you know, kind of coming from your side, one thing like I've been kind of hinting at with RPI is that so much of this relationship has been like a learning experience for us, too. So I'd love for you to get into kind of the same side of what kind of lessons have been learned and carried forward from this install. >> Yeah, no. I mean, I would have a second everything that John said, right? The the collaborative troubleshooting was incredible. Amazing, right? So everybody kind of had that date in mind, but it really was everybody left off of each other to kind of figure out, "Okay, we have a new problem. How are we going to solve it?" And certainly, what we did with the glass was a key piece, but even leading up to it, right? So if you look at the elements that went into putting the computer together, whether it was making sure that the facilities that were going top to bottom, right? That was the first time we had we were doing that in any quantum computer installation. And just making sure that we had I mean, effectively, that we weren't going to miss something, that it wouldn't cause additional vibrations at Johnson said, "We started off at a very stable floor. We wanted to make sure that didn't change." Um you know, lengths of pipes were changing. We know that some of the cryogenics can get impacted if the lengths can change, right? In terms of cooling efficiency. So, we kind of had to make sure that wasn't um going wrong. There was another piece that related to, you know, we have certain security protocols on how we keep our hardware, what's you know, behind certain protected areas, what's not. Uh Short answer is I think what we did there became also a template for some of the other System 1 installations. Uh I think soon after that we had uh installed one in uh Yonsei University in Korea, right? Um and in fact, one thing that was common that we learned off of the one in RPI, again, it comes to the glass is the RPI one, beautiful as it looks from the chapel, also has people that are walking at a level an elevated level in the corridor. And so, they get to see the top of the system. And we wanted to make sure that some portion of that is not necessarily blocked off, but also for glass so that they could get a nice visual view of what it may look from the top. And the one in Yonsei had something similar. Uh and so, what we learned there is in how to really it again became the mechanics of how these glass pieces fit together, what kind of hardware needs to go in and so on. So I think that translated forward to what we did uh in the future as well as for us to understand that as as unique as this particular one was, we would want to probably keep the mechanical room side by side rather than top bottom. We certainly made that work, right? We made it work in here. I think it works incredibly well so far. We haven't seen any unique issues because of it. But in terms of the additional work that had been done, and almost knowing that there are chances that uh the location or the uh the our our team, our clients like RPI, may not be as amenable to ripping things up as as RPI was to be very honest and transparent, right? It was quite a bit of work that you had to do to make that happen. And not everybody would be ready to do that, right? For a variety of reasons. And so So I I would say there was quite a bit of learning mechanically, what are the things to do? How do we start setting tighter specific not tighter is a wrong word, but uh um make them as clear up front as possible. That makes it easy to prep the location to accept a quantum computer. While at the same time it makes it a little easier for us to say, "Okay, once we have this in place, we have a process to go install." So I think that uh I think we learned a lot from RPI from the RPI installation at that time to say, "We need to be a bit better in how we specify. How do we say it has to be this and not this? There that be a range, but let's try and do that." So, I think those those helped a lot in everything we did in the future, whether they were system ones or anything else. >> And uh you know, not just talking about system one, but we can even take it forward into the future. And I wanted to transition here to talking not just about quantum computing, but part of what makes RPI special in this case is that you also have this overarching, you know, computing center and program that's kind of fusing quantum computing into an overarching supercomputing framework that includes your Amos classical computer as well. Um so, I mean, you know, just thinking about that, John, can you tell me about how you're integrating quantum into this overarching program of yours and how you kind of envision it clicking into your supercomputer program? >> Yeah, a couple of things there. One is we've we're on our third generation. You mentioned Amos supercomputer. Uh we put about a 100 teraflop system in place in 2007. Six years later, it went to about a petaflop, and Amos is about I'm just under 10 petaflops worth of HPC. Um we see um quantum computing is a a much different approach to computing, a much different paradigm for computing. Uh we all I think understand that here. Um but there are all these programs that have been written for the more classical computing platforms, including supercomputers. And can you leverage that work and use HPC as an on-ramp to quantum computing? And it's not a completely replacive in all cases, but it might be replacive in some cases. So, can you take pieces of your high-performance computing and go to a hybrid model where quantum may be better at certain things, and then bringing those back? You know, quantum is not going to be better at data processing, for instance. So, it's a it's a lose to try to say we're going to give up on system Z or other things because uh have quantum systems. But quantum may be much better at some optimization routines and so on. And, I think that's where we're going to start to see the the tradeoffs. And, rather than try to go completely one way or the other, can you start to make um uh high-performance computing an on-ramp to quantum computing and back. And, so we've done some of the first experiments uh in the world with IBM. And, the other place that's notable is RIKEN, uh where we've had a HPC system call a quantum routine uh do some work on the quantum system, then return a result to the HPC routine without any manual intervention. And, I think that's a harbinger of what's coming in this quantum-centric supercomputing world that you'll see CPUs, GPUs, and QPUs uh all either on the same bus or in the same architecture uh and working on uh similar problem or the the pieces of the problem that they're best at. Uh and, I think we're at the beginning of of building those types of architectures right now. And, we have a interesting testbed with Ames and the quantum system one uh that we have in place. >> Yeah, I think this is what is sort of the this is the this is what we want. I mean, the idea here is that the quantum computer is like an accelerator in the same way that a GPU might be considered an accelerator. Now, of course, it's like a lot bigger and requires a refrigerator in order to do this accelerating. But, at the same time, you know, we need to start building these systems and think of them as that accelerator. And, that's kind of what I'm thinking about here with like you actually have this the start of this model where there's almost like a logical evolution from there, right? Where you have your supercomputer, and then nearby you have your quantum computer. You're starting to run these coupled workflows. How do you make them tightly coupled workflows? How do you make them like quantum accelerated workflows? But, you need the quantum computer and the classical computer in order to do that. And, I feel like I'm you know, that's what we're seeing, which is really exciting. >> And, the hidden gem there, I think uh Ryan, is the energy usage. Um if we we algorithms that are much faster than the quantum system, um they're going to as we scale them up, they're not going to go up uh in a linear fashion with respect to power in, which is not true as we're seeing in the classical computing world right now. These hyperscalers uh in these large data centers, uh the amount of power in is going up pretty much linear to the the amount of compute out, and we that's not sustainable in the long run. We've got to find a different model. Whereas the quantum, for the most part, once you've gotten to a certain level, the incremental compute is not really costing you much in terms of power. >> It's I mean, it's amazing. And that's the other thing people don't talk too much about is you see this big giant box, you think it's probably oh, that I know how much a GPU costs to run. That thing must cost a lot to run, but it's actually it scales quite a quite a bit better, and it's you know, the cooling is the hard part, but then it's not so bad. >> Yeah, and I and Rajiv, you can correct me on the figures, but I think our system is less than 100 kilowatts uh for the whole system. And we were talking to a group the other day uh and I won't name the group, but uh when we put out that number, and their their supercomputing was costing them 2 megawatts, uh they were like, "Huh, there there might be some real cash savings here to uh be able to fund uh quantum and so on. So, I I think there's something very real there. >> Just to kind of riff off of that, right? I agree with you. I think the the power of compute is not talked about enough with respect to quantum, right? Exactly that, right? We've done some calculations. If you take a look at today's GPU racks, right? And you look at power per square foot, if you will, right? Even if these are kind of large systems, and you start to do that, and the potential for what they could compute, we are looking at almost an order an order of magnitude difference between what is being run today versus what quantum could do, right? So, in terms of what you can get out in terms of computing power versus how much power you need to consume, has a pretty significant impact. And, you know, just again to what John said, right? And I think it was mentioned earlier, too, the reference architecture that IBM has put out with respect to where which are the right workloads that CPUs should work on, which are the ones that GPUs need to do, and which are the ones that quantum is best at. Once you start to really form workflows that can integrate them, I think one of many of those advantages is going to be in terms of reduced power or power per compute. Uh there's no doubt about it, right? And and I kind of mentioned an interesting part of that, too, where work is yet to be done, is how you schedule between the two, right? We all know, you know, if we we really heard of supercomputers are an expensive resource. And it's something you have to, you know, book ahead of time. You have to do this, okay, I'm going to get this time. Quantum computers are similar, right? You have a chunk, but how do you make sure that while you're computing on the quantum computer, your expensive supercomputer isn't sitting around twiddling their thumbs, and vice versa. Right? So, to be able to jointly schedule the two so that the jobs that are going between the quantum computer and the supercomputer are tightly interlocked, and there is that scheduling is certainly a lot of research work that has to be done. And I 100% right, RPI and and RIKEN in Japan are really leading the way for the entire world to know what's the right way to do it. And I can tell you that many of our clients going forward have looked at that work, have looked at this reference architecture, and are already thinking of what they're going to do together. >> Have you started to generate an idea from these installs about what it's going to look like? Like do you starting to envision kind of a future quantum computing system that actually has these more tightly integrated and coupled workflows? You >> Absolutely. So well in the sense that suddenly there's the quantum computer, right? We are we suddenly already have some level of classical resource integrated in our quantum computer, right? I think the discussion to be had right now is we are trying out different models of people our clients are trying out different models between how close in terms of geographical vicinity does the supercomputer need to be to the quantum computer, right? We already have an experiment going where RPI has it a little distance away. But then kind of the same campus a little distance away. Whereas, you know, the one in Japan for example are a few floors away, right? And I think that'll be part of it is which workflow requires what level of latency. And that's what we are trying to kind of figure out if the latency really needs to be that low and the intent is, "Okay, maybe not yet." But there could be future workflows that it would be, but I think you need these kind of experiments to happen or to be developed along these workflows to say what needs to be co-located. It absolutely has to be within the same building and what can be a few buildings away or a few miles away or a few cities away, right? Because that is also going to be the case. Today we are also finding people who have existing supercompute capacity. And for a variety of reasons they may may not want to put the quantum computer right there. Right? So I don't think there is an answer yet where it has to be co-located or has to be a certain distance away, but that's a really the research that will go on suddenly with help from RPI as well as others to say what is that optimum or maybe it's for certain workloads it has to be and for others it does And then I mean of course John, you know, as kind of one of the folks in charge of running these kinds of experiments and having, you know, basically one of the first models of what this kind of starts to look like, have you started to project out kind of the next round of experiments, the next round of kind of com- com- you know, combining quantum and classical and what's going to look like as you move forward? >> Very short answer is yes. Uh the longer answer is uh finding the right use cases to understand that in a little bit more detail. So, uh we have several researchers working on um some problems with benchmarking data that preexists or and/or data sets that are well known uh and uh either uh recalculating classical compute uh results and/or new results and seeing uh how they can go back and forth. Um it turns out right now, I I agree with Rajiv, this is a little bit of an area of uh discovery here. Uh right now, because of the way the queues work and such, um you have you have seconds perhaps uh between some of these applications going back and forth. As we get more and more tightly coupled, I think that's going to pretty quickly go down to the millisecond range and maybe be there comfortable there. Well, if you're in the millisecond range, you same building or same campus is pretty important. Um you know, I I I I think uh if I remember my fiber optics right, I think going from here to California is uh straight out fiber would be 30 ms speed of light. So, uh we're we're approaching that range of you have to be in the tens of milliseconds or such if you're going to uh start to do these experiments in a real way. And then, if you start putting these things on the same bus, you're going to go even faster than that. Uh and uh Rajiv mentioned schedulers, I think the schedulers are incredibly important to this right now Uh in terms of how do you uh uh schedule loads not only on the quantum system, but on the main system and then on both systems uh going forward. And uh uh make it such that you can get that tight coupling, really get the efficiencies out of these hybrid systems. >> Yeah, this is definitely something that we're, you know, Rajiv kind of mentioned this reference architecture, but it's something that we haven't talked enough about, I think, which is the fact that a lot of what the workflow in the future looks like is, of course, based on what the problems look like. And you know, different workflows are going to require different kinds of near-time compute, real-time compute. You start to put in things like error correction, error mitigation. All of these things are going to have their own um sort of considerations when you think about this workflow. And then, of course, with the scheduling, it's like, well, you have to run these programs on two like Rajiv said, you don't want your twiddling your thumbs. You don't want to waste your quantum resources cuz that costs money, your classical computing costs money. So, all of this has to happen in sort of a tightly coupled, efficient way, problem-specific. So, it's actually really exciting cuz these it's all happening as some of these applications and algorithms are emerging. And so, everything is all happening and pushing forward at the same time. So, um that's the cherry on the cake that I wanted to then pass back to you. I I really wanted to end with this one cuz I really am excited about it, which is just what was it like to turn the thing on? Like when you it was like last day, you you know, you finished the project and and you have a quantum computer in a church in upstate New York, you know, and you've been working months on this thing, you've broken glass, frozen timelines, whatever, whatever. So, uh either of you, you know, just tell me what it was like to be there. >> One is uh uh Ryan and the people that are listening to this, uh the person who funded most of this is a gentleman by the name of Curtis Priem, who is one of the co-founders of Nvidia. And Curtis was always disappointed that when they started Nvidia, uh they didn't document what they were doing with pictures or images or uh videos and so on. So, he wanted to make sure when we were doing this, he documented it. So, he actually paid for a film crew that came in and they actually produced a documentary. It's really pretty cool. PBS has picked it up and so if you do a Google PBS Project Chapel, you'll actually get to the the hour and a half, you know, so get your popcorn ready and everything else, but you'll get to the hour and a half Chapel Project. I think it's fascinating and goes over a lot of the things that we're just talking about and it's kind of fun to see the things that work, the things that didn't quite work as well as we would have liked and so on, but um for me uh this started at a board retreat in March of 2023. Uh great great idea came up from our board chair and our vice chair, which was Curtis and President signs our agreement in June. We do some testing over the summer. Uh we go to demolition and construction in the fall. The cryostat is at temperature in December. Uh we start going into testing and so on and then one of our research engineers actually had a working program that he worked on in February. So, by the next March, essentially a year later, we were able to show a working quantum computer on our system uh to the trustees and so a year to a year start to stop. And by the way, that researcher was at a conference working out of a hotel room in Minnesota coming back to the RPI campus to get access to this, but he made a video of himself and we showed that at the trustee retreat in 2024 and then had a grand opening in April of that year. So, it was just really neat and you you mentioned what was the moment like. I think there were a bunch of moments like and it was just very rewarding to within a year to essentially go from once again start to finish. >> Those are great stories. I agree with you. I think the time I have in my head is the moment we said go, which was that June date that John talked about, to the moment it was up was 9 months, and I think that so far has been the fastest we have gone from go to having something live. And I think that was incredible. It was an incredible um moment for everybody who had spent their time on it. I will say one thing, John, you may not know this, but I'm going to say it. I really, really, really wanted to be at the inauguration. I couldn't for a reason. I don't know if I ever mentioned the reason to you. Mike knew the reason I couldn't, but I had planned on going. Since it's World Cup soccer time, I'll mention it. I'd planned on going to a Premier League UK game around that time. And so I ended up going to see uh which was just pre-planned. I was going with a bunch of college friends. So, it just turned out that the the day was almost exactly the same day. And so that's why I missed it, but I know uh it was a fantastic day to really have everything come to you know, come to a perfect conclusion. Yes, the earthquake was on that day as well, but it was just fantastic. And I think we continue to cherish this relationship that IBM and RPI have had. Um I should mention, right, that particular system a little over 2 years later is going through an upgrade. And uh RPI is going to get a phenomenal new uh upgraded quantum computer. And it's just it's it's it's gone on incredibly well as well. And this has all come as a result of some fantastic collaboration so far between the teams and and I hope that continues uh for multiple generations of quantum computers. Did your team win? Oh, yes, they did. >> [laughter] >> Very good. All right. So, it was all worth it in the end. That's it for this episode of the Coherence Times. If you enjoyed our conversation, please be sure to subscribe wherever you get your podcasts, comment in the comments section, and share it with someone curious about quantum. You can find us on Spotify, Apple Podcasts, and YouTube via the IBM Research Channel. And for more episodes, resources, and deep dives, please visit us at ibm.com/think/podcasts. I'm Ryan Mandelbaum. Thanks for tuning in, and remember, the quantum future isn't just coming, we're building it right now.