Bản chép lời đầy đủ của Quantum Error Correction and the Road to Fault Tolerance with Benjamin Bloom

told me the process forced Atom to sharpen its pencils. They went back to the drawing board more than once, and now they have hundreds of pages of documents describing their roadmap in exquisite detail, which they now have to go build. There's also a competitive story. Google announced in March that it's building neutral atom systems, and it hired Adam Kaufman, who comes out of JILA, Ben's own scientific lineage in Boulder, to lead that effort. Neutral atoms were the dark horse a few years ago. Now the field is crowded with well-funded players, and I asked Ben where he thinks Atom's real differentiation is. We get into what the toric code result shows and what it doesn't prove yet, how atom loss went from the platform's biggest weakness to an engineering problem the team can manage, what Magne will let its first users do, and what Ben thinks utility-scale neutral atom computing looks like in 2030 and beyond. Ben also stepped back to CTO for a while when Rob Hays ran the company, then came back as CEO, and we talk about that too. Let's get into it. Sebastian Hassinger • 05:24 three or f three or so years ago. I saw the hundred cubit prototype machine. You had just moved to Colorado Benjamin Bloom • 05:34 Yeah, so that's where we started building our next generation. Sebastian Hassinger • 05:37 Yeah, and that was kind of a return to Colorado for you because it I mean do you sort of see your stint in JILA and the in the atomic clock engineer's laboratory as sort of the genesis of your quantum journey? Benjamin Bloom • 05:53 Yeah. I mean I think if we go way back to when I was a grad student presenting posters. the posters were all about how you make these amazing, amazing neutral atom quantum clocks. And then the little corner of the poster, there would always be like Oh, could you do quantum computing with this? Question mark, question mark. Sebastian Hassinger • 06:11 That's really cool. So I mean that's and you were part of the you're an author on the paper that was the clock that's is it ten to the minus eighteen. Benjamin Bloom • 06:21 First Riesel Adam clocks was ever. Incredible. Sebastian Hassinger • 06:24 Incredible. I really find I've seen, you know, I've known Janine. I went to QIP in Boulder that he was sort of the host for when I was at IBM. I've seen the documentaries being the what was the Edge of All We Know or something like that Benjamin Bloom • 06:38 he's featured in. Sebastian Hassinger • 06:40 Fantastic. And the work at Jilla is really extraordinary. I think that is such a beacon of how far you can take a quantum technology. So what made you sort of shift into the computing side of things? Benjamin Bloom • 06:53 yeah, I mean I did my PhD working on these, you know i kind of physics challenges that the border in engineering, like building these clocks and things like that. and then it was kind of more of a clean movement from there into just classical computing. so I think this is a long time ago. So this is,

you know, twenty fourteen or something like that. not that long ago, but fair bit of time. and I really wanted to go work on something that was, you know, gonna affect every single person's life. And so I went and joined Intel and started, you know, fabbing instead of you know, building one quantum, you know, clock. I was building, you know, billions upon billions of transistors. and it was fun and it was i it was cool like working in a group, you know, setting of a hundred something thousand people but you know, after two years I was like, well, I really want to go back to something where it's a small company, you know, there's there's huge technical risk and high impact from what you're doing. and it was there that I actually went and joined the quantum computing industry. So I was started as an IC at Rigetti. I don't know what employee number I was at Regeti, but it was within the first 20 or so. started as an IC became a manager, learned all about superconducting qubits and, you know, went from you know chips that had, you know, zero or one qubits on them when we're doing our initial testing, all the way to being able to make chips of twenty qubits. And that was a great transition because I learned all about all the pieces you need to make a quantum computer. and it was around that time when I started thinking, you know, the work I did as a grad student Actually pretty similar and could potentially be useful. Sebastian Hassinger • 08:38 And I mean that was at the beginning of sort of looking at neutral atoms as a as a potential modality. Benjamin Bloom • 08:44 Yeah. Sebastian Hassinger • 08:45 But you made sort of a unique choice, them your your qubits are the strontium, right? Is that correct? Benjamin Bloom • 08:53 Well we started with strontium, but we switched to uterbium, but i even it doesn't I mean I think that as we've learned like What you have to build is the technology around your system. Like we have Uturbium, we continue to love Uturbium But I also at this point don't know if it really matters anymore. Like the neutral atom is a platform and it's all about what are the pieces of technology you build. to make that platform scale. Sebastian Hassinger • 09:18 Right. Benjamin Bloom • 09:19 And I think that uterbium is a really good candidate for that, but the technology is kind of Atom agnostic. Right. Sebastian Hassinger • 09:28 It's neutral in atoms as well as neutral in atoms. and you know you take an an interest, I mean it's it's always I think at this stage of an early technology, as you said, high risk and high potential impact, there's no blueprint for how you take you know, d what startup path you go down as there is with more of a proven technology where it's incremental innovation rather than you know paradigm busting in innovation. The the path you've gone is been you know we you've got a fairly tight partnership with Microsoft at times and particularly in development of the error correction codes and implementation of error correction. And in June you brought out a paper that I think Matthias Troyer called, you know, Earth Shattering or something like that. It was a pretty it was a pretty good quote. So what was that paper? What was the result of the error the error correction paper machine? Benjamin Bloom • 10:26 I mean we've taken this idea that you have to build a neutral atom system in a way that you will use a utility-scale neutral atom system. And so you can imagine that there are ways to do shortcuts or clever clever things that allow you to kind of jump ahead in certain directions, but kind of don't solve the fundamental challenges. We always take into this other approach, which is you only get one shot at everything, so everything you have to do has to be a stepping stone towards an utility-scale quantum computer. So what we've done over the last few years is we feel like we've solved the kind of binary challenges with neutral atoms like Can you get them to do mid-circuit measurement? Can you do reloading? And all these pieces that people kept saying, oh, if you had this, you could build a utility-scale quantum computer with neutral atoms. We took a no you there's no There's no ifs, ands, or buts about it. You just have to build the things you need. Sebastian Hassinger • 11:21 Right. Benjamin Bloom • 11:22 and so the paper in June, it is very, very similar to what's been done at USTC, what's been done at Google, which is You have an error correction code running, and it's running many, many cycles. In this case, it was like 90 cycles. I'm sure in future versions it'll be more than that. But the idea was You get it to work and operate exactly like you would a utility scale quantum computer, which is you have mid-circuit measurements, you deal with the fact that some pieces of your system are imperfect, so you're gonna have atom loss. But you make sure you have a solution to that. And that's what we showed is how you actually can reload the atoms. You can kind of cycle which ones are the data qubits and which ones are the ancillas. And you can do all of this in honestly, the exact same way it will be done five or six or seven years from now. on systems that are, you know, a thousand times larger. Sebastian Hassinger • 12:13 Right. Benjamin Bloom • 12:13 probably better too, but at the same time now there's no binary lift anymore. Sebastian Hassinger • 12:18 Right. Benjamin Bloom • 12:19 It's now just, okay, increment all these things. Sebastian Hassinger • 12:21 That's really interesting. And how many atoms were in that in that device? Benjamin Bloom • 12:26 that's a great question. I don't know what to do on the hand. So that's one of the crazy things about it. It was probably on the order of Well, it's actually even crazier than that. So it's probably on the order of like two hundred or three hundred atoms or something like that. Because you're like cycling atoms in and out of them. Sebastian Hassinger • 12:45

Right. Benjamin Bloom • 12:45 And so I would have to actually think about well what is If we counted up all the atoms that pass through that circuit, what did do? but so I think there's this funny thing with neutral atoms, which is that At the end of the day, you build all these optics and you know each optics system you build is like a tool in your toolbox. But one way or another, you're gonna make that software programmable like that tool. Sebastian Hassinger • 13:06 Yeah, yeah. Benjamin Bloom • 13:06 And then when you actually kind of sit there and program the circuits on it, everything is software programmable. Sebastian Hassinger • 13:12 Right. Benjamin Bloom • 13:12 So it's like if you operate with this number of qubits, if you have this qubit talk to this qubit, if it's moving from here to there. all of its software programmable. And so I think a lot of the time people ask us, you know, they look at our roadmap and they say, well how many logical qubits are in this system and how many logical qubits are in this system? And we kind of give approximate numbers because we pretty much say Different people are gonna run different error tracking codes. Sometimes they'll want way more logical qubits. Right. And sometimes they'll want very small numbers. And is that true Sebastian Hassinger • 13:41 off sort of in the distance like you get in? Yeah. It's fewer logical qubits for higher D and more resiliency essentially for errors. That's interesting. So and it was it's toric code you were implementing. Benjamin Bloom • 13:53 Yeah, so that one was like a very rotated toric code. Or Sebastian Hassinger • 13:58 or I guess it's actually the it's the ancestor of the Surface Code. I guess because Katev can came up with toric before Surface Code was formalized. And Is that was the choice of Torque because it lends itself to your to the t the platform that you're building better or what was the what was the rationale? Benjamin Bloom • 14:17 I mean I think we had already been thinking about it. It's like one of these like base you know, codes of like people learn about these codes in like the evolution of quantum error correction and since we can create these periodic boundary conditions just by moving the atoms around we can enact this. I think there's been lots and lots of great work, and I think cited in that paper, the work of like Microsoft and David Desan about rotated Torah codes and how you can actually create these like savings in the qubit number by actually playing some games about what you're calling what and so on and so forth. But I feel like the reason why we did it is because at the end of the day And I wish I knew the numbers off top my head. At the end of the day, y you're just gonna have some code block, which is like NKD, and That's how you're going to use a neutral atom quantic Peter. It's like it doesn't matter what is the exact connectivity. You're gonna end up moving the atoms around and just dealing with whatever the connectivity is. And so when I think of how you imagine a neutral atom quantum in the end. It's not going to be some giant slab of surface code or something running on the machine. Sebastian Hassinger • 15:20 Right. Benjamin Bloom • 15:20 It's going to be these block decodes. Yeah, yeah, yeah. Sebastian Hassinger • 15:23 That's interesting. And obviously that's That's leaning into the strength of the neutral atom platform, which is it i the topology is entirely plastic. You can rearrange the atoms in any way you want in that three-dimensional space. Benjamin Bloom • 15:35 Yeah. Sebastian Hassinger • 15:36 As you said, you move it, you know, have encillas over here and memory over here and all the I mean you can sort of build I think I when I interviewed Vladin Vulatic, I said the way he was describing sounded like he was building a von Neumann machine in thin air, basically. Yeah Benjamin Bloom • 15:51 and Vladin Vulatic was my undergraduate advisor, actually. Oh wow Interesting. That's really interesting. Sebastian Hassinger • 15:58 I'm just gonna pause for a second because I forgot to ask you to take your name tag off. So I just heard a jingle. Benjamin Bloom • 16:03 So yeah. So Sebastian Hassinger • 16:05 easily edited out. that's really cool. So obviously so the Microsoft partnership collaboration has been primarily around error correction, but there's also the delivery of the Magna system to QNorth in Denmark. what what's that sort of that's looks almost like a joint go to market rather than a system RD collaboration. Benjamin Bloom • 16:31 I mean it is. I mean I think there's in that's actually kind of how if we could go back like how we started the relationship with Microsoft was RD collaboration and then as we got more and more confidence about what we could do, we started adding this go to market space and so I think they're a great partner. I think they bring, you know, a lot of software above the quantum computer and they also bring a lot of expertise in quantum error correction. so they're a great partner to work with and I mean when I think of the future, I mean an end to end, you know, workload on a quantum computer, it's going to touch a lot of different pieces. Sebastian Hassinger • 17:08 Yes. Benjamin Bloom • 17:08 And I think it makes sense that there's going to be partnerships like this, partnerships with different pieces in the stack, whether that's algorithm companies, whether that's, you know, specific quantum error correction companies. I think All of that is going to end up having to be part of it. And then you are going to have to bring in classical computing. Sebastian Hassinger • 17:25 Right. Benjamin Bloom • 17:25 So like when we talk about, you know, what is the workload for a quantum computer? It's going to be something complex that honestly now I believe an AI agent is going to figure out how to take a problem, cut it up between CPU, GPU, and QPU, and say, well, this piece goes here,

this piece comes here and so on. Sebastian Hassinger • 17:42 Yeah, yeah, yeah. That's really interesting. And the design of Magna has been referred to as the first commercially available level two machine. What's meant by level two? Benjamin Bloom • 17:53 for them it's all like you know, this idea that you can use it as a logical qubit machine. So it has enough logical qubits that you can kind of do a step beyond what you could right now if it's just playing with a few number of logical qubits. And that these are real logical qubits with, you know, quantum error correction. Like not just detection or anything like that, but actually have a large enough distance that you're doing some really clever things. and so I think they've been looking at a few different ways of utilizing that system. And I think that obviously people can try other quantum error correction on the system. Sebastian Hassinger • 18:27 Yeah. Benjamin Bloom • 18:28 But I also think one of the goals and one of the you know strengths of actually having mart Microsoft as a partner for this is they're going to come pre-baked with, hey, this is a quantum error correction layer that is proven and works really well on the system. Sebastian Hassinger • 18:41 Yeah. Benjamin Bloom • 18:42 And then you can use the logical qubits and Of course you can skip it, but now there's something that you could start and say, Great, this works well. Sebastian Hassinger • 18:49 That's really cool. And was the system were there a set of requirements from QNorth that were expressed in number of logical qubits they wanted to support or the distance of the error code or was it in terms of We want to be able to support this use case. Benjamin Bloom • 19:10 It was more like a use case actually. but yeah, no, there there's some specific like, hey, we're gonna run this kind of you call it an experiment, or we're gonna run this benchmark with logical qubits, and that is going to be how we measure that the machine is really doing what we say it's gonna be doing. Sebastian Hassinger • 19:32 That's such an interesting target, in terms of specing out a machine. And again, it feels like that's one of the strengths of the neutral atom platform is that it's not you know, you have an enormous amount of freedom of degrees of freedom in terms of what you what you actually assemble. Benjamin Bloom • 19:48 Yeah. Sebastian Hassinger • 19:49 is that is the flip side of that set of strengths the trade-off is the speed of operation, or this sort of it's more difficult in my mind, at least maybe I'm wrong, to sort of make that into a hardened product, a repeatable product of with some standardization. Is that is that fair to say? Benjamin Bloom • 20:10 I mean I don't know about the last one. I mean I think one of the things that we're learning how to do is create very, very good optical systems and utilize, you know, manufacturing both in-house and out-of-house to make these systems repeatable and buildable and easier to deploy and everything like that. now I think that the other piece to this is you have so much freedom, it's pretty easy, and I think we've even done this, to accidentally you know, force yourself into a box and then get to, like you said, slow run times. I think as we've learned in the last maybe nine months or twelve months, it turns out to also be easy to rephrase the problem again, which is how do we utilize the strengths And then have a goal, not necessarily of saying what is the smallest qubit quantum number of quantum computer you could have at utility scale, but we go back and say, great. We want to solve this problem in X time because people are not going to wait longer than five hours or one hour. Sebastian Hassinger • 21:19 Right. Benjamin Bloom • 21:19 And it's a question of, okay. What could we do with neutral atoms to do that? And I think that has actually turned out to be a huge strength of neutral atoms, which is now going back and saying, great, what were the things we could do that no one else could do? It's really complex connectivity. It's actually fast gates. Like if you actually think about other modalities that have, you know, atoms that are kind of or ions or things like that, you know, the gates are a little slower on ions and stuff And so the question becomes, great, you have all these nice properties, but you have this one negative thing, which is movement. Movement takes time. Sebastian Hassinger • 21:52 Yeah. Benjamin Bloom • 21:52 And it's like, how do you honestly pipeline around that? Sebastian Hassinger • 21:56 Right. Benjamin Bloom • 21:56 And I think we've come up with some really, really clever ways now where we can very, very cleverly pipeline and we can talk about runtimes that are as fast or faster. than some of the leading platforms in the world. Sebastian Hassinger • 22:09 That's really cool. I l I like the idea of sort of reverse engineering from like it's almost like time value of knowledge. Benjamin Bloom • 22:15 Yeah. Sebastian Hassinger • 22:15 Right instead of a money. It's sort of like, well, if it if it takes over X amount of time, it's a worthless. Yeah, yeah, exactly. Exactly. Benjamin Bloom • 22:22 No one wants to wait like whatever is days, 100 years, or 6,000 years for a thing You have to find a way to Sebastian Hassinger • 22:28 draw the line a hundred years. That's why I'm never gonna build a cathedral. So clearly this approach, this partnership s as you said, starting with R D, now go to market is working. I the Magna machine is that sort of seen as a prototype of a product that you'll repeat in or is every subsequent order going to be sort of tailored to the use cases that customer wants to tackle. Benjamin Bloom • 23:00 No, it is it's actually almost the opposite. So we had a prototype and what we've done is we've actually kind of hardened it and we've created a few of these hardened ones and one of them is Magni, one of them is in Boulder. I

think that, you know, for the right customer, for the people who want to have a machine like that, we're happy to work with them and get them access to that. I also think we're not slowing down. Like I think we laid this out, I don't know when, but a few years ago with our roadmap at like a quantum world congress. We want to 10x the number of qubits inside the module every time. And so, yeah, we're gonna go 10x this time, or 10x another time. Like so we will s be building systems with many thousands of qubits soon. And I think that's important because at the end of the day, you have to build useful quantum computers for all this to work out. And the only way to do that is to aggressively scale. Yeah. Sebastian Hassinger • 23:51 Yeah. Do you have a sense of w if there's a physical limit to the number of atoms that you can have in a lattice in a single chamber before you have to sort of have multiple devices Benjamin Bloom • 24:03 Yeah, I mean we so just to say this, we have Sebastian Hassinger • 24:06 I mean I know the atoms don't take enough space. The optics to actually hold them in place and the tweezers and all that stuff. All those lasers take space. Benjamin Bloom • 24:15 I mean what we do is Yeah, we kind of believe that maybe there's not a hard limit, but there's an engineering limit. Sebastian Hassinger • 24:22 Yeah. Benjamin Bloom • 24:22 So in the meantime, what we've done for the past about one and a half years, almost two years, has been building networking equipment so that you can have many of these systems talk to each other. And I would say that will slide in at the right time into our roadmap for our products. But when we think about what you can do and we look at technology right now, it seems like you can get to tens of thousands to hundreds of thousands of qubits inside a single module. and yeah, it might be hard to get to a million, will it be impossible? I don't even know if it's really impossible, but like at some point you also want to stop doing engineering. So the way we look at the networking is including m a networking that's you know works really, really well with our system. is also the point at which you can maybe not stop doing an engineering, but you can sort of slow down your engineering and you can pretty much say, Great Here's a module, let's make 50 of these or let's make a hundred of these. Sebastian Hassinger • 25:20 Well, I mean I would assume there's even potential parallelism to overcome the kind of performance issues. That's another way to get around the performance issues. have a bunch of movement going on at the same time in multiple chambers. Yeah. That's really cool. And again, I mean like it's clearly working because not only have you got the Magna system on the way to being delivered. It's twenty twenty seven I think is when that will be operational, but you have new news about the platform being accepted. So Benjamin Bloom • 25:52 Yeah. So they so we're in DARPA phase C, which is a

Congratulations. Thank you. Sebastian Hassinger • 25:58 massive Massive win. Benjamin Bloom • 25:59 Yeah, I mean I also think it's just more of a validation also that neutral atoms are, you know, went from a dark horse candidate a few years ago. Like I remember when I started the company, there was even like a came from like a nature news article or science news article where they refer to as the dark horse candidate. I don't think neutral islands are the dark horse candidate anymore. And I think that part of that is you know, now with enough scientists and engineers throughout the national labs as part of that QBI program, looking at what we're building, looking at technology and saying, wow A lot of this has a huge potential to work. Sebastian Hassinger • 26:34 Yeah. Benjamin Bloom • 26:34 And it has a huge potential to create utility scale quantum computers. And I think that's what they've seen after having worked with us for the last like year and a half or so. Sebastian Hassinger • 26:42 Yeah. Yeah. I mean I was just speaking with Charlie Tahan, of course Microsoft's Majorana two is also in the program and he was saying that he feels like that external audit or benchmarking will be sort of a really powerful steel to hone Benjamin Bloom • 27:01 Yeah. Oh, I mean I think it forced us, I can't remember who told me this, but I was like, yeah, that's a good analogy. They're like, it really forced us to sharpen our pencils. Sebastian Hassinger • 27:10 Yeah. Benjamin Bloom • 27:10 Like and I think we were surprised by some of the things we learned and we had to go back to the drawing board a few times and make sure what we are working on and what we are building actually scales. So that when someone says, I want to run this crazy quantum chemistry problem, not only does it actually run end to end but the time it takes to run is a reasonable amount of time and you've spent a reasonable amount of money on that the hardware, on the power, on everything about that. Sebastian Hassinger • 27:38 Yeah. Benjamin Bloom • 27:38 Yeah. Sebastian Hassinger • 27:38 Yeah. And it's I mean, as you said, it's a validation for the neutral atom as a modality. What do you how do you sort of perceive you know QR has been at this for a while, Pascal's been at it for a while, inflection is one of the OGs in the space. And there's now Kaufman was just hired by Google from also worked for GLUD. and Oratomic just raised three hundred million off a research paper. Benjamin Bloom • 28:06 Yeah, yeah. Sebastian Hassinger • 28:07 Are you worried about like your ability to differentiate with that many competitors or are there unique ways in which you approach neutral atoms that are that are atoms Benjamin Bloom • 28:18 Well strangely enough, yeah. Strangely enough, we were actually one of the first mutual item companies. so in fact, I talked to the founders of Pascal and then I got told later on hilariously that conversation convinced them to form Pascal.

Sebastian Hassinger • 28:31 Oh wow Benjamin Bloom • 28:32 I actually talked to the folks who, you know, founded QRA and I think they started off like with analog computing and I think that kind of push them down that direction for a while. you know inflection's been around for a while making centers. So I guess you could say they're they're one of the OGs because they started the center. I mean I do think that the government engagement we have is a huge differentiator. my guess is the tech Technology is differentiated, just not in a place that people see. Sebastian Hassinger • 29:02 Yeah. Benjamin Bloom • 29:03 And so when we talk about Sebastian Hassinger • 29:04 atoms being atoms. Benjamin Bloom • 29:05 Atoms being atoms. But like when we talk about actually all the technology we're building to Make the neutral atom platform fast. There's no outside papers about it. There's no know anything about it except for what we've given to say the DARPA QBI folks. It's pretty unique to Adam, I would say. And when people come and interview, we tell them about it and so on and so forth. But I have a feeling like that is where the battleground is gonna be had, which is look, everyone has the same atoms. Sebastian Hassinger • 29:35 So operational actually in the in the actual technology itself. Yeah, yeah, yeah. That makes sense. I mean performance is obviously incredibly critical. Benjamin Bloom • 29:43 Yeah. Sebastian Hassinger • 29:44 and so like that's that makes a lot of sense. the other thing I was curious about, like You founded the company and by the way you do also have the advantage of the generic brand. Benjamin Bloom • 30:00 Adam Sebastian Hassinger • 30:00 Adam quantum computing is sort of like Kleenex, right? Benjamin Bloom • 30:03 Yeah, yeah, yeah. No, that was the greatest thing I've ever done in my life is name. company. Sebastian Hassinger • 30:09 It's it's a great name. It really is. there's there's a very strong brand value in it, I think. but you f you founded it and stepped back to CTO or Rob Hayes ran the company and then came back as CEO. Benjamin Bloom • 30:22 So Sebastian Hassinger • 30:23 what just talk about like the difference in the roles and what did you learn th in that in that transition, sort of from founding to CTO to CEO? What's sort of that perspective? Benjamin Bloom • 30:33 I mean I think when we started the company, I mean this is the true I'm sure of all companies as you grow, which is that it at every time, you know, is the right person to bring on at the right time. I mean when we started the company you know, we were just a bunch of physicists. Like there were a bunch of physicists in a room and that's how we built our first setup. and I think at that time we probably maybe when you came visit, we're probably like I don't thirty thirty five people or something like that. Sebastian Hassinger • 31:00 The team in Colorado had just been stood up at that point, so I think that was just

five to ten people was just starting to build. So Benjamin Bloom • 31:07 Yeah. I mean I think that you know the company grew to about seventy people. We started off with relationships with Microsoft and stuff like that. And so I think that having that CEO who can go and talk to hyperscalers and stuff made a lot of sense. 'Cause I think hyperscalers were still trying to understand neutral atom. Yeah. Sebastian Hassinger • 31:25 Yeah. Benjamin Bloom • 31:26 and then when we took when I came back, you know, I mean I think then the goal was great. Now we have a base of partners, how do we go and continue just pushing on the technology? Because I mean utility scale quantum is a long, long road. And you know, this is a very, very technical company, honestly. When I took over again, we're probably like seventy-five or so people. now we're about a hundred eighty people. We'll probably be in the next year and a half or so, I think we'll double in size. and I think part of that is that we're trying to build this amazing roadmap we've made. You know, if you go back to what I was just saying about QBI We really did sharpen our pencils and now there's, for better or worse, hundreds and hundreds and hundreds of pages of documents Sebastian Hassinger • 32:18 that Benjamin Bloom • 32:18 explain our roadmap. In exquisite detail and now we have to go and build those things. So I think that we've expanded engineering a huge amount. Like I think that when we talk about you know the makeup of the company It was a bunch of physicists, and then it was physicists and electrical engineers and like a beginning business team. Now it's like we have the business team, it's still growing. But it's not just physicists and electrical engineers. Now we've had optical engineers and mechanical engineers and now system engineers and manufacturing engineers are coming. And like all of these different aspects of the company are kind of coming together as quickly as possible because it we have to build so much stuff. It's not just build the machines, it's build all that technology I'm talking about, whether it's how do you steer laser beams super super quickly. How do you integrate the networking pieces into the actual compute so that they can actually all simultaneously work? How do you make the system easier to build? Like I think easier to build and easier to integrate is really, really important for neutral atoms. I mean I'll be honest, it's probably important for all of quantum computing, but Sebastian Hassinger • 33:29 Yeah, I mean that's in that general category of productization I was referring to. So it that's a really good perspective. Do you think that draws more on your experience from Intel or more from Regetti or Equal? Benjamin Bloom • 33:41 I've probably more from Morghetti. I mean in Morgetty it was like, you know, building all of these different pieces at the same time to try to be able to stand up systems. And that was that was more focused on the NISC

kind of error systems, but at the same time you had packaging engineers that were coming up and you had cryo specialists and you had fab and so on. I mean the stuff I learned at Intel was actually interesting because it's more like I was there for a few years and I kind of just saw, you know, what does integration do in a fad? And actually, you know, when I go and hire physicists now with this huge plethora of other teams I tell physicists, hey, you're pretty much the integration in the fab. Sebastian Hassinger • 34:19 Right. Benjamin Bloom • 34:19 You understand the quantum computer from end to end. Sebastian Hassinger • 34:21 Right. Benjamin Bloom • 34:22 The optical engineer doesn't necessarily have to understand the quantum computer from end to end. Sebastian Hassinger • 34:25 Right. Benjamin Bloom • 34:26 The electrical engineer doesn't have to understand the quantum computer from end to end. You, the physicist, you have to understand the quantum computer from end to end. Sebastian Hassinger • 34:33 Interesting. That's really interesting. Okay, so you know, the The it's now 2030. You've delivered multiple Benjamin Bloom • 34:43 fast. Sebastian Hassinger • 34:47 I know. So what does a utility scale Atom computer look like and what types of Things are your end user customers, what are they doing with it at that point? Benjamin Bloom • 34:57 So I think in 2030, I mean I think we are capable of deploying networked systems and we are starting to do that. But I'll be honest, the first utility-scale quantum computers don't have to be networked. Like they will be one system with many tens of thousands, potentially hundred thousand qubits in it. And They will be doing calculations for materials modeling and things like that, or maybe, you know, for spookier sections in the government doing cool things. But generally speaking You can get a lot done for, I would say, these national lab, these big science customers with one system. I think if you go a little further into like 2031, 2032, Then all of a sudden you're talking about, well actually now I have systems that have potentially millions of qubits across like multiple different machines. And you are doing things for pharmaceutical companies, you are doing things for commercial material companies, commercial chemical companies, Sebastian Hassinger • 35:55 optimization, optimization. Horizontal. Yeah, yeah. Benjamin Bloom • 35:58 And I think at that point you're also saying, look, you know, you have a fleet of these things and for the most part, you know, the pricing and how much of it you're using in that fleet is all based off of like time to solution. Like if someone wants a problem done faster, maybe they pay more and they reserve more of that fleet. They get their answer and then the fleet goes back together and now all of a sudden you can have ten different companies utilizing that fleet with different, you know, workloads and so on. So I think that's where you get back

into the kind of cloud business. Sebastian Hassinger • 36:29 Right. Benjamin Bloom • 36:29 and you get a little bit farther away from these machines sitting in these big science or big compute scientists. Sebastian Hassinger • 36:35 Do you anticipate that network connectivity between modules is that sort of relatively long range as well as as sort of because I think of coupling sort of you know interconnects and then actual networking, which which category you sort of imagining in that scale. Benjamin Bloom • 36:51 I guess when the way you describe it, I actually think that maybe what I'm referring to is interconnect. Sebastian Hassinger • 36:56 Yeah. Benjamin Bloom • 36:56 Which is that I think that in our base. Sebastian Hassinger • 37:00 In the data center, yeah. Benjamin Bloom • 37:01 I think we have a lot of clear plans about how it works in the data center. and we kind of have an idea of what we can do ourselves. We have this partnership with Cisco, partnership with New Quantum. I'm sure we'll partner with others as well. about how you go a step beyond that and have more than just a minimum viable product, but how do you make this thing super resilient? How do you have amazing switches and so on and so forth? But when I think about it, I always think of, well, this is the data center. This is the data center. Sebastian Hassinger • 37:27 Yeah. Benjamin Bloom • 37:27 Now a part of that is that like if you want to do computation, like most of what we focus on is computation. Once you do these like long reaches between different systems, it's it's not the best for computation, is what I'll say. Right, right. Sebastian Hassinger • 37:43 Yeah. That's a amazing vision and it is I mean, you just made a joke, it went by fast. It's gonna be twenty thirty before we before we realize it. It's happening so quickly now. So thank you so much, Ben. This has been really, really great. I feel like I have a much better grasp. for what Adam's up to and again congratulations into getting to the QBI program at Stage C because that's a huge accomplishment and a real vote of confidence in what you're doing. Benjamin Bloom • 38:09 No thank you. Sebastian Hassinger • 38:13 Thanks so much to Ben Bloom for joining me today. What's going to stay with me is how Atom changed the question it designs around. Instead of asking for the smallest quantum computer that could work at utility scale, they ask what it takes to solve a real problem in the amount of time a real customer is willing to wait. That moves the focus from qubit count to time to solution, and I think that's where the whole field is heading right now. If you want to go deeper, the show notes have links to the June 2026 toric code paper on arXiv, coverage of the Magne system and the QuNorth deal, and background on Atom's part in DARPA's QBI and the Department of Commerce LOI. It's all at newquantumera.com, along with the full transcript. While you're there, sign up for the newsletter. It's the best way to hear about new episodes, plus the occasional longer piece on what's happening in quantum. If you listen on Apple Podcasts, Spotify, YouTube, or Amazon Music, please subscribe if you haven't already. And if you liked this episode, share it with someone who's following the race to fault tolerance. That really is the most helpful thing you can do for the show. Thanks for listening. I'm Sebastian Hassinger, and this has been The New Quantum Era. Theme music by OCH. See you next time.