Transcript: Dr. Cameron Cogburn interview

The following is a transcript of our September 3, 2026 interview with Dr. Cameron Cogburn. The transcript has been edited for clarity.

The Poly: I listened to your talk at the Engineering Quantum Future of Precision Medicine and Life Sciences conference. You narrated a fun story about how you first interacted with the Nighthawk processor. Do you want to share more about that?

Dr. Cogburn: When IBM first released the Nighthawk chip, the R1, the first revision of the chip, it was December 19 of [2025]. And, you know, we have our own on-premises computer, IBM Rensselaer, but we also have some limited access to these systems on the cloud. And I knew that this was going to be an important step from the current architecture which we have, which is this heavy hex topology on IBM Rensselaer at the time. I had a problem which I'd been working on and kind of got to a point where I was stuck. I couldn't get the result I wanted on our hardware, so I was waiting to run it on this new chip. I thought it would do really well there. It took a couple of days because I was actually traveling with my family back home for the holidays. It was Christmas morning when I was able to finally submit the job to the hardware.  I'd submitted the job early in the morning and then everyone woke up and we went downstairs and had breakfast and Christmas presents and all this stuff. And I had come back upstairs and the result from the job was on my screen. It just showed exactly what I'd hoped it would show. It showed an agreement that I couldn't get it on the earlier heavy hex topology. So, that really kind of emphasized to me that number one, technology is changing very quickly. And number two, that when it does, it can just open up a whole new class of problems that you couldn't [solve] before. And number three, if you want to be on the front of that wave, you need to start thinking now about the problems that you can run on the hardware you'll get in the future or the frontier hardware right now.

What was the job that you were running? What was the problem?

So, the problem I was looking at was looking at the potential energy, the interaction potential between these two objects in a model that is used as kind of a toy model for quantum chromodynamics. So the theory of the strong force, right? These objects, they're called a kink and an anti-kink, but basically, it was looking at the interaction potential between these two objects. And when they're close together, when the strong force is very strong, when the interaction strength is strong, that is precisely where on the old hardware, the plot is supposed to basically asymptotically exponentially fall down, but instead it kind of did the opposite. It exponentially went up just because of noise and how the circuit was run. But on the Nighthawk chip, I was able to arrange the qubits used such that noise was much less of a factor and the curve I got from the quantum data really almost overlapped with the [numerical values]… It's a small system, so with the exact numerical result, I was able to do. So, yeah, it was really neat. 

Pretty solid. The Nighthawk architecture obviously changed the physical layer of the qubits into a grid rather than the heavy hexagon. How would you describe/quantify or visualize the interactions that you were looking at on this chip versus the other chip?

 So, I would think of it like this. The Nighthawk chip has the qubits arranged in a two-dimensional square grid. Whereas the machine… And I should say that we actually have this on campus right now, and it's been live since the 15th. Two weeks, three weeks at this point. So, we have our own right now. It has this 2D grid, whereas the previous generation chip, they called it a heavy hex architecture. And that's simply because if you kind of squinted your eyes, it looked like hexagons that were a little lopsided, so that's why they called it heavy hex. But the bottom line is in the square grid, you can imagine that each qubit has four neighbors, it has a coordination number of four. Whereas in this heavy hex architecture, most of the qubits could only have two neighbors, and some had three neighbors. In our new chip, each qubit touches four other qubits and can talk to four other qubits. Whereas in the old one most of them can only talk to two other qubits. When you put your problem on the quantum computer, a lot of times, you'll need various qubits to talk to each other. And if they're right next to each other, then the noise is minimized. And so to do that on either type of architecture, if a qubit's far away and not immediately touching another one, you have to use what's called a swap operation, a swap gate. And that's fine. Basically, you can think of it as bringing a faraway qubit and putting it right next to the qubit it needs to interact with. So on the heavy hex architecture, you had to do this swap operation a lot of times. Just because most of the qubits only talk to two instead of four. Whereas on the Nighthawk, you don't have to do it as much. And in fact, there was a special thing about the system I had set up where the qubits it needed to talk to, I could set up one line of what I call the physical qubits, the data qubits. And then on the outside was kind of these, they're called ancillas or measurement qubits. They were right next to the data qubit they needed to measure. So there was essentially no swap operations. And that made it possible to get this result. Because with a lot of swap operations comes a lot of noise in the result, and that was the fundamental problem happening on the earlier heavy hex architecture.

How much has the noise reduction been?

It really depends on the problem. So all things being equal, it just depends on the problem. But basically—I don't know if this is mathematically true or not—I essentially think whatever factor of swap operations you can reduce, the noise goes down by that much. It might not be that simple. But [it’s] a huge difference. It might be even more than just saying, “oh you know, half the number of swap operations, you get half the amount of noise.” It's probably even exponential. And then of course, with the new chip, the qubits themselves are better, meaning that the technology under them is more robust. The so-called lifetimes of the qubits are longer. It's not only that there's a better layout, just the technology itself is better.

So, this is a 120 qubit processor. How does that impact efficiency at all, in other senses? Or  how quickly it runs? Or does that depend on the problem as well?

You mean like how fast a job runs on it? I think, not the number of qubits does. Without getting too technical, in each of these processors, there's something called a repetition time. And you can imagine like as you run a circuit on the quantum computer,  each qubit has gates applied to it, which is just the shining of microwaves onto these chips. There's nothing preventing you from performing all the operations extremely quickly. But the qubits might not have enough time to be reset. And hence the result might not make any sense. It's almost like if you have to use the same paper with a pencil, you have to give time to erase the previous marks before you write on it again or it's just going to look like junk. If you do it too quickly, it's like you have no time to erase. And so you're just writing on top of the old writing and the result is just noise. It's just messy. So it turns out that the Nighthawk chip actually at the moment has a longer [repetition] time so, by default, and you can change this by default to get good quality results, it actually waits a longer amount of time before it applies the next operation. This is just a really complicated way to say that by default, it actually takes four times longer to run a job, once it gets to the quantum computer itself. There's a ton of steps before the actual circuit is run. Not only just waiting in a queue if you have to, but it has to compile and then transpile. And then finally, the running time itself is actually not that big of the whole amount of time. But the good news is that reset time is going to drastically shorten very soon. And fundamentally, this reset time doesn't matter at the moment because the qubit quality is better, the layout is better, so we can get better results even though it actually takes longer to run a job.

I see.  Why did you say that the reset time wouldn't matter in the future?

Well, on the cloud, IBM has just released  an upgrade to the Nighthawk chip. It's called r2, revision 2. And we have access to it, number one. It also means that there is the potential that our chip is upgraded to an r2 at some point.

And when do you expect that will happen?

I don't know if it's going to happen, but if it does happen, I'd expect it within less than a year.

Okay. I was hearing from a bunch of other faculty in a few different departments … there was some confusion about RPI acquiring a second quantum computer somewhere?

Right, right.

Is that this?

No, no. That's not this. It’s just unclear, in the sense that these things aren’t cheap? So even beyond what we have now, if we get a second computer, is that something that RPI needs to fund all by itself? Or do we work with the state and then allow other institutions to maybe access it? I don't know if buying time is the correct way to phrase it, but can we do this as a collaborative effort? Kind of like how Empire AI or these other massive classical high-performance computing centers are run. And so how much does the state of New York help pay for this as well if it's a SUNY collaboration? So there's a lot of details to be worked out. I think we'll absolutely have access to a next-generation system at some point in the next couple of years, for sure. And where does it go? We don't want to replace the System 1 that we have here. So does it go at the tech park? Or does it go somewhere else? Because one of the things that's unique about RPI is that we're very fortunate to have a quantum computer, but we also have a supercomputer at the tech park. And that's going to be hopefully upgraded very soon as well. There's a direct link between the two, a direct fiber, so you don't have to go through the cloud. I mean, most of our jobs you don't have to, but I should say that even when you access our computer right now, it's through the cloud. But there is a way to access it directly between the supercomputer and the quantum computer. But in the future, you could imagine maybe you just build a system where they're just integrated together. Where do you put that system? What [are] its power needs and stuff? So this is a question, and this is why probably other people are confused about it, because it's not been settled yet. But it just depends on who our partners are, what amount of funding is available. And then you can decide what system you're able to provide and then where to put it and all this stuff. So that's the next step. But I should just say that we're constantly thinking about what is the next step. I think you have to. 

That's actually a very interesting point, because last semester we got to interview President Schmidt, and he spoke about his [Department of Energy] appointment. And under Dario Gil, who is the ex-head of research at IBM, he spoke a bit about the Genesis mission to consolidate data from national labs and research done at those labs, combined with quantum research and AI-assisted methods, to streamline and accelerate scientific discovery, versus competing with the private sector. So where do you think RPI fits into that right now?

Yeah, that's a great question. I think we don't and we should not compete with the private sector. I think we should work with them. And in fact, we do have some partnerships with the private sector. I think it's mutually beneficial. One that I'm involved with is Western Digital. They're interested in a number of questions. We have a setup where they funded a student to work with me. We're working on implementing, we're trying to look at basically cryptographic type questions on the quantum hardware and HPC. But it's really to set up the infrastructure to train students. {it’s] a pipeline such that they can learn here the knowledge base needed to pursue problems that not only academia is interested in, but is relevant for industry. So that's why I think we're not competing at all. We need to actually collaborate as much as possible. It's really hard to find talent right now,so if you can find them, train them, and then get them where they're needed, I think that's ideal. And same with government. We have potential collaborations with agencies, similar type problem. I think it's crucial that we're all kind of moving together. Because not only does it benefit all of us individually, but it benefits the national interests and security and benefits the economy. I think it's just one of these: a rising tide lifts all boats. 

Thank you. I'm going to backtrack a bit. In and around RPI, because of the DOE appointment, and because of the quantum system one in 2021, there's been a lot of public hype around quantum utility. Your perspective actually working and, running and testing your theories on this hardware. What is your analysis of using quantum computer versus the classical supercomputer?

I don't think it's just one versus the other. I think what quantum advantage, whatever you want to call it, is going to be using both of them to go farther than you could with either alone. And I think a great analogy is just how, you know, we have CPUs and GPUs, right? Now they work both together. They're both sometimes on the same chip, definitely in the same box or laptop at this point. And, ideally in the future, it's going to be the same thing. You're going to have your CPU, your GPU, and your QPU. And now the QPU might not be able to fit into the laptop. What I mean by that is you're going to have algorithms or problems that are going to be solved. They're going to have mostly a classical CPU or GPU component to it. And then when it gets to a really tough part that you do need, like a quantum computer to make headway in, that's where the QPU is going to be very useful. And so it's critical to have all of them because I think it's not a competition. In a different context, it's not a competition between classical and quantum. It's more figuring out how we can use quantum as an accelerator to get to better computational algorithms. I actually just like to call it computational advantage, not quantum advantage. Because I think quantum advantage came from  Shor's algorithm and a few others, but Shor's is definitely the most impactful or useful. There is a mathematically proven speed up, right, that you need quantum to do. But most of the algorithms are not necessarily some mathematically proven exponential speed up. In fact, they're going to be kind of murkier. They're still going to be useful, but it's going to be just again, taking your classical algorithm and modifying it to allow some quantum in there. And then together, you're going to be able to solve a problem to higher accuracy or larger system size or in shorter runtime than you could just with classical alone.

There was an article that I read about a problem that a quantum algorithm had not previously solved before, but using tensor and matrix algebra, with a higher dimensional matrix, it had just been solved. I don't remember the problem. I'll leave [out] the specifics of it, but it was something interesting. Do you know what I'm talking about?

I don't know. So it was solved with just a classical tensor network method? Yeah I think that the tensor network methods are extremely powerful. I think a lot of the questions that quantum hopes to tackle, tensor networks are very able to solve them. Now, there is a point where they can't solve them anymore. So it's a little bit technical, but let me just say it like this. Any problem that you can formulate as a one-dimensional problem, a tensor network is going to be able to solve it efficiently. But when you start getting problems that are two-dimensional or three-dimensional, things like  —how a tensor network works, this thing called the bond dimension in a tensor network is going to explode exponentially. There's a provable point, there's going to be a point where the tensor network cannot simulate a system of a certain size no matter what. But the neat thing is, though,from an abstract mathematical perspective, a quantum computer and a tensor network, you can almost think from, again, just an abstract mathematical method, they're the same thing.

Adopting or using tensor network methods, you can actually make progress on quantum computation calculations. You can actually use a tensor network to almost warm start your problem. Basically, do a calculation up to a point and then let the quantum computer do the rest of it. This would be an example of this hybrid classical quantum method that we're trying to look at right now using our supercomputer and quantum computer together.

Along the direct line of access that they have.

Exactly. But at the moment, tensor networks are very powerful if you say you can do something with quantum that you can't do with classical, the tensor network people love to come out and prove you wrong. I think for a one-dimensional problem, they'll always be able to do that. I mean, up to a point, but there is a point where you can do a calculation and say, hey, there's a system of this size, you could take, you would need as much memory, classical memory, as fills the observable universe to simulate this system. So there is a point where you fundamentally can't do it anymore. And we're going to get there within five years at least. If not less.

Yeah. Maybe on a slight parting note. There's a lot of physics and computer science and engineering undergrads who really want to get involved with a quantum computer.And are trying to understand how it works, but they don't know where to get started.

Yeah. The 4000 level classes.

Sure. What do you recommend? What's a good resource to actually understand what's happening?

So I think I should point out that RPI does have a quantum minor right now.

Yes. Right.

And so first off, you can just look at the requirements for that. And we've deliberately set it up so that there are some 4000 level classes.  But there's also classes before that. And so if you really want to, you get to graduate with a quantum minor. And that could be very beneficial for employment. But like,  just look at that minor. There's like four classes. If you just take them in order, that can set you up to get to those 4000 level classes. But if you're just curious and want to learn more about the quantum computer, you can. So IBM Qiskit, like has amazing resources.

Yeah.

So they have a YouTube channel, they have textbooks, they have everything you could want. I think that's a great place to start. And then if you're part of the RPI community, you have an RPI email, you just send a request in. It'll go to my colleague. We'll just make sure everything checks out. And then you can get access to the computer. And so that way, you can start learning about what quantum computing is. You can run small programs on the computer.

You can run simulators and stuff too. And so then you can see the difference between a simulated job and the actual job. Although the hardware is getting so good now that we used to do that and used to be able to see the noise very easily, now it's a lot more difficult. I do think there's a lot of resources out there. And don't get overwhelmed with the 4000-level classes and all the math behind it. At first, I think you can just understand how revolutionary this is just by looking at some of the Qiskit stuff. And if it does interest you, then yes,  definitely learn the theory behind it. You know, take these classes. It'll enable you to do research in it and understand it deeper. Because I think why people struggle with quantum mechanics and just quantum in general is it's not intuitive. We don't interact on a daily basis with it.

It can really only be understood with math. The math, though, I should say is just linear algebra. It's just simple. So if you can understand linear algebra, you'll understand quantum mechanics at a much deeper level. But you don't necessarily need to know that at first.

Does anyone ever truly understand quantum mechanics?

I don't think so. I think it's just foreign to our daily experience.

Yeah. I just wanted to know, I was reading there was this Canadian company, Xanadu. Has RPI had any relationship with them?

Not Xanadu, no. We obviously have a great relationship with IBM. That doesn't mean we only need to have a relationship with IBM for sure. We do have people on the experimental side, like Xiangyi [Meng] in the physics department, and material science department. Like, this could be an interesting project for the more hands-on people out there. There's nothing stopping anyone from making their own qubits right now. And in fact, we have a relationship with GlobalFoundries. There's actually other places, I forget where it is, but they will make quantum qubits if you send it to them, right? Now, the difficulty is you need to take your chip and you need to cool it down and you need to be able to read in and read out. So all this electronics behind interpreting what's going on and how to control it, that's not taken care of. But to make an actual qubit, it's a quite simple thing. One thing people can look at is coming up with new modalities or new systems for quantum computers. Xanadu, or Quantinium, or IBM, they all have different approaches to building a quantum computer, but that’s just their approach. There’s many other ways to do it. It’s not easy, that’s why there’s only a handful of companies, but if you have a clever idea, it might not be as difficult as it seems.