At the Quantum World Congress (QWC) in mid-September, trapped ion quantum computing pioneer Quantinuum will unveil more about its expanding roadmap. Its current state of the art system, H2-1, has 56 qubits, and is no longer “classically simulatable,” says the company. We’ll learn more about its next-gen systems at QWC. Over the last year or so, Quantinuum has steadily rolled out science advances, architecture improvements, and global collaborations. And, of course, there was the recent report that Quantinuum is planning an IPO.

Among the company’s recent highlights are impressive work on error correction with Microsoft, work on proving QAOA speed-up with JPMorganChase and Argonne National Laboratory, establishing global collaborations with RIKEN, Singapore, STFC Hartree Center, and release of Nexus, a development and workflow platform for gate-based quantum systems. Perhaps the core step forward was development of Quantinuum’s quantum charge-coupled device (QCCD) architecture, which along with emerging optical-interconnect schemes, should enable scaling up system size.

These are heady times for Quantinuum.

CEO Rajeeb Hazra told HPCwire, “So I’m going to put a little teaser out there today. While we do the physical scaling, we are also looking at more defined paths of getting to fault tolerance right and even improving error resiliency in the NISQ era, and we now even have a vision of how we want to implement fully fault-tolerant quantum computing. I won’t divulge the details of the roadmap, but we are now within a few weeks of being able to talk about it with much more specifics.”

Are we nearing a break-out moment for Quantinuum?  Time and the market, of course, will be the final arbiter.

Hazra says Quantinuum is transitioning from the lengthy discovery stage — a decade during which the “ground truth” of much of its underlying technology was hammered out — to an acceleration stage during which quantum computing becomes a key ingredient of the next business cycle. This next business cycle, says Hazra, will involve the alignment of classical compute, Gen-AI, and quantum. More than big machines acting alone, he says, the next cycle is about new workflows and broader business practices.

What are the key elements of Quantinuum’s plan to succeed?

“[The] first is we have conviction that our H series roadmap is the winning path, and we have the recipes now in place, and that’s why we are accelerating,” said Hazra. “[The second is] we are accelerating the commercial viability by accelerating the roadmap, and we are doing that with very key partnerships. In Singapore, for example, it’s all about computational biology and we have some very specific outcomes that we believe are no longer [expected] in 2035 but by the end of the decade, again, based on acceleration of hardware and software. The third is that as we do this you will see Quantinuum far more visible, and you’ll hear from us more often than you’ve ever done before when while we were in the discovery phase.

“I think the other part we could have a longer chat about is what’s the relationship of AI and quantum, right? We have a very unique point of view on classical AI and how quantum is relevant to it.”

Intrigued?

Hazra shared his view of Quantinuum’s transition from discovery to acceleration during a lengthy interview with HPCwire last week. While the conversation steered clear of technical granularity, it broadly covered Hazra’s ambitious vision of Quantinuum’s roadmap to success in what’s still a very competitive quantum computing landscape. Portions of the conversation (lightly edited) are presented below.

By way of background: Quantinuum was formed in 2021 when Honeywell’s Honeywell Quantum Solutions unit was merged with the U.K.-based start-up Cambridge Quantum Computing. It has been a leader in developing trapped ion technology. Honeywell owns a 54 percent stake in the company. IBM is also an investor. Hazra — a longtime electronics industry senior executive, including at Intel — joined as CEO in February of 2023. Ilyas Khan, then CEO, remains as chief product officer and vice chair of the board.

If you’re a quantum watcher, you know there’s a vigorous ‘battle’ going on between competing qubit modalities. In essence, we don’t really know yet what the ‘quantum transistor’ will be. Among the contenders are superconducting qubits, silicon dots, trapped ions, neutral atoms, photonic, diamond (nitrogen-vacancy), and others. They all have strengths and weaknesses (operating environment, speed, control electronics, susceptibility to manufacturing inhomogeneity, etc.). Quantinuum has bet big on trapped ions, known for inherent high fidelity but sometimes knocked on scalability and speed issues. Hazra touches briefly on both.

HPCwire: I was hoping you could provide us with a snapshot of the company vision for where quantum is heading.

Rajeeb Hazra: I’ll start by saying what you are seeing in the last six months is that we are accelerating. If you think about it, even before we were formed as Quantinuum, but even on the hardware side, and Cambridge quantum as a software entity looking to codesign with hardware, which is really where we came together. There was a period of discovery. Honeywell had gone through a soul searching over what modality to pick. It wasn’t just somehow we all grew up thinking trapped ions, right? Cambridge was also looking at what are the best modalities. Industry was slowly trying figure out what quantum was and tying to figure out who Quantinuum was. It was kind of an age of discovery for the last 10 years or so, including the last two or three at Quantinuum.

That’s largely over for Quantinuum. For us, very specifically, the change you see in our external persona and the kind of the voice we have now is because we are now accelerating through all of that discovery.

We have deep conviction in trapped ions as a very important, if not the only modality. I share your views that it’s too early to say there’s going to be only one modality, but we believe trapped ion is going to be a very important part of the modality set that survives. We have conviction in our scaling [capability] and true to our nature, we publish things instead of just claiming that we can get to 1000 logical qubits or something like that. So the whole wiring and sorting problem paper kind of tells you why we think we have the solution, and then we embody it in our products — Helios H3, H4, systems that increasingly scale without compromising fidelity. We could have built a lot of trapped ion injection (capability) into the trap and not managed decoherence error, memory error, crosstalk error, state preparation error, and then those systems would be largely useless.

HPCwire: So you’re changing gears after roughly a decade?

Rajeeb Hazra: Yes. We’ve got acceleration of our roadmap. We now believe that we are well on a path to having commercial systems that scale to tipping points where large commercial applications or discovery becomes possible. This was the first year the H2 the system we have is no longer classically simulatable. So running the system, you can get quantum information out that you could you could never get from simulating a classical system. So when you take that one step forward and say, ‘now I’m going to use that data coming out to model a quantum phenomena,’ that’s data you could have never had before, and there’s no going back.

Along with that we’re accelerating what I would call platform orchestration elements like Nexus, building up the developer platform, and then maturing what we would call domain-specific — I don’t think of them as application, even though we say the word application — platforms or discovery platforms like Quantinuum’s inQuanto for chemistry. We have now a very integrated view of how our hardware and software can accelerate based on the outcomes of discovery that have told us how we want to build the hardware, how we want to scale it, how do we develop very successful error detection and correction codes to further improve the fidelity into logical qubits.

The other place you’ll see acceleration is in partnerships, and you’re seeing that already. We announced, you know, we’ve got seminal work going on with Microsoft (error correction), and watch the space in the near future as well. We announced major partnerships with sovereign countries, Singapore, UK, and there’ll be more to come in the rest of the year. Why? Not because we’re just out there trying to get MOUs in place, because these sovereign countries, as they energize and invest in their quantum ecosystem, are looking for the best partner that includes not just the best machine, but also the capabilities we bring, not just on the software side, the research side, but even sometimes the partnership side, right. So you’ll see an acceleration of that work, because that’s important for the acceleration to commercialization, because those partnerships help us define, refine, and accelerate use cases.

HPCwire: Does the increasing collaboration mean getting more involved industry issues?

Rajeeb Hazra: Yes, you’ll see us be far more aggressive as leaders in establishing our footprint on the leadership debates of the day, whether it’s new benchmarks, whether it’s debate around what is actually a logical qubit versus not. We’ve been in the discovery phase, more focused on getting to ground truth ourselves, and just creating the machines with the ground truth in it, you will see us also take a much more proactive position on leading the industry through what’s called the, you know, the hum and drum noise into these critical elements. Why? We cannot afford that same level of confusion [that exists now] because now we believe we’ve taken quantum into the relevance of the next business cycle. Now you need to know the truth, because quantum is intercepting your next three-to-five year business cycle.

HPCwire: What do you mean by the next business cycle and how will quantum intersect it?

Rajeeb Hazra: It’s the simple case of three things coming together at the right time — high performance computing, the classical side, and what I would call generative AI, and quantum computing, these three coming together. It’s not just, you know, the age old we are just going to achieve quantum advantage or quantum supremacy, right? It is about real workflows, like how would you design a new drug? Where you need the classical compute, you need AI agents to do the work of what I would call the digital lab of experts, and you need quantum computing to actually give you the data at chemical accuracy, which you would never have even if you had an army of super fast lab researchers. Those three things coming together to a tipping point is what has made it absolutely relevant in the current business side. Companies are planning for how they take, for instance, discovery, whether it’s materials chemicals, or how do they take supply chain planning with these three things [HPC-AI-Quantum] working together into a new era.

It’s just like many years ago; if you weren’t doing computing, you were losing out. To compete is to compute. I’ve stood on stage in HPC events and said this a trillion times. Then there was the, “If you don’t have a cloud strategy, you’re not going to have a scalable compute infrastructure.” Now it’s if you don’t have the next generation compute strategy around AI, classical computing, and quantum, you’re not going to be able to compete.

So that is why we are within the next business cycle. As you’ll see in a short time, we are within levels of capability that can actually meaningfully accelerate process of chemical discovery. I mean, if you read the blog that Microsoft wrote with us on the work that we’ve been doing; it isn’t like it’s suddenly there’s one application runs 10,000 times faster. No, that’s not what quantum is about. It’s about now you have a different way of discovering new chemicals or creating new chemicals or optimizing them that you couldn’t have done before, and that has tremendous economic value. You could send it out (out source), but then someone else is going to figure it out, and you’re going to be at a disadvantage. That is why it’s relevant in the current business cycle.

HPCwire: When do you think you’ll have a collaborator, a customer who will have done exactly that, and that you could point to it as a case history?

Rajeeb Hazra: That’s a great question. We’ve been working with the biggest names for the last five years, even before it was Quantinuum. Some of them, we’ve made very public. Some of them don’t want to be public. Ones we made public, BMW, Airbus, Total on chemistry. [The] BMW Airbus work was for hydrogen as fuel. We work with fertilizer companies on food security and improving methane production. I almost have that chart [of collaborators] in my head.

HPCwire: Is it safe to infer that some of those companies are now or on the brink of using this technology in production?

Rajeeb Hazra: I want to make sure I’m not misleading you. Because what companies are doing is saying, “What is the quantum algorithm? What is the use case? And how do I know that it actually is producing the output we want? That’s phase one. Then at scale, can I apply the problem at scale? We are kind of in that second phase. Then it is, “How do I introduce it into my production workflow? Now, production means building something, right? If you take a look at what Microsoft has announced, it’s like companies that are doing new drugs, building essentially new drug molecules, are going to have to go through a phase of discovery that is part of the workflow. And when it goes into production, means that is the workflow that is being used.

It’s not like in classical [HPC], let’s say BP is doing discovery upstream; they use real-time, full wave inversion or reverse time migration, and have a bank of computers sitting there running [and can say] “Okay, we’ve now installed them and that is what we are using.” This is a new workflow. In many cases, it’s a new workflow that is part of the upstream that says we are going to do this differently, the discovery of new molecules, and we are starting now to perfect and make the algorithms, number one, work and number two, [work at] scale. [Making the algorithm] work [was] like the last couple of years. Scaling is now what’s happening and that is why we have to accelerate the roadmap. Otherwise we wouldn’t just build more and more powerful quantum computers. We have very specific feedback through these partnerships on what scale is needed and at what scale you can do this at a production level, etc.. That is why it’s in the business cycle, because if there wasn’t a production horizon to it, people would be like, ‘Okay, sure but it’s just research”.

HPCwire: So we’re starting the next business cycle, which is the next 3-to-5 years?

Rajeeb Hazra: Yes. You will absolutely see those tipping points and people saying, here’s how we have used quantum computing to change the way we are running our business, whether it’s the business of creating something or it’s an internal supply chain management or securing the enterprise through Quantum hardening.

HPCwire: Quantinuum has bet on trapped ion. Neutral atoms have recently gained more prominence. IBM and Google (superconducting qubits) have been in the game forever.  It seems like there a lot of activity on very many modalities. What’s your take on winners and losers in the qubit wars?

Rajeeb Hazra: I will just tell you, this is my strong view and getting stronger every day — in the next three years or five years, you will see the haves and have-nots. You will see people who are still trying to build their first machine and make it work for real applications or real algorithms, versus people who are doing it and customers are using it and saying, here’s the result we’re getting. So I think for neutral atoms, and PsiQuantum and some of the companies that are working on the photonic modality, there’s been bold claims being made, but they’re yet to build their first machine that customers are using and paying for. We are on our third generation of that machine, of that cycle, with customers.

HPCwire: Quantinuum’s high Quantum Value (QV) benchmark score (262,144) has attracted attention. Even IBM is impressed. 

Rajeeb Hazra: The reason our results are where they are versus others is because we’ve gone through multiple cycles of understanding what the key challenges are, coming up with solutions that are scalable on that results in our physical machine being high enough fidelity that we can get stunning results like the one Microsoft did with error detection and correction, even in the NISQ era, right? I mean, there Microsoft statement was, we have now entered the reliability of Level 2.

HPCwire: Talk about Quantinuum’s bet on trapped ion technology (see Quantinuum summary below).

Rajeeb Hazra: What trapped iron has going for it is versus a superconducting is that nature creates every iron identical. The superconducting guys actually have to create a qubit. So we start with an advantage, and you see that today, where the physical qubits for superconducting are hard to do and still have low enough fidelities that it is hard to cross the threshold the way you can do detection and correction. So they’re still using [error] mitigation, which is a very expensive way of actually getting correct qubits. The second thing is, we have worked really, really hard. It’s not like we haven’t had our own trials and tribulations, but we have now the solution to scaling trapped ions without losing fidelity, actually, while improving fidelity.

HPCwire: What about the slow speed of trapped ions, relatively-speaking?

Rajeeb Hazra: Speed is important, but in the right context. Time to solution is important. So. If you give me a blazingly fast shot, but it gives me the wrong answers 80% of the time, your time distribution is still worse, right?

HPCwire: So you have higher reliability, and perhaps fewer shots which, effectively translates into improved time to solution. Let’s switch gears for a moment. What your sense of the U.S. government role in making quantum systems broadly available?

Rajeeb Hazra: We’ve worked for a long time work with the Department of Energy through Oak Ridge National Lab  (Oak Ridge Quantum Computing User Access Program (QCUP)) and in my mind, it has created a very efficient system for researchers to use quantum machines, including ours. So that’s a large number of folks. We’re also working with the other side of DOE on what their needs are into the future. So we are very well engaged. If there is a point of frustration on my part, it is that I think, of all the geographies we are engaged in, the US is still under investing in kind of quantum and getting quantum capability for open science. Let’s not talk about the dark side, but at least for open science.

HPCwire: There’s an observation out there that the U.S is trailing, for example, the UK through its testbed program, and Europe through things like the Munich Quantum Valley program, in which other countries are investing in bringing in a wide variety of devices — regardless of where they’re built — and then learning how to use them so they can get to practical use of these devices, no matter where they come from.

Rajeeb Hazra: I think the US could definitely do more. I won’t say they are not doing that, but we could do more. I think the hangover a little bit for us has been in the U.S. we looked at Quantum through a very cryptographic lens, right, as a threat, and what do you do? And that was necessary, by the way. The work NIST just did is amazing, and it’s going to actually help post quantum cryptography become real. But that isn’t the only lens, right? There’s all of these other things and it’s not putting quantum tech computing in the right balance with AI, putting quantum computing in the right balance with quantum sensing, so you can have quantum sensors be the data sources, and quantum computing be the way you actually do very large scale computational exercises.

I feel like there’s more to be done, and there has to be coherence between all the different quantum networking and on, not just how you secure large scale networks, but how do you scale up larger and larger machines. I still feel like the entire quantum computing industry is trying to build a 10 gigahertz processor, and then the 50 gigahertz processor; it still scaling up a single machine getting and that’s going to be necessary, but at some point it will be [require] a scale-out solution, because that will scale more easily than trying to make a single quantum computer

HPCwire: We hearing more about quantum and AI coming together but without a lot of detail.  What’s your take?

Rajeeb Hazra: [We] could have a longer chat about is what’s the relationship of AI and quantum, right? We have a very unique point of view on classical AI and how quantum is relevant to it. But we also have a longer term view of what is the right generative AI technology of the future. Is it still transformer based approaches to manipulating training and manipulating neural networks, or is it a whole different way to represent what Hidary (Jack Hidary, CEO SandboxAQ) calls large quantitative model, not just language models, but large quantitative models. And we have a very, very active research and we’ll be publishing a few things over the next, I would say, six months or so.

When today’s AI, as successful as it is, when today’s generative AI technology starts running into sustainability challenges or traceability challenges or interpretability challenges, what’s after this? And how is quantum relevant? We believe it is highly relevant, and we believe what we are doing is leading the path for the next phase. This is kind of like saying in the height of summer, “What happens when winter arrives?” Most people don’t pay attention to it, because they’re still enjoying summer, but we see large challenges on the horizon.

These data centers are getting impossibly large to power. Maybe space is available. It’s not a very sustainable solution, and the black box nature of generative AI is making it hard to get into regulated industries and core free enterprises. It’s great for conversational interfaces and [to] give me suggestions. But would you replace your department that does your portfolio management in a large bank? Or deep hedging [with] a generative AI agent that you cannot explain in a regulated environment how it came to certain solutions? No. It’s a responsible AI or interpretable AI — that is something we are deeply, deeply working on, and we have some papers coming on it.

HPCwire: It seems like more quantum companies are building computers intended to be deployed on collaborator and customer premises.

Rajeeb Hazra: We are seeing a lot of interesting demand for our systems in various geographies. We are now scaling our ability to manufacture these systems, and rather than build these as research driven systems that then get commercially used, we are building this so that we can go land it in Germany or UK or Singapore, and build these systems in a way that they can actually become the center point of you know, it’s not like everything has to be handled either in Broomfield CO (Quantinuum’s headquarters). Then we can make multiple copies of it as a business.

HPCwire: Raj, thank you for your time.

Link to QCCD paper (Scalable Multispecies Ion Transport in a Grid Based Surface-Electrode Trap),  https://arxiv.org/abs/2403.00756