The Quantum Control Stack with Niels Bultink

Sebastian Hassinger (00:00.876)
Hi, Niels. Thanks for joining me. I really appreciate your time. Tell me a bit about to start, what was the inspiration for starting QBlocks? What was the origin story of the company?

Niels (00:16.401)
Yes, hi Sebastian, great to be here. Thanks for having me. So let me go back a few years indeed. So I founded QBlocks six years ago together with one co-founder called Juven Ove. He's a long friend of mine. We go back about 20 years now. And yeah, I have to say, since I've met

This co-founder Jules, we've always been thinking, tinkering about starting a company one day, both going on our individual tracks, met at university, but going into various businesses individually. So I've been working in the semiconductor industry. He was building seismic sensors for oil exploration entirely different fields.

Sebastian Hassinger (01:05.966)
Hmm.

Sebastian Hassinger (01:13.998)
Mm-hmm.

Niels (01:14.501)
But after a couple of years working, I wanted to go back to school. I missed quantum engineering, quantum technology, and I went back for a PhD. So relatively late in my education and career. Also with the mindset of developing a business out of it. So after about two years, I was doing very fundamental work on quantum error correction.

Sebastian Hassinger (01:37.838)
Hmm.

Niels (01:43.933)
making quantum computers that can correct their own errors and doing the first initial demonstrations of that. And for those demonstrations, we needed new innovative control solutions because quantum error correction by nature means measuring and correcting. So that's a fundamental control problem. So the engineers around at the university, but also at the governmental labs here,

Sebastian Hassinger (01:48.302)
Hmm.

Sebastian Hassinger (01:58.926)
Hmm.

Sebastian Hassinger (02:04.312)
Right.

Niels (02:12.699)
they were building these sort of prototype control devices to make my research work. And that's where the initial idea for QBlock sparked. If we need this now for our scientific breakthroughs, the entire field and the development of quantum computing will need these type of control systems to move forward. So just using that advantage, that initial head start,

Sebastian Hassinger (02:17.39)
brain.

Niels (02:40.817)
and building a company out of it. So that's how we decided to go for it and build QBlocks and run with it. Correct. So you're based in Delft, the Netherlands. Yes. Correct. Yeah.

Sebastian Hassinger (02:46.658)
And that was a to you delft, right? Yeah.

Yeah. And was that, that you went back to get your PhD with the idea of, finding sort of a commercial path. Did that, was that inspired by, was quantum Delta already, in existence or did quantum Delta happen while you were doing your PhD?

Niels (03:12.593)
The Quantum Delta as a fund came on a bit later, but we were working in an institute called QTAC that exists very, very pioneering, always top three, top five in terms of scientific output. And they were also very early on in seeing quantum, not just a science, but bringing in lot of engineering disciplines. So microwave engineering, cryogenic engineering to

Sebastian Hassinger (03:22.071)
Yeah.

Sebastian Hassinger (03:26.327)
Right.

Sebastian Hassinger (03:37.165)
Right.

Niels (03:41.949)
to build those next scientific milestones.

Sebastian Hassinger (03:45.964)
Right. Interesting. so, I what was the, you know, the, in the experimental physics lab, there are, you know, whatever AWGs and all sorts of, of control, electronics for experiments. And I mean, you know, when you're not talking about a commercial setting, people sort of cobbled together control systems out of existing, units. What, were the,

the limitations that you saw that sort of spurred the need for a custom built specific qubit control kind of device.

Niels (04:23.741)
Yeah, one of the most funny and typical examples I always find the IBM quantum experience, so the first quantum computer that was live in the cloud. So it allowed people for the first time to drag and drop quantum operations on screen and just with the very simple interface control a quantum computer. It becomes funny if you look under the hood behind the kind of flashy gooey, what the first control stack looked like. So that simple

Sebastian Hassinger (04:39.224)
Mm-hmm.

Niels (04:53.489)
web interface translates into code and that code translates into programs for these very large scientific equipment pieces. AWGs by companies like Tectronix, Keysight, measurement equipment that is not at all tailored for quantum computing. It was just built for general purpose lab purposes. And those first quantum experiments and even the IBM quantum experience

experience was built up of those devices. And you would need about one meter or three feet per qubit of shelf space, of rack space to control it. So the five qubit computer just had six meters of equipment to do it. And yeah, it's just size, cost, speed of operation, the bugginess of those approaches. It's very easy to see that it's not at all scalable towards the.

Sebastian Hassinger (05:32.065)
You

Sebastian Hassinger (05:37.838)
Yeah

Sebastian Hassinger (05:45.848)
Yeah.

Niels (05:50.981)
Even the systems that we have now with hundreds of quantum bits, but especially if you look ahead, where we want to go in the next few years towards tens of thousands of qubits, you really need this hardcore engineering to bring down scale calls to manage the complexity and speeds that are needed to make it all work together.

Sebastian Hassinger (06:00.856)
Right.

Yeah.

Sebastian Hassinger (06:15.406)
That makes sense. so, you I mean, you performed IQ tech or what's described as the first sort of feedback on solid state qubits in, I think, 2012. And you the the cube blocks platform is purposefully qubit agnostic. How difficult is it to create to build a control system that has that sort of agnostic

I mean, in other words, how similar are the different modalities of qubits or what, you know, how do you, how do you maintain that flexibility?

Niels (06:52.763)
Yeah, so it's somewhat of a nuanced answer, but it's surprisingly easy to make a control system valid, applicable to many qubit types. Well, if you look at it, especially from the digital side, so the process or architecture that we've built, and how it sort of sequences operations, sequences quantum operations, and does measurements.

how it relays information throughout the system. That's the bulk of our engineering, the bulk of our AP. That is almost 100 % mappable to every qubit type. Where it varies per modality is the frequencies that you need to address them. Some of the qubits operate at microwave frequencies, some of them at optical frequencies. So you need different sort of analog techniques to...

to get from the digital framework towards those qubits. But the bulk of the system really is very widely applicable. And that's also the reason for us to exist in this space and the greatness of why a value chain is very advantageous to this field. sometimes companies like us are called picks and shovels. It's a nice analogy.

but it doesn't hold entirely. I would more look at it this way. A quantum computer is much more than just qubits and qubits are the most hard part to develop. But if you look at the stack that is required, the qubits are just the bottom layer of it. And all the other layers of the stack are also crucial to develop. So I think most of the companies and big companies and small companies are really focused on engineering those qubits.

But it's very important that those other layers of the stack also reach scale, reach maturity. And that's what companies like us do. And often at those other layers, you can also be much more qubit agnostic. So it's also advantageous to the field to do it as a value change, much more efficient that we do this for the hundreds of clients that we have instead of everyone doing it themselves. It's just a matter of efficiency for the entire field.

Sebastian Hassinger (08:59.112)
Mm. Right.

Sebastian Hassinger (09:08.718)
Yeah, that makes sense. the so what you're saying is a cross modality send it's still very similar to what you described with the IBM quantum experience. It's still a classical code that gets translated into a series of signals in whatever frequency range is appropriate for for the particular qubits sent down to the qubits and then and then there's sensing or a feedback loop that that

Set a signal comes back up from the qubits and you interpret that and then turn it back into classical data essentially, right? That's that's essentially the loop that has to happen continuously.

Niels (09:48.893)
Yeah, well, apart from the size and the cost perspective, there's quite a fundamental difference of how we have built up quantum control versus how it was back in the day. And that has to do with the rigidity of the system early on. So those early systems, would sort of precompile the whole quantum program. Everything that happens down to the nanoseconds has to be precompiled. So quite memory heavy.

Sebastian Hassinger (10:04.558)
Mmm.

Niels (10:16.293)
and slow approach because you have to upload gigabytes of data to your control system for a simple program. So heavy in terms of memory usage, but also slow. And there's no interactivity possible because everything is sort of predetermined, pre-compiled. If you look at our system and the way we operate Qubits, it's much more dynamic. So we have a process that in real time on the fly build up these

Sebastian Hassinger (10:16.471)
Mm-hmm.

Sebastian Hassinger (10:20.462)
Hmm.

Niels (10:44.701)
quantum programs build up these sequences of pulses. And also they can be adaptively programmed. in real time at microsecond time scale, measurement information can be used to adapt the control flow. And that sounds maybe flexible or nice, but that's at the core of the elements needed for quantum error correction, quantum fault tolerance. It requires you to continuously track errors in the system

Sebastian Hassinger (11:10.808)
Right.

Niels (11:14.461)
and correct them so that flexibility is not just a nice to have or a performance increase, but it's a basic need for this new era of quantum fault tolerance.

Sebastian Hassinger (11:21.441)
Right.

Sebastian Hassinger (11:26.646)
And that, I mean, that is critical for, I mean, that speed, if you don't have that speed and responsiveness, then your, whatever advantage your qubits are affording you is going to be burnt away by the classical loop for, for trying to correct it, right? If that's too slow, then there's no point in actually trying to the quantum computer. So it's a, it's a very, very crucial task that you're trying to take on. and, and is that.

I read a bit about the sync protocol. Is that part of what allows you to operate at that extremely fast speed?

Niels (12:05.019)
Yes, indeed. So we have two fundamental protocols heavily protected that are very fundamental to our processor architecture. They are called Sync and Link. Sync is about synchronizing all those processor cores. So we have about one per qubit. So hundreds of these cores currently operate at the same time. And the Sync protocol allows all those processors to be in sync.

down to the picosecond even. there's no, whenever you reboot the system, there are no indeterministic changes in the delays of the signals with respect to each other. And that's quite a difficult, but required a trait of the system to be able to scale. The other one is linked. That's about the exchange of information between those cores. So how

Sebastian Hassinger (12:35.128)
Hmm, wow.

Niels (12:59.517)
The behavior on qubit X can depend on the outcome of qubit Y anywhere in the system. that bandwidth and speed of how fast any qubit in the system can use information of another qubit, that's very unique in our platform. Like how fast and how many qubits you can connect in this way with each other.

Sebastian Hassinger (13:21.646)
Hmm. And those are, those chorus are FPGAs, is that right?

Niels (13:27.229)
Yeah, the cores, the processor cores are currently being implemented in our product on FPGAs. So FPGA is a sort of a flexible logic platform where you can put signal processing elements or you can put, but you can even put your own processor core on an FPGA. So that's currently how it's implemented, but.

Sebastian Hassinger (13:47.426)
Right. Right.

Niels (13:52.317)
For scalable systems, you will see the field moving away from that paradigm towards more hard-baked processors because already at the current production volumes that we have, it makes more sense to put things in chips in ASICs.

Sebastian Hassinger (14:08.056)
Custom silicon, yeah, yeah. That's what I was gonna ask. It seems like the FPGA is a very useful platform because of its flexibility, because of its programmability, but in order to get to higher scales, it's very good in this experimental stage of the industry where we're trying to figure out how to tackle these basic challenges of fidelity, coherence time, error correction, all the rest of it. But...

I assume that to move forward, you'll have to do something like an ASIC or like your own custom silicon to get that ratio down to one core per cubit is probably not sustainable when you get to millions of cubits. Hopefully we get to millions of cubits. Yeah, interesting. Yeah.

Niels (14:53.511)
Yeah, working hard on it. But I think exciting times for the field right now because the goalposts for quantum advantage have moved significantly closer over the past month. So many of the algorithms, one example is Shor's algorithm to break RSA cryptography. A year ago, we thought we would need

Sebastian Hassinger (15:07.436)
Yes.

Niels (15:22.461)
tens of millions of qubits for that, but the best estimates currently have gone down towards 10,000 to 100,000. So it's really a factor hundred smaller what we now think we need versus a year ago. And those numbers will likely even go down instead of up. So the large pie in the sky that was there the last three years has really become within reach right now. That's a really exciting times.

Sebastian Hassinger (15:31.022)
It's extraordinary. Yeah.

Yeah. Yeah.

Sebastian Hassinger (15:49.122)
Yeah. Yeah. That's really cool. So, so there's three things in the recent development of the company that I wanted to talk about. and, the first actually is, is related to that FPTA topic. It's, it's the, partnership with the kick, team or the kick board out of Fermilab. I've worked with that team before, Gustavo and Sho. it's a really interesting platform because it's, it's

Niels (16:12.509)
Great.

Sebastian Hassinger (16:17.932)
I almost, describe it as sort of being a Swiss army knife for a physics lab. It's, you know, it's a FPGA board with an IO, a custom IO board, which they've used for cubic control, also Gustavo was telling me about a dark matter detection experiment where they multiplexed, think it was a hundred thousand or something CCDs detectors for trying to sense dark matter at the bottom of a mine. it's a...

Niels (16:23.889)
Yeah.

Niels (16:43.015)
Mm-hmm.

Mm-hmm.

Sebastian Hassinger (16:46.262)
It's a really fascinating platform for scientific discovery. I'm curious what the motivation was behind that partnership.

Niels (16:56.701)
Yeah, good point. So I think a couple of years back when we started the company, it was very much an academic field, almost exclusively. And with the development of the company and the market, we now see sort of a split of that that market. Large part of our revenue of our customers are now business to business and are working on scaling up the systems. And

Sebastian Hassinger (17:06.808)
Mm-hmm.

Sebastian Hassinger (17:22.03)
Mm-hmm.

Niels (17:26.503)
the technology needed for that scaling challenge is diverging quite a bit from what platforms you need and want for academic research or even one level earlier, workforce education. And yeah, for those needs, developing the right workforce and having very flexible platforms even down to the FPGA code.

Sebastian Hassinger (17:43.214)
Hmm.

Niels (17:55.157)
Kik has been an amazing system and it has allowed many people in the field to do things that are not possible with off-the-shelf products like ours. Our products are used a lot in science and help a lot in scientific discovery also at a small scale, but it doesn't allow some of the flexibility that a platform like Kik has. We really like

Sebastian Hassinger (18:17.774)
Hmm.

Niels (18:20.049)
those projects, we really like open source. We have had a strong dedication to open source software already from our inception. We've been doing everything in software open source. So we liked helping that community and that project move forward because it has been run from Fermilab, of course, which is kind of a mismatch sometimes because you're doing customer support or support to the field.

And for them it became sort of not maintainable as a project. And for us, it's actually quite synergetic if you look at our current sort of split of the different markets to adopt a project like that.

Sebastian Hassinger (18:54.766)
Hmm.

I see. Interesting. Yeah, that's really interesting. So in other words, would address the academic and educational side of the market where ultimate flexibility and openness is actually the most important values. And then the QBlocks product line, the rest of the product line would address more the B2B, the commercial applications where

robustness and support and specific functionality is really, really important because everybody's trying to scale up. That makes a lot of sense. I love the board. It's really an extraordinarily powerful tool for exploration. So I was excited when you guys partnered with them. So that's great. And then along the same lines, the quantum utility block that you launched with Quantware and Qt control, that sort of

elements of the supply chain, you know, coming together in an open way. Tell me more about that. That's really fascinating.

Niels (20:03.303)
So it's a great example of showing, demonstrating to the field the value of open architecture of a value chain. we have, as a company, we supply products and those products always end up in a stack. And in that stack, connects to software layers above us, it connects to hardware layers below us. And sometimes,

for our customers, it can be challenging to integrate those different layers, whether they are doing in-house developed software engineering, or whether they're buying products from another vendor for those layers. So what this project is really for is to come together as companies and really work on that integration and understand how can we tailor the products that we have together in such a way that they work better together and to deliver more.

values to the customer, more performance, but also ease their integration work. And that's, I think, really a great collaboration with these two companies, QuantRack and QControl, to do that. And also quite successfully in what we have done over the past APS Mars meeting, so the biggest physics event, we have done many demonstrations where we show in a weekend, you can actually set up a full stack from zero.

And on day two of the conference, we could show live performance on qubits. We could actually measure data on those qubits that was used in talks during the same event. So it really is just a nice, really tangible showcase of what you can do with off-the-shelf products rather than inventing things from the ground up.

Sebastian Hassinger (21:38.638)
Amazing.

Sebastian Hassinger (21:46.786)
Yeah. And commercially, I mean, you described sort of the stack of vendors, the advantage of QBlock specializing in control systems instead of customers doing, QBlocks is doing that in house. But there is sort of, there's a risk in this market to that approach, which is a vertically integrated approach like IBM or Google or Quentinium or IonQ.

They're trying to build the entire thing in-house essentially. Do you see that as just a competitive threat for QBlocks or do you see that as a risk to the market development as a whole?

Niels (22:31.733)
Do I see it as a risk? I would say the trend is quite positive. The trend is more more companies realize we shouldn't do everything ourselves. And if you look at competitive developments, I would say that the speed at which QBlocks develops control stacks and the amount of people and the amount of specialists we can dedicate to it.

just makes it very hard or not wise for many companies to do it themselves. Makes it much cheaper and easier for those companies to buy from us and to focus on their core IP. If you look at classical computing, it's a huge value chain with so many specialisms everywhere down the stack. What we have done to isolate control is just quite a broad layer actually still.

Sebastian Hassinger (23:28.014)
Hmm.

Niels (23:28.593)
So I would actually expect the trend is going to be the other way around, where you will see more more specialization in different layers of the stack. yeah, the field needs that. That's my opinion.

Sebastian Hassinger (23:37.101)
Interesting.

Sebastian Hassinger (23:41.528)
That's really interesting. potentially even more, as you said, specialization or fragmentation of the stack from qubits all the way up to the classical interfaces. I mean, it makes sense. If you take the analogy of the classical information technology market, there were early contenders that were vertically integrated from the ground up, but then...

Certainly by the PC era, it was open systems and a richer and richer supply chain with more more participants. So that's interesting. It's going to be very, I mean, no matter what, it's going be very interesting to see how this all evolves. And so like in terms of evolution, the most recent sort of development for QBlocks is expansion into the US, right? was just earlier this month that your facility in Massachusetts

sort of went live and that's a manufacturing site. What's the significance of a US site to QBlocks and how does that factor into your strategy?

Niels (24:48.497)
Yeah, so we have been serving the American market from day one. So from six years ago, we started with it's just like NIST and many of the national labs started to buy from us. But we also noticed that it was hard to serve those customers in the same way how we can do it here locally in Europe. So that was a reason two years ago to start build up

team across North America. And it has been quite successful if you see how fast we could grow a team and also grow our market share in the field. And for us, a natural next step has always been to produce locally. think with geopolitical developments, it is seen as very important to be able to serve us, to be able to support.

old technology to be able to produce old technology locally. And that's for us the reason to make that step. So yeah, very, very good progress. this quarter or last quarter, we started with our first production batches made in the Boston area.

Sebastian Hassinger (26:06.67)
That's fantastic. That's fantastic. And, and, know, I guess the, the, the next question is basically, you know, you came out of Delft and Q-Tech. there's a very impressive, but yourself, I Q-Blocks itself has actually been primarily a bootstrap company. You haven't really raised a lot of capital for, for the footprint you have in the market and the maturity of the product that you have, which is kind of amazing.

And you also have a lot of peer companies that also came out of Q-Tech and Delft. You know, we've talked about QuantWare already, but there's also Delft Networks, Orange, also Q-Fox and Delft, single quantum magic where is there something, is there some special ingredient in the Dutch quantum ecosystem that you can

in that too, like what's in the water there, guess is what I'm asking.

Niels (27:12.349)
Yeah, good question. I've been asked this a couple of times already and I think there are some commonalities, some basic threats. I think for one, there's quite a strong collaborative mindset that's quite deeply embedded in the culture. And you see it reflected in the ecosystem in the Netherlands and Delft specifically. All those companies almost, they work with each other, they do joint go-to-market.

They do co-developments and they really try to understand how can we together grow as an ecosystem. And the word ecosystem is used almost everywhere in the world, but I see less of this synergistic behavior that fits it well. So I think that has been a strong element of the Dutch companies. The other one is, it fits into it is this

more value chain approach. Most of the companies you mentioned are not full-stack companies, but they are building an element of the value chain. And that also helps in being less venture-backed and being a bit more bootstrapped. The ultimate extreme, of course, but being more revenue-driven and more customer-driven. And that's also very much a commonality in the Dutch landscape.

Sebastian Hassinger (28:32.258)
Right.

Sebastian Hassinger (28:40.408)
That's really interesting. so last question, you kind of have a unique viewpoint because kube blocks is qubit agnostic. So if you had to place a bet, what modality do you think will get to sort of commercial advantage, commercial application first, putting you on the spot? I know, I know I had to ask though.

Niels (28:58.919)
hitting dangerous terrain there.

No, think, well, the field is developing quite a bit. I think recently all platforms have been progressing very well. think most notably past 12 months, think developments in cold atoms have taken quite a flight, where the amount of physical and also even logical qubits have gone up quite quickly. The other one that...

that is really showing a lot of progress in the spin qubit. So for decades, they started out as the only sort of solid state qubit. And then for decades, we're kind of stuck in this two qubit phase. But over the past year, the progress in those spin qubit platforms has gone up immensely. And yeah, they now are

Sebastian Hassinger (29:37.294)
Mm.

Sebastian Hassinger (29:46.382)
Yeah.

Niels (30:00.337)
close to competing with superconducting qubits in terms of the size of the processor. So also quite interesting to see it develop. But at the same time, superconducting qubits trapped ions are solid, are still the most performant platforms and have clear rather low risks passed towards these tens of thousands of qubits. Because the fact that those goalposts are moving closer by also makes it for superconducting qubits more

Sebastian Hassinger (30:20.972)
Right.

Niels (30:28.727)
likely to become to reach that level of maturity. They're big in terms just they're just big in terms of size. But if it's about 10,000 qubits, you can actually reach that quite elegantly.

Sebastian Hassinger (30:28.867)
Right?

Sebastian Hassinger (30:36.739)
Mm-hmm.

Sebastian Hassinger (30:41.635)
Right.

Yeah. Yeah. Fascinating. All right. Excellent. Thank you so much, Niels. This has been really interesting. I think what you're doing with QBlocks is really fascinating as a pioneer in the field. And I very much look forward to seeing what the future lies for the product line and for the company. So thank you very much.

Niels (31:03.143)
Thanks a lot, Sebastian. Thanks for having me.

Sebastian Hassinger (31:06.72)
All right.