Entrepreneurs to Watch 2023: Carlos Abellan

The July/August 2023 issue of Optics & Photonics News featured the magazine’s biennial feature spotlighting 10 Entrepreneurs to Watch. Here, we offer an interview with one of those entrepreneurs, Carlos Abellan, the cofounder and CEO of Spain-based Quside. Quside is exploiting quantum processes to produce truly random numbers for cybersecurity, cryptography and randomness-based computing workloads, while also providing the ability to monitor their quality.

Can you give us an overview of what your company’s doing and where you’re taking it?

Carlos Abellan: We are a deep-tech company based in Barcelona and spinning out from research. So we come up really from the scientific world. Basically, we focus on delivering the most advanced randomness solutions to the market for three main topics: the generation of randomness, the monitoring of randomness and the processing of randomness.

And we do that for two markets. The first is to help people build the strongest cryptographic foundation. So it’s about cybersecurity and making the transition to future-proven security. And then the other market is the computational market. So it’s about accelerating what are called randomized workloads. And you’ll find many, many use cases, from cryptography to finance, insurance, logistics—many domains in which these workloads are present. And we can help customers run them faster and more efficiently, with a lower energy footprint.

A lot of people are gravitating toward quantum technology as a commercial opportunity. You’re focusing on quantum as a source of genuine randomness or entropy. What makes this such a significant market opportunity right now?

One of the interesting things is that random numbers are foundational in many IT domains. So there are a lot of software algorithms that rely on randomness—or in other words: no randomness, no application.

Security is one example. If you don’t have random numbers, and I mean hardware-generated random numbers, you don’t have security. It doesn’t matter how much you invest in algorithms and mathematics; you don’t get security. So inherently, in cryptography alone, the market is already huge. Every single device would need something like this.

And then the other market is in computation. There are a lot of algorithms that, unless you use random numbers, you cannot solve them. These were invented in the first half of the last century, algorithms like the Monte Carlo method. And random numbers allow you to get a sufficiently good solution. It is not exact. So the thing here is not what makes this a huge market. The question for us is, why would people move into doing it with quantum means, rather than what they are using today? But the market is already there, in a sense.

So why would people use quantum?

In cryptography, you need to generate what are called unpredictable digits, so that no one else can predict them. One means used to do so is pseudo-randomness, which is catastrophic—you cannot use that “as is” for security. And I hope not many people are doing that today. The other method is using other sorts of classical systems that are more about your ignorance of the process than actual, fundamental unpredictability.

One example I normally give is the weather forecast. If I tried to predict whether it will rain or not tomorrow where you are, I can make a guess. I will most likely fail—first, because I don’t know where you are. But even if I know where you are, I will still fail. But if I look at the weather forecast, and I know where you are, my chances to predict tomorrow’s weather are better, right?  So based on how much effort I put into predicting the outcome, I get better at predicting it. And no matter how complex you make it, this is always true.

“With quantum physics ... It doesn’t matter how much effort you put on trying to predict those random numbers; you will never succeed. It will always have innate unpredictability.”—Carlos Abellan

With quantum physics, that’s not the case. It doesn’t matter how much effort you put on trying to predict those random numbers; you will never succeed. It will always have innate unpredictability. That is one fundamental thing.

The second aspect is that for us, it’s not about only generating the randomness—you also need to monitor it. This is hardware, and we always tell the same story: in maybe 50 years of IT, no device has been sent to the workshop to fix a broken random number generator. No server, no laptop. And most likely many broke. But no one is looking at that, or no one is looking carefully enough to care about that.

And quantum physics has something very unique. Because its randomness is fundamental, you can test the physics and make sure that it is working well and is able to produce randomness. With other technologies, you cannot. So the fundamental unpredictability plus the ability to measure the quality are the unique features of quantum physics for these markets.

How does this all work in practice?

We use a technology that we call phase diffusion in semiconductor lasers. So basically, we exploit the fundamental process of spontaneous emission in a semiconductor laser. The laser is switched off, and at that point, there is a significant amount of spontaneous emission noise. And that means that there are photons that are emitted with a random phase, which is what we want to measure.

But measuring the phase of a single photon is complex in that domain. So we play this trick where we gain-switch the laser. That means you suddenly switch it on, and you generate millions of “copies” of whatever was inside the cavity. At that point, all of the evolution is deterministic. But that initial random seed will determine the phase of whatever outcome is in there. So you generate optical pulses that all have the same intensity, but each of those has a random phase.

Next you use interferometrics. So for instance, you can use a Mach-Zehnder or heterogenous detection, and you convert the random phase to random amplitude. And when you have random amplitude, you can use a photodetector or a digitizer to move into the electronic domain, then convert the analog signal to digital with standard electronics. The trick is amplifying that microscopic phase into a macroscopic observable.

Is there anything specific that you’d say especially distinguishes Quside’s approach?

There are a few aspects of this core technology that are unique for Quside. One, it leverages the semiconductor industry for optical communication. So it can run very, very fast, and it’s ready for going scalable. We can already produce full wafers of our chips; this is something we’ve demonstrated, and we’re technologically ready.

It is also very high quality in terms of noise. The challenge, typically, is that you get the quantum signal, you’re observing it, but how much additional noise is in there? How much noise comes from the detector, from the electronics? How strong is the quantum signal? Normally, it is not that strong compared with the others. So distinguishing the pure quantum signal from the rest may not be obvious.

In this case, because of the interference, we end up with a very strong signal that is very, very large. And the quantum noise that we can measure compared to the other noises is very good. So it allows us to do a very detailed inspection of the quality in runtime.

Quside has what you might characterize as two product lines: quantum random number generators (QRNGs) of various types, and a “randomness processing unit” (RPU). What are the distinctions between those?

The traditional QRNG, for us, is to help customers build the strongest cryptographic foundation. So it is about producing the right amount of cryptographic material to protect communications. That is its purpose, that defines the market and who our customers are—OEM manufacturers, infrastructure enterprises. That is sort of the market that will consume this.

What is an RPU instead? An RPU is a hardware accelerator. The name came in a one-on-one with a colleague, and he just jokingly said, “the RPU.” And I said, “That’s the name I’ve been looking for for a year!” It gives the customer an impression of what it is—you use a GPU for machine learning or whatever, you use an RPU for randomized workloads.

And what’s a randomized workload? It goes from cryptography, you can put a cryptographic algorithm there, and the good thing is that you already have very good, high-quality random numbers. So it’s a great pack. But you can also use it for financial simulations and other randomized algorithms. Heuristic optimization is one. We’re now exploring synthetic data in artificial intelligence as another use case, which is about generating artificial data randomly.

What customers can get from this is two things. They can do it on a CPU, but it will take them longer, and it will consume a lot more energy. So when performance or efficiency is something you look at, this product can help you.

So it’s like mining Bitcoin—you’d want to do that on a GPU rather than a CPU. If people have a heavy load of randomness they need to process, they go for the hardware accelerator.

“What customers can get from [the RPU] is two things. They can do it on a CPU, but it will take them longer, and it will consume a lot more energy. So when performance or efficiency is something you look at, this product can help you.”—Carlos Abellan

Exactly right. And that is the way we explain it to the customer. We say, “Give me a profile of your code. How much time are you spending on randomness generation and randomness processing?” If you’re spending a significant amount of time on those two things, we can help. If that’s not a significant amount, you can use still use it for the quality. Again, having high-quality randomness is important to avoid correlations, to avoid patterns. So that is another element, and some customers go in that direction.

I remember when I was doing my Ph.D., I was in Japan on a trip. And we did this seminar, and this guy stands up and says, “I’m not using pseudo-randomness in my simulations. I don’t care. I’m not using it. I have 16 slots, and I want to build them with as much speed as I can.” So this is because the quality is another element.

Let’s talk now about your road to Quside. How did you first become interested in science and in optics and photonics?

I am a telecom engineer by training, and I arrived there because I love sound engineering. And at some point, I thought, I can take sound engineering, which was a shorter degree, or I can do telecom engineering. And I took the longer road.

Then when I was there, I learned I loved cryptography and I got interested in quantum physics. And I guess that was inspired by the TV series “Big Bang Theory” at the time. I thought it sounded super-interesting. So I took some lessons, I loved it. And then I went to one of my high school teachers, and I said, “I love these two things, but I don’t know what to do for my degree and for my master’s thesis and all of that.” And he said, “You should go and look at this professor, his name is Ignacio Cirac. He’s been doing a lot of things in quantum.”

So I contacted Ignacio Cirac, and he put me in contact with other people. And I discovered the field of quantum cryptography. That was the thing that combined the two things that I liked. So I joined ICFO, the Institute of Photonic Sciences, for my thesis. And from there, my project was quantum randomness for cryptography and so on.

Were there any skills that you’ve learned as a Ph.D. student that that became applicable to running a business?

During my Ph.D., I was always very involved in outreach activities with the communication department. And in particular, there was one project that was a Friday afternoon project that turned into a full-year project for a lot of people; it was called the Big Bell Test. So that was a full year of massive communications and outreach and thinking about how to talk to people, how to communicate complex concepts in simple ways that people can follow.

I would say all of that communication and outreach exposure was an amazing experience. And I believe it has been helpful, because in the CEO mode or the “start a company” mode, you’re selling all the time. It doesn’t matter if you’re selling your company to investors or you’re finding customers or you’re recruiting people—you’re always telling a story that people need to follow. So that outreach, I believe, was a very important part of my journey.

Have you always had an interest in starting a company? Or is that something that came later?

I actually started another business before Quside and had tried to start another one before that. I would say it was not that much about starting a business. That was not the motivation. It was about starting something new—this is something I like.

Were there lessons learned from founding those previous companies that you carried forward to Quside?

One of those companies, we created it and grew it, and it’s still growing healthily and very excitingly. I was there at the very beginning, during say the “zero-to-one” step and doing the first fundraising. So I got a lot of exposure to talking to investors before going into Quside, and that was helpful. And starting to think about the customer. That was a very good experience because it was not related to the research I was doing; it was in parallel to my Ph.D. So I learned a lot from the process.

“The first [company] was a complete disaster even before it took off. The lesson learned was, don’t start the company with your group of friends; that will never work.”—Carlos Abellan

The first one was a complete disaster even before it took off. The lesson learned was, don’t start the company with your group of friends; that will never work. So that was the big lesson learned.

In between, as part of our research, we applied for funding from the regional government here for training with the Haas Business School at Berkeley. That experience was really, really good. The way that those guys gave the training, how they pushed us to think about the customer—that was really helpful. It was not a full-year MBA program at that time, it doesn’t matter. It’s just exposure to a different way of thinking. That was very helpful.

You mentioned ICFO, where this technology was, in some sense, incubated. Could you talk about some of the highlights of building the technology there?

ICFO is a research center, and I was on this research line with two professors, Valerio Pruneri and Morgan W. Mitchell. One of the interesting things about the institute is they have a unit that is called KTT, Knowledge and Technology Transfer. And basically they scout potential technologies that may have an impact, and they try to start to build some patents, talk to customers and build the journey. So from the very beginning, they were involved in this project.

It was a combination of, on one side, the physics of the process, understanding it and pushing it, how to measure the quality, so on and so forth. Then pushing the technology, first with prototypes and then with photonic integrated chips. At that time, photonic integrated chips were a new thing; unless you were a foundry, you could not build your chips. But you could start to do multi-project wafer runs. So we pushed the technology

And actually that combination in parallel was interesting because we built some devices; we built seven and shipped six to three research groups. One in Vienna to Anton Zeilinger, one to TU Delft to Ronald Hanson and one to NIST to Krister Shalm, because they were doing these loophole-free Bell test experiments, and they all used the randomness technology because of the speed and because we could measure the quality. So at the end, the technology and the quality merged together into very unique combination for those experiments.

I really like to build things that people can use, looking at the application level. So building for something bigger, the institute framework, and the state-of-art research that we undertook was, I would say, the perfect storm in a sense.

“I really like to build things that people can use, looking at the application level.”—Carlos Abellan

When did you realize that this was something with commercial potential and start talking about spinning it out?

The commercial potential was almost from the beginning. The exploitation path was undecided, because one path could be licensing the technology to someone else, and then someone else builds it. The Institute has a preference for spinning out companies and having the researchers take the technology further. So they had that preference, and I had that preference. So even if we explored some potential licensing opportunities, we decided to take the spin-out path.

And how did the team initially come together?

I left the Institute 100% and joined the company, as did another colleague, Waldimar Amaya. We had started together on the research program; he was a research engineer. We were missing one piece—well, we were missing many pieces, but at the technology level we were missing one piece.

We understood the technology, built the prototypes. But we needed to build photonic integrated chips because we always had the vision that this technology could reach everyone, everywhere. So scaling was an inherent part of our vision and that of our early supporters. We could have built appliances for governments only, and then we then wouldn’t need the photonic chip. But our vision was always to scale and to serve everyone.

So we hired the former Ph.D. student of the group, Domenico Tulli, who had gone to the US for a postdoc and then came back to work for a company in Spain. He joined as a cofounder at Quside, to lead all of the photonic chip and scaling platform. So it was the three of us at the start. We were, of course, missing all of the business and financial expertise and things that a company needs.

But at the beginning, the challenges were of a technological nature. We had contact with key customers, who provided confidence that a successful product would have demand. But at first, it was about questions like, “can we produce this thing at a relevant price point?” etc. Then we started to recruit some folks in engineering, but it was not until almost 18 or 24 months into it that we hired a director of operations. That was our first nontechnical hire, and it was almost three years to the first salesperson.

What were some of the milestones once the company had been started, in getting it off the ground and getting it funded?

Our first backer was purely strategic, as they were interested in having that technology and demonstrating it could be scaled. So the first milestone was to build the first products and put them out in a way that someone else could use. And repeat that process and demonstrate that you can produce the chipsets at a significant scale. At that point it was undefined what scale meant. And actually, our first approach was, “let’s do whatever we did in research,” multi-project wafer runs or whatever. And that was our plan, and all of our financials were built around that.

But then three months into it, we said, this is bringing us nowhere, because even if we succeed, they will not be able to scale. It will be wasting time and money. So we decided to take a pure-play foundry approach that took us a lot more money and a lot more time. But today, we can produce at an extreme quality for wafers.

That was a pivot very early on, but everyone was aligned on the purpose. And that was the first set of milestones. Now you have the products, you have the capability—what about the sales and the commercials and the customers? So we have a lot more focus on selling it, supporting it, solving the customers’ pains in a more efficient way.

So this is the next generation from the growth perspective, but we never lose the innovation aspect. We’re very active in creating patents and exploring innovation, launching the RPU and all of the innovation around it. We have a very strong pipeline of innovation, but it is always driven toward the customer. So it’s not like crazy moonshot research. It’s real life and customer driven.

“We have a very strong pipeline of innovation, but it is always driven toward the customer. So it’s not like crazy moonshot research. It’s real life and customer driven.”—Carlos Abellan

Can you talk about the state of investments right now and what the current goals of the company are?

We recently secured a series A funding round with two lead investors. One is Trumpf Ventures and the other is a deep-tech investor in Spain called Bullnet Capital. Other investors also joined this round.

The main focus today for the company is twofold. We’ve built the capabilities, we have products, and we can support those products. They are different form factors, but all of them are relatively high-end products targeting a high-end market; we’re pushing maximum gigabit performance. So one part is consolidating the company there and growing the sales, supporting these customers, helping them scale.

The second part is continuing to focus on our vision. We want to make technology available for everyone, everywhere. How can we go to the next stage? Because we strongly believe that this technology is for any connected device. We’re now focused on satellites, data center equipment, enterprise-driven appliances, that sort of thing.

How do we go into the large-scale data infrastructure market, into every server, into every firewall, into every gateway? What do we need to develop? What is the next technology we need to pull to reduce the cost so that it is attractive to that market? And eventually, how do we make it into every phone?

We need to invest our limited resources smartly to bring us closer to shipping the data infrastructure market and position us to eventually serve the consumer market. So one goal is very operational. The other is more about reducing the cost of the supply and making it more attractive to enter new markets.

What do you think are some of the biggest lessons you’ve learned during this process?

I would say, the perception of what the challenges are changes dramatically during the journey. In research, when we demonstrate that the chip can do that phase diffusion thing, and that we can measure it—you think that from there, it’s all downhill. You publish the paper and then you realize, oh, that wasn’t everything. You need to build the product and that product needs to work all the time.

“You publish the paper and then you realize, oh, that wasn’t everything. You need to build the product and that product needs to work all the time.”—Carlos Abellan

At that point, unless you do the homework, you don’t have a line of 10,000 customers placing orders. So all of the commerciality, the marketing, support for the customer—all of that is really tough, right?

That’s a transition in how you think when you move from research into shipping product. The perception is that it’s the paper, and then the product. But no, it’s actually all about the customer. If there is no customer, there is no business around it. And that focus is not always well placed. That is one thing.

And the other thing is that managing people and finding the right team is everything. Wrong choices on the team will set you back two years. It will be a terrible mistake. And managing people, and specially a high-performing team is also a lot more time consuming than you may expect.