Fusion systems always need sophisticated control systems, even if they don’t always know that they need them
Commercial fusion energy deployments will rely on strong and capable control systems. As The Fusion Report documents this evolution, we will report on more of the supply chain for fusion energy, of which control systems are an important part. Today we’re interviewing Mark Pleško, CEO of Cosylab, a 25-year-old European company with 250-plus employees and 30+ PhDs that develop (among other things) control systems for fusion projects. The company is headquartered in Ljubljana, Slovenia, with several other locations worldwide including Menlo Park, CA USA; Brugg, Switzerland; Tokyo, Japan; and Suzhou and Guangzhou, China.
Tell us the about the origin of your company: I have established the company 25 years ago with a couple of students of mine who are co-founders. I was a researcher for 20 years, working on particle physics, nuclear physics, particle accelerators, and other systems like that. As a part of a research collaboration, we have built a control system for a particle accelerator at the Karlsruhe Institute of Technology in Germany. And then we said, well maybe we can make a living out of that, so we have established a company: Cosylab, which stands for COntrol SYstem LABoratory. We have then started to develop control systems and software development services for people in the particle accelerator community. Since that time, the company has been growing to the point that we now have roughly 250 people. Our major customers are almost all the big research laboratories around the world, such as SLAC at Stanford, FermiLab in Chicago, CERN in Switzerland, the European Southern Observatory (ESO) located in Chile and many others.
How did you get involved in fusion energy? Fusion is for us a big opportunity. We’ve been working with ITER for 18 years already; they were our first customer in fusion.
We’ve worked for a number of years with the German fusion institutes; we also work with the UKAEA, and provide a major part of the control system software for the U.S. contribution to ITER – collaboration with Oak Ridge National Laboratory and Princeton Plasma Physics Laboratory.
Who are your investors? There are none! We have started with small services, the first contract was 10,000 Swiss francs, then the next contract was 22,000 US dollars. When we started the company, I went to a conference and met all my research friends and asked whether they had any work for us? Some of them said, “Well, it’s end of the year and we have a little budget left”, and that’s how we got started. And so slowly, but steadily, we have bootstrapped ourselves up to the current size.
Who are your target customers? Initially research institutions and then also companies who are building very complex machines. A case in point is fusion: When we have started working for ITER, we said it’s a reasonable business on itself and although it’s not smashing due to relatively low hourly rates, it helps us to build competences. We thought that if we have the stamina to work with them for another 10 or 20 years, and assuming that fusion will work, then there will be a lot of business later with fusion companies. And this is exactly what is happening now, when we are talking with all the major fusion start-ups, and working with a number of them.
What makes your approach unique in the market? The fusion startups have brilliant physicists and engineers, and they should develop the physics dynamics, the control of the plasma, how it’s moving, new engineering solutions, new types of magnets and lasers, and the like. But control systems is something that, while it’s not a commodity and it’s difficult per se, is something that can be solved conventionally, if you have sufficient understanding of how a fusion machine works. So we thought, okay, why don’t we make something for that?
This is actually what the industry is starting to call Fusionics — a term modeled on avionics in aerospace. Just as avionics refers to the electronic systems that make an aircraft airworthy and certifiable, Fusionics refers to the control and instrumentation discipline that makes a fusion plant commercially operable. We co-founded this initiative together with UKAEA, and the goal is exactly what I described: a shared framework that every fusion company can build on, without having to reinvent it from scratch.
If you want to do magnetic fusion, you have to control the magnets, and you have to control the timing. If you have lasers, you have to control the shots. And obviously a lot of other things, too. In total, just the engineering and programming effort can easily be in the hundreds of man-years, not to mention the cost of electronics, networking and computer hardware, which usually amounts to 10% of the total machine. But we can make it less expensive, because we standardize it. There is a certain workflow, and if you would package this as a standard product, or a standard platform, then the fusion startups can develop anything they want on top of it, and also keep it proprietary, because that’s also important. But the framework for the workflow, while still adapted for fusion, doesn’t have to be proprietary and can come from Cosylab. So that’s a win-win for everybody, right?
What is your biggest issue getting visibility into the supply chain requirements of fusion companies? I think the biggest issue we have is that controls are not treated as a main problem, until it becomes the main problem. But by then it’s already too late. Typical fusion startups have physicists, machine designers, hardware people, scientists, so they have a good team, but of course, they have wrong assumptions about the problems they don’t know,. And software always seems easy. I’m a physicist, so I can make jokes about physicists: for physicists, if it’s supposed to be complicated, there has to be a mathematical formula, like a differential equation. Then they believe it’s difficult. But programming? Any kid can do it. Like all physicists, I was programming when I did my PhD, but of course, one thing is writing a program that makes a physics calculation, which you can correct in case of bugs, and the other thing is getting an industrial, production-ready system working autonomously.
And then we have to talk about safety and about cybersecurity, which must be controlled all the time. You might have an original genius who has developed the software and control system, and then this genius is on vacation, or changes the job – who takes it over?
These are the typical things that we have to evangelize to the fusion people. And again, they usually don’t worry, because for them, of course, getting the plasma ignited is the key thing. And everything else can be, in principle, solved later. Thus they create what we call technical debt. But at the end, you have to pay back this technical debt. And the later you start, the more technical debt you have to pay back at a shorter time frame. We are there to help them avoiding this debt altogether.
What’s your thought on how important is AI, and how hard is it to compete with it for investment? I think, actually, AI is a godsend for fusion for several reasons. One reason is that data centers will need a lot of energy. People are already looking into alternative sources of energy like nuclear, solar, etc. You cannot have the data center not work at night because of solar, so you will need a mix, and fusion is a potential candidate. Not for today, obviously, but for the future. The other good thing of AI for fusion, which is indirect, is that people have learned to invest a lot of money into relatively risky businesses. That means if you invest now in AI, you may as well invest in fusion. The risk is probably comparable, but the gain could be also comparable. Obviously, AI will be exponential if it works, but then also fusion will be exponential if it works, because everybody will use fusion as the energy source. The only thing that could be detrimental is that all the money goes into AI, and then there is no money left for fusion. However, this is a general problem for any companies that want to IPO. Fusion is still a small fish in the pond, so there will probably be enough money left one way or another. Finally, a bigger danger is that people will burn their fingers when the AI bubble bursts, and then they will not want to invest into anything that is remotely risky for a while.
Control Systems – Often Forgotten About, But Seldom Unimportant
For better or worse, control systems are often the face of systems such as nuclear reactors, fossil fuel power plants, hydroelectric dams, and commercial fusion deployments. Their screens and the information they provide are often the critical piece between a first-of-a-kind (FOAK) system and a production system. More importantly, control systems mirror in software the complexity that fusion systems embody in hardware. This is where Cosylab exists – at the nexus of that hardware and software.
