E626 | EUCVC Summit 2025: Nadia Carlsten, DCAI & Bjarke Ruse Sejrsen, Go Autonomous: AI Factories in Practice

0:00Or get the AI hype. This panel brings together builders deploying AI in real operational settings. From pharma supercomputing to autonomous workflows. With DCAI's Nadia and Go Autonomous' Bjark on the mic, we get a ground-level view of how AI is changing work, value chains, and competitiveness. Thank you so much. Now I'm also with sound. Wow, a little crowded stage here. Nice and comfy. Thank you for coming. Nadja, let's jump straight into it. We don't have that much time, so straight in. Why is it so crucial for Denmark and Europe to have access to a supercomputer like Gifion? What strategic advantages does that provide in a global scale?

0:48The main reason is it's basically impossible to do anything without AI these days. You can't innovate without AI. You can't be the best in your industry without learning how to use AI, learning how to innovate on AI, learning how to deploy AI. And it's really difficult to do that if you are dependent on access based on either paying a lot of money to get access because you're a large company and you can afford to throw money at GPUs, or you are dependent on some other mechanism that somebody else controls to get access. So the reason it's important for Gifion to be in Denmark is because it's a boost to the Danish local companies and researchers that need that access in order to innovate.

1:33Thank you. So, Bjarke, you have been lucky enough to try to use the Gifion computer. From a business perspective, how has that materialized in your startup Go Autonomous? You can say GIFION for us is not just a bigger and faster computer. It's been a strategic enable for us. We are a classical, you can say, startup scaler. That has been getting far by doing what 99 % of other companies are doing on 99.9%, fine-tuning the models that are openly available. and we could see, like, we could get even further if we trained a foundational model on all that proprietary data that we have. So one thing is to have that data, have that idea.

2:22Another thing is also to, when you then get access to a computer like yours, you actually need to know how to utilize the power efficiently, right? So suddenly it's not just sitting there and doing what most have been trying to do, but you need to sort of actually dance in a way together with the hardware and that's where you can say the match were really unique for us also the timing and so on was perfect so what has it done for us now essentially we sit on top of a world first model that is trained on pure b2b data grounded in that that enables us and our customers to do things that you could say I think the best proof honestly is I would challenge all of you in the room find a company that is AI native that is today handling roughly 30 billion euros on an annual rate for their customers.

3:12That is what we do. And that we can only do because we are running very tailored models for that. Yeah. A question for both of you. And I will let the ladies go first, right? That's how it is, right? So, Nadja, how do you see, you know, these opportunities for other companies in Denmark, you know, how do you leverage such a powerful computer, right? What's your wish for Danish companies and startups out there when they reach out to you? Yeah, I've been making the same wish since I got to Denmark, right? It's for people to move faster, to adopt AI faster. That is the one area where Denmark is lagging, is in AI adoption.

3:52And by that, I mean adoption for useful use cases that will actually transform business, not some POC about a chatbot. We actually don't even do that in Gaveon. Our computer is an AI-optimized cluster. We are designed for flexibility. We are designed for workloads pertaining to innovation of different types, different areas. We have customers doing everything from like what Go Autonomous is doing with the business data, but we also have drug discovery workloads. We have quantum simulation workload. But the common thread is these customers are innovating at large scales. They are learning AI. They are deploying AI within their industries.

4:35And they also don't want to trade off data security and data sovereignty in order to be leaders in AI. So that is really what we're here for, is to enable customers on that journey. Bjarke, can you add some startup flavor to that one? What do you see? What do you hear? What do you believe? I think the biggest challenge actually for Europe to create the next sort of AI native startup is that now you're all in this room, right? That is actually the capital environment, to be honest, right? If we take ourselves, we got this opportunity, but we are one of the few. And if I'm looking to U.S., they are much more comfortable in deploying risky capital, essentially.

5:18And this sort of training your own models is not just something you do, right? It's not one-time training and then you just run with it. We have so much we can do with it now, but it also demands much more R &D than what you normally would do or see in a startup. So that changes the spreadsheet for the VCs, right? And we need to see a change in mindset for us to actually create those companies or else we will see, and that is also great, we will see a lot of the corporates taking advantage of these resources that we now have available, but not necessarily the startup scene. Nadja, where do you see this technology take us?

5:54I mean, the technology can take us many places, right? But I just want to add to what Bjarke was saying. I think the investors have a key role here. I talked to many investors who have companies in their portfolio that are doing, quote, something with AI. And when I ask them what that something is, they don't know. And there is a big difference between the companies that are leveraging existing model and companies that are training new foundation models, right? And there's pros and cons. But as an investor, you should know who is doing what because their rate of growth is going to be very, very different.

6:32And to get back to your question for where that gets us, what we would love to see is for Denmark to be a leader in AI. So not just a consumer of these models that are made elsewhere, but a leader in development of new models and also a leader on how the technology is shaped and deployed. For us as an infrastructure provider, that also means not trading things like sustainability, commitment to green energy, just to do AI. And the beauty of being in Denmark, this is one of the few places where I can say large-scale AI infrastructure and say green energy in the same sentence, and the two are not in conflict.

7:11I couldn't do this in the U.S. But it's super interesting. And just, you know, holding on to it, I would like to understand, you know, what is your biggest pain point, Nadia? What do you miss from startups and corporates when they reach out, right? Because we need ambition here, right? We need to fight for the European future, right? We need to develop in Europe. Yeah. I would say greater readiness for AI. And that means everything from knowing where your data is, knowing how to actually wrangle your data into something useful, but also enough tangible use cases. So people have usually a vague notion when we talk to them about how they could use AI to improve the business.

7:56But they need almost a translation between that and actually being able to do something useful with the computer. So we've been finding that we brought a lot of expertise, of course, at the infrastructure layer, at anything that has to do with actual computes. And now we're having to go up the stack and advise customers on how to actually use these GPUs in a way that's efficient. What can you do with 1 ,500 GPUs that you couldn't do with 15 or 150? So we would love to partner with more people in the ecosystem to actually think through some of these use cases, because that takes expertise both in the domain knowledge, but also in the AI knowledge, which we bring.

8:42Bjarke, last popcorn question for you. Does compute power matter more than talent in the next five years? I actually want to answer another question. So I think a shout out to everybody. to make it happen what you just said, right? It's, I think, an obvious opportunity we have here is actually getting the corporate environment and the startups together. Go what's out of us, nobody knows, right? But the only reason we are at the level we are at is actually because one of the major, I call it the kings of Denmark, right? Kronfoss essentially bootstrapped us. He solved such an immediate pain they had, such a big pain, that they were actually willing to put in quite significant capital.

9:24and that was not even to train like foundational models, but like fine-tuned models. So we are actually a result of that, not in a structured setup. It was not intentional actually, but they just had such a big problem. Imagine that we create that environment. I think we could have a superpower here because we have great talent, no doubt, and we have the resources and all that stuff available. We just need those parties to come together. Awesome. That's it for now, but compute power over talent? No. Exactly. Thank you, Nadja.