#50 Magnus Grünewald: On AI Sovereignty, Deep Tech Infrastructure and Building High-Performance Teams

If what we are trying to do is becoming a reality, it's going to be one of the biggest companies in the world. I think the upside is infinite. The importance of it is infinite and also doing it. Some Eurobites, something that's very thrilling to a lot of people. Welcome to the Mostly Awesome podcast. My name is Katie and I'm Hannah and together we're the co-host of this podcast. Our guest today is right at the center of one of the biggest strategic battles in tech, AI infrastructure. We're joined by Magnus Greenavite, the co-founder and CEO of Lysium, a startup building Europe's sovereign GPU cloud. To make this visionary reality, he recently raised a massive 10.3 million euros pre-seed round and he's only 23 years old. Before founding Lysium, Magnus served as chief of staff at NPUL where he helped scale the company's heat pump business from a small team into a major operation. His colleagues describe him as having incredible execution of strength and a superpower for recruiting. Today we'll dive both into Magnus's personal journey and Lysium's mission, exploring how he raised over 10 million far deep tech, the convinced top tier talent to join a brand new mission and what it really takes to build the impossible. Magnus, welcome. Thanks a lot for having me. Very, very cool introduction. Yes, also warm welcome from my side. I would like to start with your personal journey. So we learned that you began your first entrepreneurial and the worst at just 14 years old from running your own podcast studio to funding a marketing agency. Later, you went on to study business administration, completing internships and going back. Could you take us back to these early days? What were some of your first attempts at building something from scratch and what did you learn from those side quests along the way? Of course, so for me, I've always been very intrigued by everything that's entrepreneurial. From very early days, I've always been inspired by building something on my own. Getting things from zero to one has always been something I'm really intrigued by. As you mentioned, very early on I started a small podcasting company in a sense of when I was doing a podcast for other people, moderated them, if edited them, I published them, which was very exciting. I've been in this creative bubble for some while, doing also some music video production. And some other creative stuff. Later, I started an event manager in a company where we created bigger parties in a city. I think stuff that you do when you're young and trying to earn a quick back, which is super, super exciting. I think it taught me a lot of getting things from zero to one, trying out and being willing to fail. And just basically cutting corners to achieve something with a very limited means. Which I think really, really helped me know with getting stuff from God, which I think is great. Then, actually, I wanted to become a physicist during the time, wanting to learn something real, wanting to learn something that can help the world in just being meaningful. But the side quests taught me that maybe I'm not the best physicist, but I'm better doing something on my own. Therefore, study business, didn't have internships, then ended up in an environment now. Now, I'm doing what I see. Okay. That's super cool. So you thought about becoming a physicist, but then quickly decided against it, probably rightfully. Let's hope for it at least. Let's hope for it. We don't know yet. Exactly. But what made you decide to go to Monheim at that time? I think it was a very close friend of mine who managed me into doing this. I applied to my home town in Dresden to study physics, and I applied to the Monheim very last minute. I think it was the last day. It was on a deadline. I got accepted to both. And then it was just a gut feeling. And I had this urge to go out into the world. You go somewhere else, move away from my home town, and then once Monheim. So you mentioned that you were already very early uninterested in an entrepreneurial project. Did your studies then impact your decision of becoming an entrepreneur? I think it was rather the other way around to us. Well, the endeavors motivated me to switch to business and find the medium of lives in that, which turned out maybe I actually studied something else because it didn't really suit me that well. But I think it helped me in a sense that it gave me a little background knowledge about the entrepreneur and so I think it helps me with the job of the basic, and just gives you this broader picture and just self-confidence that you're doing. So I think it helps. But other than that, I believe that for building a company, it might be actually more useful to learn something other than business because business in a sense is quite trivial. Some people might now disagree with that, but I think it is learning something meaningful that you can then turn into a product, which then can be turned into a company might be more useful. But yeah, one of the many learnings along the way. Nice. And so you're an entrepreneurial driver that developed, as you just mentioned, led eventually to the founding of Lysium. Let's talk about that for a moment or two. For those of you who haven't heard of Lysium yet, could you maybe explain in one minute what it exactly is that you're building? Of course. I'm happy to talk about it for ages, actually. Not only a minute, but I try to keep it as short as possible. What are we trying to do with Lysium? We're trying to run managed and optimized GP workload or computer workloads in general for our customers with the vision of making it as easy as possible for the end user. I sometimes like to compare it with a power socket. The thing when you're charging your phone, when you're using electricity, you don't know and you don't care about what's happening and how the energy is produced, how it's maintained in the crit, how it's pushed through substations and everything like it, it doesn't really matter to the end user, right? They just want to charge their phone, for example, and plug something into the walk. And then it's what we want to do with computers, right? We don't want people to worry about the infrastructure. We don't want to take care about the orchestration, everything that happened in the back. We want them to focus on their work, because we want to enable science and users of compute to focus on building their amazing products. And we take care of the infrastructure in the back. Okay. Yeah. That's super interesting. In AI infrastructure, there are so many things happening and also components. Do you have the idea for a spark to exactly do what you guys do with the GPU providing? And what give you the push to turn that idea specifically into reality? I think in the beginning, I have to be honest, it was the illusion. I think it is very big to look at when it comes to doing this. I think it started actually before CHGPT happened, it was very intriguing. But I, as can say, told me I looked at it, it was very, very fascinating. GPT-3 was just announced, so before CHGPT in the 3.5 turbo came and just said, "Magical" and I was very, very intrigued by all the power and GPUs it was necessary in order to train such models. And during my time at Enbal, then where obviously I did a lot to do with infrastructure and the energy and everything adjacent to it, it came natural to me that I was thinking about data centers on an infrastructure level and thinking about how can we provide those assets with this critical infrastructure out of Europe, because everything to that point was on. But you were hyperscalers and looked at it from the asset level and then after meeting my call phone and speaking with them a lot, to also speak it all with customers. Problem that occurred is way, way bigger than just the building of the data center, but it goes into the user and the workflows and the adoption of workflows. You're going from the energy sector, thinking about data centers and then going to the direction of GPU, providing that and now you're competing as you already mentioned with big hyperscalers. And the question is, maybe it is delusion in the beginning, but how do you differentiate? What does that Lysium offers that the others don't? It's a, you could make the point that everybody in this space, including the hyperscapers, offer compute, I mean, everybody builds data centers, Rensal, the GPU capacity. So the end user, what makes us different is that we don't think about selling compute like the hyperscalers, what I'm a competitor to, they think about it like building out assets, renting it out of it, selling services on top of it, and we see two instances need to select, it's normally complicated, it's very tedious, very expensive and it's very prone to error, it's just a dislodging. We think about it, if it's like this power socket approach, we don't want people to rent out machines and do that where we want them to basically pay for the execution of workload, we just want them to think about it on workload level and not on like a rental based model. So I think this is the core difference. We don't offer a service and a rental based business to the workload execution, we just offer the workload execution and therefore the user flow is a completely different what people just actually never think about. This is the core conditions, I would say. Why do you focus specifically on GPU availability and not, for example, CPU as well? Because after all, it's not only AI applications that need digital sovereignty, right? Yeah, of course, you could make the point that everything is rewamped and it's been learned in Europe. We just focus on what we believe is a, you know, the biggest pain point in the market right now in talking about startups and things that are narrowing down during one thing, it makes me well in the beginning. It's crucial to success and we decided to focus on GPU workshops in the beginning because they are very new in the market and then GPUs just popped up when I popped up really, they have been around before but not to that extent. And therefore the software and the orchestration is not very sophisticated yet. There's a lot of problems with interacting with it, it's very complicated. A lot of machines fail, GPUs are underutilized. So therefore the problem for the end user in the interaction and also in the cost side of things due to, for example, on the utilization is just the biggest and therefore by completely rethinking and rebuilding it, we have the biggest impact, the biggest improvement that we can show to the end user. Now looking at the deeper on your customers, do you see any patterns in who your customers are? What type of companies care most about the environmental and sovereign aspects of what you're building? I think when it comes to sovereignity actually, two groups of customers, first one being them who are very eager, I need focal clients in looking at, for example, meta companies who just need to store very sensitive data, in data centers in the country of origin, basically, therefore might be also biotech companies or just other highly regulated industries that are required to upload data stored in the U that have the requirements to have their data computed within the U and everything associated with it. And then there's the second group of customers who don't require EU sovereignty, but want to use it as a hatch for potential future, I think, shifting compliance and shifting regulation and just want to use it as a setting point for their customers as well. Because I think to basically trick us down the value trade, as you can say, it's completely made in Europe. A lot of people see this as a very, very high value proposition. So this is for the sub-renity part, there is also the product part, right, it's making it easy. This is what makes us special. I always like to say this is what makes us great as a being in Europe, but it's actually the product of the orchestras on top because we have a lot of features that are novel that at once in the world, they're very, very high tech, very, very low level that would invest at a lot in R&D, make it tremendously amazing. And those are customers in the research space, for example, people that have a very high need for compute in order to get their workload executed, it might not have the biggest understanding of it. For example, now, in the biotech space, everybody needs to use compute. If you run the physics or simulation company, you need to, you have compute and you need to run it, but the people that work with it might not have the biggest expertise in running on large scale distributed systems. And they shouldn't care about, right? I mean, they should focus on the research that should focus on their simulations and the novel protein engineering pipelines, whatever it might be. And for them, we can just create so much value because they know don't have to worry about writing. And they can just focus on the research again. And I think this, if you look at it on a band diagram, it's got a disease, so run it's you need and also the ease of use need. There's a subset of customer who we are basically the only solution that they can use. Well, moving on, we also want to understand how you work. And some of your former colleagues described your time as Npal as extremely formative. You also helped scale the heat pump business there from just a few people to over 100 employees. So a startup within a startup. What would you say were your biggest takeaways from that time? And how are you applying those learnings now at Lycium when building your own teams? And then I had the luck of being a tiny part in a very big, amazing, amazing team with the most amazing colleagues down, just the most amazing learnings one could ask for. And I think Npal helped me in a sense that it really took away the field of doing something very, very capital of things, and very asset-heavy. I think everybody could have as a sense of how to get software and how to scale it. Kind of, yeah, we can iterate quickly and we can do X, Y and that. As soon as you get into operations heavy and also like capital and things, businesses, big basket signs, it's a very non-trivial approach of looking at problems because the kind of ability to iterate isn't a bit slower, and then I was still very high. But it's slower than when software businesses. And therefore, you think about problems on a way bigger scale. It's not about how can I tweak a button, how can I do this, but it's a very big out-of-the-box thinking, how can we scale up processes that are not a tiny bit better, but they're 10x, but how can we complete the re-engineer processes in due and a way that's very structurally very smart. And I think there's really, really help, especially in industry that is very capex in time. That is very complicated, especially in the vertical integration that we do controlling a lot on the asset level as well and not only on the software as that. So I think that's really helped me. On another note, I think what we did extremely well was hiring. So I think just the way to understand how people work or they're motivated or to get the most out of them and how to give them an environment where they can strive at is something which is on there, in which I hope not to apply this company as well to build a place where the most amazing people can meet and work efficiently. And I mean, as you mentioned already, I think startups inherit parts of the people that work there or especially the founders came from, maybe something that is unexpected. Do you have a cultural practice or something that you do at Lyceum that others might find surprising or even counterintuitive? I think we have a very extreme delegation in a sense. What we do very, very early on with people is just give them a tremendous amount of responsibility and just make them own topics from the beginning when they start. This leads to a situation where people just take on so much ownership and they're indoctrinated that they want to own as much work streams as possible, which leads to a very, very quick iteration and the ability for us to think about new problems very quickly because you just have so much ownership on other things that you can basically leave this be thinking about new topics and have an amazing team that takes care of it, which I think is only enabled by just the ability for Lyceum to attract very smart and very hard working people. I think this is something that most other founders and also other companies don't do to that extreme. I think they very want to be involved very little, never they want to make a lot of decisions on their own and just keep it very close from the beginning, which I think we distribute it very, very quickly and basically get experts on the field that know as we get their job better than we do and give them as much resources as they need to know to do their job. Very interesting. So you mentioned already one of the things you do to keep people motivated is that you give them a lot of ownership from the beginning on, do you have other techniques how to get people motivated in your team and how to make them work as efficient as possible? I think motivation is probably one of the biggest challenges of a company, right? I mean, if you have happy employees and if you're motivated and pleased and they don't very skilled, you've basically done everything that you have to and the company runs itself in a sense. I think motivation can be done by classical factors, like compensation, I think everybody in the company is incentivized, but shares everybody is all over the company as well, which I think really helps. Other than that, I think there are like, I like to call it basic adjacent motivators that everybody has as well, which for example, is feeling of community, but also the sense of moving the company forward. So I think it's very, very important that we share successes in the thing that we, everybody tells if they do something right to basically get the ceiling that we're moving very, very quickly. And I think this is like a certain setting prophecy, right? If people are motivated to look to companies going forward, they're like, people are feeling even harder and chased at high, basically, of progress, which I think is something you can steer, which is something you can incentivize and amplify by communicating it very well. And I think then there are individual motivators, like factors where everybody is motivated differently. Some people might need X, some people might need Y, some people might need Z, as that in order to be motivated, and I think therefore we spend a lot of time in one of ones and like to go to lunch, you know, with people individually, try to understand what are the things that keep them up right now? What are the things that stop them from being motivated? And what are things that they might need as an individual to feel motivated? And this is something that which I think just requires a lot of time and attention to deep down and adopting and just building the individual players here for those people to feel happy. And the combination of those three things makes the thing very good basis for motivation. Really cool. And then maybe taking one step back in the recruiting process, one thing that you mentioned that really stands out when looking at your highs is the number of very talented people what convinced the people to join your mission? Yes, so I think people from the best of the best and not of people, like PhD backgrounds or from quantum hedge funds, for example, were just in general very, very smart people. I think it boils a bit down to our philosophy of hiring, which is getting people on board that are just incredibly talented, just a lot of raw talent, combined with a very high sense of urgency and very high level of conscientiousness. And if you have those three things together, super smart, high sense of urgency, conscientious people and you achieve a setting where they are very aligned with the vision of the company each, then you have people that can be missionaries and just run source problems and just do it find problems on their own. And I think this is the base that you need to find very heterogeneous setup of different skill sets. So get some people on board that have some expertise, I think especially in our case where you build something that has long development cycles and it's going to be very reliable and it's going to be very robust and scalable and safe that basically have done some mistakes in the past. It can give some background in that can tell some, can tell the story and can tell the mistake is that people should avoid and combine it with just people that don't have that experience, but it's just so hungry and so willing to work that they are willing to move mountains and just try stuff, sail, get bigger on their feet and continue the process. And this is a set up I think we did very well is putting together a team that is very well and sitting together, it's good genius enough to bring a very vast level of skill and also ideas, but it's still homogeneous enough that it has the same culture and everything basically works together. So we spent a lot of time on putting together this team and I think we did a very amazing job until now. When you're hiring the best of the best though of each of the fields, they have many opportunities. Yeah. And why are they choosing a CM out of all places? It sounds exciting, but a set they also have many different opportunities. I think it's vision, it's problem and it's team. First of all, the vision of what we're building is just in my opinion, I think if what we are trying to do is becoming a reality, it's going to be one of the biggest companies in the world. If we are the ones that can do the easiest way accessing this incredibly important resource of the future compute, and we can make it in a sense that everybody has access to it and we can make it super-sitioned or the usage in a cost side of things, then yeah, I think the upside is infinite, the importance of it is infinite and also doing it some Eurobites, something that's very thrilling to a lot of people. So this is the vision part. The problem part is something especially on the engineering side of things that really attract a lot of people, because the stuff that we are doing is not trivial, it's not something, yeah, we might need to write some typescript and move some buttons around and try to get it good user experience, but it's very on a low level optimising and doing a lot of research also and just, yeah, breaking a problem where a lot of people are working on and this is something where I think especially kind of tech people are really, really intrigued with the problems that itself. And the third thing, I think this is probably the most important thing is if you have an amazing team, other amazing people are attracted by it. So if you're very good, you want to work for somebody that's also very good, you want to learn from them, kind of get a speed of iteration going, I think we see this when we go into the office and they're just so many renovally smart people, just exchanging ideas, and your frequency that you can't even suddenly is just so quick and so intense and so small that a lot of people are intrigued by this and this is a culture where you get people together that want to work on a very, very big problem, a very complicated problem with a lot of talented people and this is just a work environment where a lot of people see a lot of upside and a lot of sun and trying it out at least. Very understandable. So taking one more step back into the holy grain of founding decisions, choosing your co-founder based on what criteria did he decides on Max Nero meant and how did he do actually means for the first time? So I think, first of all, I think the most amazing co-founder in the world, it was Max basically the luckiest thing that ever happened to me. How do I do it? A big shout out to the guy, really, really low on. We met through a mutual friend, we got connected, we started speaking, I think it just covered a lot of kind of pain points on the user side of compute me as I alluded to earlier and thinking a lot about the infrastructure side of things and I think we immediately clicked because we have the very same culture and making decisions and just iterating. We're very done, we're very direct, there is no bullshitting and I think it's just incredibly nice to work with MBAs, making decisions so quick and I think basically we don't disagree, we come to a conclusion in like 10 seconds because we have a lot of trust in each other and we know that the other person is thinking X, it's very based on a lot of research, a lot of experience and I think therefore we can iterate a lot and we can find solutions to problems so incredibly well, step, the capability of us working together is just amazing. I think I've never worked with anyone so incredibly efficient, which also void the fact that I said earlier, right? We've worked very well together to trust and to fight itself and therefore it's just a self-fulfilling prophecy. And I think the second part that really makes a lot of sense is we know the strength of each other so we know where the other person is very, very good at and given our background is very clear who owns which topics. For example, Max is the smartest person and there is so everything that is on that side, it's his car. For example, if I talk to customers and I learn something, I tell them we talk about it and he takes it and so his position is based for creativity. One of the other sort of thing is when it comes to business, finance and go to market, there's what I've been doing more and therefore I probably can get some amazing ideas on the table as well and doing just an incredible way of working together and then to know I can't complain. I would love to say something bad but I can't. Okay, yeah, I mean, after our recording, you can also know it's getting. Yeah, yeah, no, sounds incredible. The team spirit that you have and also the chemistry between the two of you and speaking of which, what do you focus on? It's business, finance, to go to market, and I see you made headlines with it's 10 million pre-seed around recently, so congrats to that again and it is a huge milestone for such an early. Like such. Yes, early stage, digital company. Could you take us behind the scenes of that process, egg? How does fundraising work and how did that come about? Of course, one thing I want to put before the inspection of the finance and everything else you mentioned. I'm not an expert as well, so we have the most amazing team. Especially a big shot at the bus here on the signing side of people who are way smarter, way more creative or way better at all of those things than I ever will be and I just have the honor of being able to work with them. And to see my job is tying everybody together and keeping them happy and just steering the ship until the right direction. I think they're doing a tremendous job of subsidising and putting everything together. But now coming to the sundries, I think the ability for us to raise this round was by, first of all, the problem and the vision again and then second, combining it with a team. They're a very, very strong founding team and in the beginning, I think not only maximise me, but many amazing other people who have joined us very early on who have enabled us to show to investors that we had a team that can actually build something that could pull something together and secondly, it's just the order of magnitude that this company can grow to. Because again, the problem is, on the opportunities in fact, the problem is very tough to grow. But the upside is very, very interesting and in that regard, we showed to investors that look, the problem is very, very tough but it's a past mechanism impact. So give us a lot of resources. We have a team that can utilise those resources effectively to then tackle this problem to get a good shot at achieving the vision and building something meaningful in space and then we're also lucky. We had investors that trusted in us and understood the vision and believes in us from early on and that was our job to make it worthwhile for them. Yeah, amazing. So you just got started actually and the most interesting part is what's next for Lissiem, what's next for you as a CEO, can you tell us what we can expect in the next month? So I think there's just so much work to do on the infrastructure side of things, actually building, financing, getting the assets up to speed, so having the bone of compute that can actually cater to all the demand that there is. And secondly, I think we have an incredible big challenge at building this orchestration layer because as I alluded to, it's one thing to get something up and running that people can use but it's another thing to make it reliable and scalable for an incredible big size and also optimise it. We have so many incredible ideas when it comes to optimising and features we want to implement and so it's just a iterating, building features, building infrastructure and basically never stopping, but it's very big. Yeah, sounds exciting. And since you are also in this big AI space with a lot of debates going on and we always like a hot take question here is if there's anything about the current AI debate that bothers you or that you think it's overrated on realistic? I think everybody is talking about a bubble and it's which money being brewed into this space. I think the problem or what actually needs to be looked at is not only the money, I think especially going to run out a lot of people talking about those enormous investment that being made which I think can be a very official way of allocated capital. If only one thing is true which is adoption of companies and it's actually also enterprise adoption. So I think the critical capability that people should be looking at is not, you know, the finances might be employed into the space and also financing structures, but it's looking at the adoption and looking at the growth of those companies. I think this is something that's unprecedented. The ability to generate value, the ability of iterating is so high that I think this should be metric people need to be looking at and what should be the metric for making those decisions. So very interesting. So Magnus, we have a tradition in our podcast. We always play a game with our guests and so if this game are that you get 20 questions that you have to choose between two words in each question and you only get one joker in this game. Are you ready? Yes, ready as I can be. Let's get into the questions. Snues are stand up. Stand up. Order are chaos. Chaos. Focus mode are multitasking. Multitasking. Coffee are tea. Coffee. Be a mentor are stay a learner. Stay a learner. Optimist are realists. Optimist all the way. Build your own chips are rely on Nvidia. Build your own chips in the in the very far future. Cloud are edge. But intelligence are imagination. That's a very tough one. I think you can't have one with or the other. I need to put a joke of that. I think it's a joke. All right, success or respect? Respect. Respect or happiness? Happiness. Happiness or love? At least the same, right? So friends, right? I'll say happy words. All right. Vision are execution. Execution and so executing with our vision is an added to tricky questions, tricky questions. Maybe I'll switch to vision on twitch to vision. Vision. Okay. Elon Musk or Tim Kug. Take off. Algorithmic beauty are human imperfection. Human perfection all the way. Rent or buy? Buy early morning or late night. Late night all the way. Silicon Valley or Shenzhen. Silicon Valley. Silicon Valley or Berlin. We were based in Berlin. So I have to say Berlin. All right. Also Berlin. And Berlin are unique. Don't say the wrong thing. Berlin. I'm sorry. I need to disappoint you guys. I think you'll have to say we're Berlin. All right. Thank you very much. Yeah. I think that was an honest answer or authentic answer at least. We have a last question which would be interesting because you seem to be a very interesting person who has also his own philosophy. Maybe you have one book or podcast episode or anything you read recently that changed how you think about building tech. And could you also explain why? I think one of the things that influence me is not that recent to be honest, but I think it was published in 2022 or 23 from Sequoia, the 600 billion dollar question away. I initially went to kind of the necessity of scale infrastructure for kind of the AI hype. There was initially 200 billion and it's going to 600 billion. That would even more. I think just shows the orders of magnitude of capital that is required in order to get a new technology up to speed. It's a very interesting article. And I think it's a bit outdated though. But maybe it isn't. So yeah, definitely very, very good read I can recommend. Okay. Yeah. Thanks for the tip. And any last words you want to tell the listeners of this podcast. Yeah, if anybody needs computers, everybody needs needs a GPU, want to execute something, they want to try something out, they want to try in a model, there's a recent or whatever. These results will be happy to make it work. It's a job to get it done as easy as possible. Okay. Perfect. Thank you for being here, Magnus. That's it. And you can continue your nightly working session. Thanks a lot. It's been so much fun. That's been a pleasure. Thank you Magnus.