The Exchanges

Q the last, whatever, six to 12 months. Yes. But there is a question that we will move to open source models locally hosted because the cost will be too significant for some of these enterprises to burden. And we're going to see that shift happen soon. If we do, That is both damaging to the providers, OpenAI and Anthropics of the world, and to Anebius. Why is that perspective wrong?

A Uh, yeah, so first of all, I think that, um, It's not in the future. It's already in the present. Uh, again, what we see in a lot of examples at the moment when our customer or the product builder gets to the scale, they start, uh, looking, uh, to the ways to improve the economics or accelerate the growth and so on. And this is the way when they most, a lot of them start to look to alternative models. So. The best way to build today is obviously to build on the frontier models from great providers like OpenAI and Tropic, Google, because they actually provide, and that's true, they provide you the best, best capabilities in the world. But then when you figure it out, the use case, when you start seeing adoption, when you see the customer data loop, you maybe can find the cheaper or even not cheaper, but more quality, high quality way to serve the same use case. You don't need, maybe you don't need the best in the world universal model, but you can Create the specialized model that in your particular case will work even better. And that's the way where you need to shift or may consider to shift from, uh, frontier closed models to open source. The most important kind of. Um, cause of those models is not just the open source, but they are tunable. They are trainable. So you can take them and you can do something. You, you can post train them and you can create the specialized model…

AI assessment note: “why doesn't it hurt, uh, Antropic and OpenAI? Because in reality, The, they move”

Q Speaking of kind of Jevons paradox and producing more in that yielding more demand, where are you not moving fast today? Where you would like to be moving faster?

A Everywhere. So when we think about how we build the company, we talk about it in the four dimensions. One dimension is capacity, like how much megawatts, gigawatts, and the GPUs we deploy. We are infrastructure company. We need to be large. If you are not large enough, nobody needs us to exist. So this is the, the, the, the physical world expansion. The team is doing amazing job, uh, but it's never enough and you want to move as fast as possible and Uh, there are a lot of complications of the real world that prevent you to move fast enough. Sometimes, uh, to launch new data center, you need to go through the entire, like supply chain, regulatory, and the fires and waters, uh, and, uh, uh, everything that happens in the real world. Right? So this is one dimension. Uh, another dimension is the product. So you want to move fast enough to address new types of the workloads, new types of the customers that coming to the market. Think about it. We started As a industry and in this AI journey, uh, from the people who first of all, built the models, right? So the, the, the, that was like companies like OpenAI in hyperscalers, large labs and so on. And what they need from you as an infrastructure provider is barely compute, like just throw infrastructure. And we see a lot of this large bare metal deals on the market. And we also do them Uh, but this is only the first layer of what we bu…

AI assessment note: “Everywhere. So when we think about how we build the company, we talk about”

Q If you had 10 x the capacity today, what would be different? Like, could you sell it overnight?

A Yeah. Yeah. It's a good question. Not overnight, but we would definitely, we definitely have demand for that. Uh, and I think the key question for us, it's not, uh, do we have demand or not, but how we actually build the portfolio of demand, because you have so many customers on this market that you can balance between. And again, to the point of four layers of the product, you can sell bare metal, you can sell managed customers. Managed infrastructure. You can sell inference and maybe in the future you can sell, uh, some new layers of product. And I think what we try to do is to build kind of quite diversified portfolio of, uh, customers. We, we, we believe that the highest tech we move, the more value potentially we can create for the customers. And actually the highest tech we move, the Bigger population of the customers we can serve. Because again, like on bare metal level, you have maybe a dozen of the customers in the world that you can work with. On, uh, managed infrastructure, there are hundreds. On inference, there are thousands. On agentic, there will be tens of thousands of new developers that build it, right?

AI assessment note: “Not overnight, but we would definitely, we definitely have demand for that.”

Q If we move to that second layer then, if we move away slightly from capacity to GPU hours, the product itself, multi-tenant, what is the main question that you ask yourself within that segment? If in the first capacity it's how much revenue concentration we have, what is the big question in that layer of value?

A What customer needs. It's a normal, you know, you speak with a lot of product founders, uh, uh, and this is the same. Like what customer needs at the end of the day, how customers evolve in their needs, where is the demand moving? So it's like, we see all this transition from training to inference. We see transition from, uh, uh, just using the models to building agents. Uh, and we see the transition from mostly AI labs, uh, consuming AI compute to enterprises coming in the game. And all the time, if we want to be relevant, we need to follow the changes. And this is the main question which like we, we ask us in the, in the product, like what should, what customer needs and what is Nebius? What is our value that we need to create? Because again, we are a small company. We cannot build everything. Uh, and we need to be very precise on what we can do better than others. And where the value that we should focus on, ah, given how customers evolve.

AI assessment note: “What customer needs... how customers evolve in their needs, where is the demand moving?”

Q Just do your fucking job. I, Joe, I know it's funny, I like it, huh? But like, what's the hardest part of just doing your fucking job today?

A Four dimensions. Uh, build scale. Build product. Work with customers. It's actually like two dimensions we discussed, like scale and product. The thought is customers. We are in the field business. We, cloud is the, we like to say that cloud is post sales business. When you sell, you sell the promise and then the customer, you need to satisfy the customer and working with the customers, covering the customers, having this strong customer engineering, like customer facing engineering team, FDE team. This is the third dimension. Go talk to your customers, make sure that they know you, that you know them. This is the third dimension. And the fourth, the most boring, but also the most exciting is the capital. We are in the capital intensive game and we're competing with the most capitalized companies in the world.

AI assessment note: “Four dimensions. Uh, build scale. Build product. Work with customers.”

Q Does the pace of model development sustain? Like you said that I would argue respectfully, you said every couple of weeks, I'd say every couple of days, there's new. Does that sustain? In five years time, are we seeing that level of iteration?

A Well, I don't know. It's a good chances that we'll continue to see a lot of niche models, uh, show up and improved. Uh, I don't know. Again, I'm a believer that we're quite far from the wall and we will see a lot of like, uh, models improvement, uh, happening. I think that what we also see is Much more new, like modalities and specialized models, uh, coming in game. So we speak about this frontier LLMs, but there is entire world of Life science models, robotics, uh, world models, video models, image models. Uh, so, and they all have their own use cases as well. And we see more and more like small specialized models, particularly use cases coming with like very much optimized. Just this morning, I spoke with a team here in Israel that develops, uh, uh, cyber defense Uh, foundational model, like the model that optimized for, to build, uh, cyber defense, uh, agents. And again, they don't start from the scratch. They, they take some of the foundational, like, uh, some of the open source foundational model, but then they train it. For the particular case optimized for the quality and the latency that needed in this like cyber defense use cases. And I think we'll continue to see it. We'll see a lot of specialized post trained models that still need, uh, optimized inference and optimized like, uh, infrastructure around them, uh, to let customer use them.

AI assessment note: “I'm a believer that we're quite far from the wall”

Q data center build out there, we're seeing more and more public angst towards AI. Eric Schmidt's getting booed offstage, um, not because of the content, but because of the AI inventions. Um, and we're seeing like public resentment towards data centers. I think 40 out of a hundred now are not being built when they go through planning and approvals. How do you think about and reflect on that internally?

A This is the environment we need to Working. So again, there are two sides of the thing. One is how we think pragmatically as a business. That's what I said. We, we think about it as a portfolio of the projects. We need to make sure that we are like oversubscribed if you want. And if one data center will be delayed, we will still deliver enough capacity to our customers. And most of the customers, they are not locked In one physical location. They just like, it's a cloud. Uh, we, we can build, uh, in different places and then bring the workloads where we have capacity. And, but this is the pragmatical side of the things. Then what we obviously see that communities and the local authorities require the companies like us to work closely with them. And explain and show what, what, what we do and work with them on their concerns and like address them. This is the reality. I mean, Uh, you can compare it, uh, when Uber, uh, started growing and in many places there was the pushback, right? So all what's happening is something new. We it's moving too fast. We didn't, didn't expect it to move so fast and so on. And I think that you, you, you go and work and you explain, and it's a, it's a, it's just a part of your, You, of your duty to engage and work with the new communities that become dependent on you and they have concerns and sometimes they just, they have concerns because they're n…

AI assessment note: “there are two sides of the thing. One is how we think pragmatically as a business.”

Q Penultimate one. Leo Aschenbrenner is a famous investor right now. Has huge cult following. Um, he recently disclosed a very large position for him. 5.3% of the company. I think it's 15% of his portfolio. How do you guys sit internally? Are you like, yeah, go Leo.

A I wouldn't say that we didn't like notice it. Obviously like everybody noticed it and like the, the, the, the, the stock jumped and it was a big news in the, around. Uh, again, I think that we take it as a justification of what we do. Uh, and then you, you got this justification, you say yourself, okay, those people, they give you a credit. That you will execute. It's, uh, uh, I come back again and again to what we do is post-sale business. Every time we sign a deal, every time someone invests in us, They give us a credit and opportunity to deliver, then go back to your job and deliver. And I think that we are in a such a market where emotional market as well, that you should keep, keep yourself like down to the ground. Remember that all this growth, uh, all these credits that customers give you, it's opportunity to deliver, go do your job.

AI assessment note: “we take it as a justification of what we do”

Q my partner before that We have is a thing we have to discuss, and it's within these layers, but you've spoken extensively about product build-out and the importance of building the product underneath capacity. When people look at you versus other NeoClouds, you know, we look at you versus a CoreWeave, you both run GPUs, you both have Nvidia relationships, you both have Meta as a customer, What's the difference?

A I don't like compare with others. The principles we build are full stack. We call it full stack integration. And you, you can think about it like full stack down and full stack up. Full stack down is we are really deep in physical world. We build data centers, we build racks and servers. We build the platform and the, uh, and when you control this kind of Uh, things downstream. You can move faster and you can squeeze more, uh, cost and provide more economically viable solutions for the customers. And then your vertical integration upstream is actually what we spoke about, like product and how can you follow the customer's needs and customer segments and not be limited by the Small population of the people that just need infrastructure, but really serve kind of enterprises and product companies, uh, with like meet them where they need us. And this is like, I think what we different and then how it, how it like showing up, I would say is again, uh, less concentration in the, in the, in the, in the business, more diversified customer portfolio. Uh, uh, we believe long-term Uh, better positioning for going to enterprises where we believe eventually a lot of demand will come from. Again, now most of our segment is working. It's AI natives working with AI natives, but we have a huge economics, a huge market of enterprises, existing companies, and someone needs to serve them and, uh, …

AI assessment note: “The principles we build are full stack. We call it full stack integration.”

Q Where is the most interesting area to invest today? Okay. I'm giving you four options. Infrastructure. Horizontal model. Vertical model. Application layer.

A Uh, I mean, we built infrastructure, so, uh, uh, we are quite happy here. I think it's a good place to be in the current world. I think that, uh, even though like we, we, for some extent we are building kind of the easiest part, not in a way, uh, it's complex execution, but we kind of know what's needed and our customers help us to understand what's needed. I think the most amazing people in this industry are those who take a risk to go and build, uh, end user products in my view, and they actually drive the most of, uh, uh, most of growth here. Like people who take a risk, like the real, like the real risk of building something people would need or not need. Uh, I think this is the most, the heroes. Uh, of our, like, AI journey.

AI assessment note: “the most amazing people in this industry are those who take a risk to go and build, uh, end user products”

Q explosion of models and the specialization of models, like you said that, and how many will be built and the depth across different use cases. Sadly, the one thing that is quite clear is that Europe Does not have anywhere near the model build out that we've seen both in the U S and in China. How important do you think it is that nations have their own sovereign models?

A Looks like the world is divided. Well, uh, we, we can, we, we may not like it. Uh, and I think that having good enough foundational models, uh, available for the big parts of the world. Is important. And I think here in Europe, uh, we, or at least at this part of the world, we should think, uh, how we Have enough capabilities available, uh, here. And I think that we had a lot of conversations over the last couple of years in like about the serenity and so on, all of this, like so sovereign AI agenda. And I think it was too much concentrated around like mega megawatts and power rather than on what we have on the build. Builders layer.

AI assessment note: “having good enough foundational models, uh, available for the big parts of the world. Is important.”

Q Can you help me on another one? We laughed earlier when we said about space. Data centers on planet Earth is a very difficult logistical build out. Data centers in space? I love technology. I'm an optimist. I hope it works. Is that fucking nuts?

A I think everything we see is fucking nuts. Uh, no, uh, so many smart, my, my, my view is very simple. Uh, so many smart people now working to make it happen. So most likely, uh, I may be less pessimistic that we'll see, I don't know what is there, like we'll build more in space than on earth in three years. My view, I'm humble enough to say that so many smart people are trying to solve this, this task and bring compute to the space that why wouldn't I believe it will happen? And I think there are a lot of challenges still, like a lot of, like a lot of things to figure out. But if someone would say, uh, say us that, uh, even three years ago that we will build like multi-gigawatt data centers and it, it will be like large interconnected compute clusters. Would you believe? I, I, I didn't think like that. And it's, we are here. It's, it's routine.

AI assessment note: “why wouldn't I believe it will happen? And I think there are a lot”

Q Before, before we move to number two being products, just staying on capacity, given the insufficient supply of capacity today, if you doubled pricing, would you see any change to demand?

A Oh, it's a, it's a, it's a difficult question. We actually raised prices like just a couple months ago. Uh, and We still, uh, still have fair, fair kind of pipeline pressure, let's say, uh, uh, on supply. And again, uh, if we, the question, we, we don't really know where is the balance. Uh, and I will tell you why. Uh, it's not only us being greedy and want to get like as much money and like, then people in the shortage will, will have to pay. For some extent it works. Like people needs compute to build. But then there is a point, and especially it's less in, in training because in training, it's like one off cost. But if you believe that we're moving to inference and inference is, uh, uh, is the cost of serving the customer, there is a, uh, a level where economics doesn't work and The economics of the products of our customers, if they work, they can grow and then we can grow with them. It's not like just supply demand situation and then absolutely elastic prices. They are elastic for some extent. Uh, but we also want to be meaningful and we want to be thoughtful kind of what our customers need. And by the way, it's not only GPU hour cost. It's all the optimizations you do, all the real, we call it TCO, total cost of ownership that you, like, and this is partially why we build the software platform. And I'm sorry, come back to product again and again, you want to speak about c…

AI assessment note: “We actually raised prices like just a couple months ago... still have pipeline pressure”

Q With Token Factory, you run on 60 open source models, and you've said before about cutting inference cost by up to 70% through optimization. Can I ask a dumb question, which is how do you actually make a token cheaper?

A Yeah. So the, the, it's not the magic again. You take the model. Uh, like some baseline model, and then you can optimize it for particular scenarios that, uh, that you have. So you can do. Actually, you can distill the model. You can make the same, like the smaller model that works, uh, uh, with the same quality. You can do spec decoding, you can optimize caching, uh, uh, and so on, so forth. So you take the model and out of this model, you actually build the system that in your particular case works with your, with, with your requirements, with optimized economics. And by the way, one of the things that, uh, also Um, I think important for customers to use managed platforms like Token Factory. The models are changing every week, every month, like right today, maybe, Minimax III was released, and there is Nematron Ultra that was announced, released. So, and this happens every few weeks, and every time the new model released, uh, it may work better on some benchmarks and Maybe not like on other benchmarks and so on. And you want to have flexibility. You want, uh, you want someone to support you on experimenting and actually adopting the new best models for your use case every time they come online. And then like the platforms like ours, uh, actually again, abstract from you, all the work that you need to do to actually like change from one model to another, to benchmark all of th…

AI assessment note: “You can distill the model... You can do spec decoding, you can optimize caching”

Q providers because I said, listen, the biggest enterprises want reliability, they want security, and most of all, they want ease. They don't want to be tinkering around with all the architecture and shit beneath the surface. What you're telling me is you're able to be all of that to allow them to pipe away from those providers And have a cheaper, better experience because you take away the plumbing, correct?

A Yes, but again, I think it's not about My point is closed models with open source models. It's not about like reliable or not reliable. Again, the work of the companies like Nebios to make possible, like, as you say, not think about plumbing if you want to use alternative models, but I think it's about capabilities. Again, I think that closed source models like frontier models are great and they will become even better. And they will solve so many problems that we don't solve yet. And we, we have such a diversity of the use cases we want to solve that there will be market for the smartest models of the world. The fastest models of the world, the in between models of the world, smart enough, but cheap enough. And you as a customer will be able to just pick the right, you know, the, the right source of token, uh, for each particular, uh, task and back to the agentic, uh, layer point, maybe It's even one to be the customer kind of task to choose the, like, which model to call now. It will be the engine that knows, uh, all the capabilities, like all the models under the needs. And then, uh, When you go to OpenAI and you do the research, you don't think in terms of how many loops you want it to make. You don't think in terms, uh, when it should go to LLM and when it should go to search. You don't think should it now call like which prompt to call, right? It's, it's happening. You ju…

AI assessment note: “Yes, but again, I think it's not about My point is closed models”

Q What changes are you seeing in customer needs that you're not seeing discussed much in public?

A Everybody's talking about this moving from training to inference. I think it just very, uh, uh, 100,000 feet like view, because this move means actually people, um, build specific products and in those products, they have their economics, they have their trajectory of growth, and it's not just like whatever. The same GPU is just used, uh, uh, for other purposes. I think it, it brings the new requirements. You, you need to build your inference platform. You need to help your customers not only run inference, but where the model that the inference come from. Everybody is taking open source models and fine tune or RL them. So how do we help them? And then when they run them, they generate a lot of data. How do we help our customers? Like when they already run their application, their inference to collect the data, to create it, and then use it to improve, uh, uh, uh, to improve the model or the application that they run. So it's a, people like this flywheel analogy, like, uh, you, you, you run inference, you generate data, you can observe this data, then you can improve the model, uh, that you run and Kind of continue, continue, uh, um, improve the quality, uh, of the end product. So I think, uh, there are a lot of pieces, both on system level and both on, uh, um, AI magic level, if you want. Uh, and I think the, the most fascinating moment for me is that I think what we see is th…

AI assessment note: “Everybody's talking about this moving from training to inference. I think it just very... 100,000 feet”

Q now, you mentioned earlier, I've interviewed some, some big people. I go and speak to some of those big people. A theme that did come up when I was speaking to them was the relationship with NVIDIA, and is a marriage a marriage if one has more power than the other? How do you think about the power dynamics in a relationship with NVIDIA when They have so much power.

A We look at this in a very simple manner. We just need to build what we build. Uh, we need to build, uh, our product. We need to tell our story and then, uh, the rest will compliment it. Uh, I think what is the most fascinating, uh, NVIDIA is still for big extent is an engineers driven company. And I think the best thing you can do To get respect from Nvidia. It's my read. Uh, they may have a different, uh, point of view, but if engineers in Nvidia respect, uh, your engineers, you will have the right foundation for relations, let's say. And I think that, uh, we managed to prove, uh, again and again, that we Know what we build and we have a strong engineering team. And I think that they see it and they respect it. And we have a lot of like engineers to engineers relations on physical, like on a, on a hardware level, on the software layer, on the inference platform layer. And the better engineers in NVIDIA think about you, the better, uh, Relations and partnership. Uh, I think it's enables. And, uh, and again, we may be maybe wrong thinking this way, but, uh, but, but, but that, that, that, that's what we see, like we can do. And, uh, we, we, we just focus on being reasonable and being kind of focused on the long-term value. It sounds like fluffy. Everybody say it, but, uh, Just do, do your fucking job at the end of the day, right?

AI assessment note: “if engineers in Nvidia respect, uh, your engineers, you will have the right foundation”

Q On the kind of customer portfolio, I love that for the capacity. Absolutely. You want to be big enough that you're meaningful, but not too large that the business relies on them. With that difficult awareness, where do you settle on what revenue concentration with a Meta or a Microsoft you are happy with?

A It's a great question, and I would say it's a, it's a main question of, Uh, our business, uh, I mean, not Nebius even, but the product category. And we always told it publicly and to our investors and to our customers that we believe that long-term strategy of Nebius is to serve as much diversified portfolio as possible. So we, we do the best to, to, to have many customers that we work with. We build the platform. If you, in reality, Again, to serve dozens of the customers of the world and like, uh, of the level of Meta and Microsoft, which super advanced and they have their entire software stack. They literally need only physical infrastructure. They bring with, they bring everything they have deployed on your infrastructure and run, right? Uh, you have a tiny, tiny, uh, additional value that you can provide them above the, the, the physical infrastructure. By the way, to To, to satisfy them with what they need on physical infrastructure is quite a challenge because you can imagine they are quite demanding and, and, and, and they need the, like the most scaled infrastructure in the world that exists. So, uh, uh, sometimes people say it's commodity, but it's not really commodity on that scale, like nothing commodity when it comes to the, to the real scale. But again, to your point, uh, uh, This is quite a small population of the customers that you can work with, and you not nec…

AI assessment note: “long-term strategy of Nebius is to serve as much diversified portfolio as possible”

Q How much more do you think Revolut will pet you in three years time?

A I don't know. I don't know. No, I don't want to speak about that. No, but I can say that like they, in total, I think they, they grow times, like they grow like this. We all see the AI companies reporting IRR growth, right? For them, it's not IRR, it's like their budget. But I think that the most advanced companies, their AI budget And it's not like this fake or not fake, like this more, all this token maxing kind of race. Uh, we see it like how they do it in the, in the production workload. So they grow the same pace, like this, uh, AI native companies reporting. They are growing their AI, uh, whatever, uh, consumption equal to their IRR. So the companies like Revolut, they're growing the same exponential, uh, trajectory.

AI assessment note: “I don't want to speak about that. No, but I can say”

Q Um, I want to do a quick fire with you. So I say a short statement, you give me your immediate thoughts. What job does not exist today that you think will be very common in five years time?

A One thing that obviously happening is we democratizing what people like called being developer, right? Now, each of us Can be a developer and like what I mean being developer is to convert the idea in some digital, digital asset. So, and I hope that again, we have to be optimists here. And I hope that, uh, this democratizing of building, like letting each of us being builder will open up so many opportunities and like that we even don't imagine yet when we will give Like millions of new people's, tens of millions of new people's ability just to convert their idea into something that works very easily. We will see a lot of new businesses and a lot of new ideas kind of just, uh, like coming in life and they will create a lot of new works that we don't even think exist. So it's like second, you know. The, uh, second orbital of, uh, uh, of all this kind of democratizing of the building. Also what is challenging and what will need to be changed. And I think it's like as risky as opportunity as risk as an opportunity is how the education will change because, uh, now when everybody has access to intelligence, what should people learn? You definitely don't need them to learn the facts. Everything is available. Like all the knowledge is kind of available. Like how do you really like train people to think when they don't need to think so much? How to teach people to continuously change? …

AI assessment note: “they will create a lot of new works that we don't even think exist”