Should We Buy NEBIUS ($NBIS) | STOCK TANK

All right, we are live. Uh, this is going to be a very fun episode. Yes, I almost died eating the food too fast. It is my fault. We have Sam and we have Manuel. We're going to talk about Nebas. I know nothing about Nebius. We have Roy, of course, who you just saw. And we have Tanner looking more and more like Freddy Mercury stream. All right, boys. Uh, let's get it started. What are we getting started? So, okay, I'm just kidding around. I think Roy and Tanner know way more about it than I do. I know nothing about like I don't know what the company does. Uh, so they're a restaurant. Um, they sell really good fried chicken over in Europe. Uh, so they're disrupting the whole industry with their vertical integration. Yeah. Yeah. They're they're basically bringing Rosco's chicken and waffles to Europe to an audience that has not been tapped in centuries. Also a demographic as well. So, uh we're looking at about 385% year-over-year growth. Bro, if I didn't just have food, if you got if you presented this like 30 minutes earlier, I would have been sold. Okay, that's it. That's the end of live stream. Live stream is done.

So, uh, I guess guys, tell us a what space they're in, what their business is, a little bit of their history, and then what differentiates them, like what what compels you about this play. Maybe we can talk, uh, you know, some earnings, uh, talk their moat. Save that one for a little bit because I'll have some questions on that, too. And man, just as far as full disclosure, I am a shareholder. Uh, I am also a shareholder. Um, I bought in like two weeks ago or so, but it's a small position. It's a investigate then in uh invest and investigate type thing. So, I'm not sure if this is a swing or a long-term position, but I I actually thought you were making a joke. I'm not a shareholder. I own Nvidia and uh Nvidia owns uh Nebas. So, yeah, they they do. They're very close.

So anyway, uh Manuel or Sam, whichever you one of you guys want to give us just broad. Maybe we could start with like the company's origins because I think it's important for the disease here. Like very important. So maybe I can start here. I don't know if you guys know Yandex uh the company, the tech company, the Russian tech company. Sorry Manuel, I don't think it's your mic. Yeah. Can you can you check if if you have the right mic connected because you might not. Yeah, maybe I Yandex was like the Russian Google, right? Yandex was the Russian. Go ahead. Is it better now? Yes. Way better. Okay. Oh, by the way, I'm share my screen. Sorry guys. So yeah, I was referring to um the Google of Russia, but yeah, just like Google, they had like several other businesses like in right hailing uh in cloud uh like self-driving divisions, etc. And so um after uh Russia invaded Ukraine in early 2022 as you know um many of the people from the company decided to leave the country and essentially start Nebus uh from scratch. And uh what happened is that there was like a restructuring process like very complicated one. And in the end what happened is that um the the assets outside of Russia ended up being for Nebuse and uh the the other part of Yandex was sold to I I believe to the government uh by like $3 billion or something like that. And that's why Nebuse started with a huge cash pile right from the beginning. And so this this is very important and you'll understand why later for all the advantages the company has because uh this brought the company like a thousand almost a thousand of highly skilled professionals in tech cloud AI that were used to build uh like everything together for like most of them I think the the average uh experience in AI literally in AI is like 10 years and to to build a team of 850 engineers years specialized in uh AI and cloud and that are working together for many years is like something that for instance core or any other cloud provider that wants to enter the space could not do like in one or two years. So I think this is really important the the the fact that they come from um Yandex and as I said the the assets that were not from Russia uh were for Nebuse which includes Nebus but Nebus was kind of built from scratch. They like rebuilt uh their AI their cloud services from scratch to purposely uh build them from uh for uh AI workloads and not really for the general cloud purposes. Then the self-driving division, they essentially shared the IP. So Yandex still has the the autonomous cars and autonomous and delivery robots in Russia, but outside Russia, um, Nebuse now has a subsidiary which is called Avide and it's actually like US-based. Um, and they have like the the delivery robots and autonomous vehicles like really really advanced. uh then tooka which is like a data labeling company triple 10 uh and uh 28% stake in click house which is an open- source database platform so nebios group when you buy shares in nebuse group you essentially own uh those four uh those five assets sorry and and manuel real quick talk about the valuation of click house and why that's significant because I think out of those that's probably the most important there yeah so uh nebus like not last week maybe two two weeks ago started to get more attention because Click House uh got uh or not got is getting a new valuation round it seems at a valuation of $6 billion and Nebuse owns 28% of the company. So uh before dilution that's that means like close to $1.7 billion and when that happened the company was trading at like $5 billion with $2.5 billion in cash and no depth. So if you had that click out stake, it was almost the other assets were like zero. But I mean I never account for the cash when I'm valuing NAB because they're burning cash like crazy like building capacity like Corore is doing like opening more data centers. So for instance they now have they they burn like last quarter like 500 or $600 million. So that cash is going to be burned pretty quickly. So I never account uh account it like in the valuation. Yeah. Also, um on top of that, they are h they do have a capex uh for the entire year at around I think 1.6 billion uh 1.6 to 1.2 billion and they only have about $1.6 billion left in balance sheet. So, as as Manuel was saying, last quarter at the end of those earnings, they had $2.5 billion, but a lot of those proceeds did come from one the $700 million uh capital raising, which included Nvidia last December, but also the sale of the Russian assets um Russian uh PI that they that they wanted to basically get rid of in order to get listed on the NASDAQ. So yeah, I mean even in an interview with the CEO Arcadi, he did say that and it's not even a matter of if, it's a matter of when uh they're going to raise more capital. How much capital that's going to be or when that's going to be, well then that's a big question. But I think that's the expectation that people should have. They are interested in something like this where if they see the stock down maybe like 10% after hours because they're raising like $700 million in uh capex, sorry, in in I don't know, let's just say 0% uh senior notes for 2030 or something like that. um to buy Bitcoin. No, I'm just kidding. They're not going to buy Bitcoin, but to invest to a data center. Um that shouldn't be a surprise, that sort of thing, you know. So, you have to come into the expectation that that is going to happen and there might be a potential impact on the price of the stock if that does happen. Uh I have to say actually, you know, I don't I don't want to interject. You can go ahead manual for the next topic.

So, no. Yeah, that's a like somebody looking at it from the outside in knowing zero about it. So I I know that the the the folks who founded it and a large portion of the company are coming from Yandex and I know that there's like five companies broken in into one. Uh what I would like to know is how does that shake out from an actual revenue perspective in terms of like of those four that they own plus ClickHouse is the fifth. Of those four they own like is there one that's contributing to like 90% of the revenue and the others are just other bets or is it pretty equal or how does that look like? Yeah. So, oh go ahead manual. Okay. Okay. Well, I was going to say that about 90% of the revenue does come from its core business which is uh Nebius the core infrastructure AI that's basically building out the data centers and that's where the bulk of the capex come from but it is also one of the least profitable business out of all of them and that's what the leveraging of the other four businesses are for but with the exception of click house. Click house is just minority stake in the company. There was actually talk during their earnings call say that they're weighing the possibility of maybe not right now but sometime in the future that they will sell their stake or at least trim some of their stake in click to continue to feed into their capex or potential capital exponential spend for the core business. In addition to that, I believe was that was that uh even a profitable last quarter? I forget the exact number. Okay. But Avride is probably going to be the next part of the business that actually is going to become ebida to profitable. And at the end of the day, all these businesses are there to serve the purpose of funding their core infrastructure business uh which is growing at a phenomenal rate. And that's mostly because the comps are pretty easy, but also because of the expectation that they're going to reach that uh 5 to700 million in terms of nominal revenue uh by the end of the year. Yeah. And these these subsidiaries uh today they reinforced in the earnings call are very important uh for the funding needs that the company has uh to expand like Everide the autonomous uh division I mean this this is like like nine years of work because this uh the self-driving division of Yandex started in 2016 and AID uh most people don't know about it but it's actually one of the most advanced autonomous like driving companies in the world. They will have like 100 robo taxis in the US like later this year with a partnership with Yundai. Uh they have over uh they will have over a thousand delivery robots like by the end of the year as well. So if you look at the peers this company this company alone could be like if it was like traded in the private markets or something like that. It could be tra traded at like five to10 billion dollars alone. That's the valuation of that is Tesla. Actually what my vehicles uh in August. So sorry Tesla will have 10 to 12 vehicles that are autonomous in Austin in August. So we're talking I mean we can't really compare obviously we can't compare Whimo or uh Tesla robo taxi to avide mostly because of the scalability when it comes to those companies as far as like that also being like that's what the company is about you know Tesla or Elon Musk about robo taxi right now um that's what is like that's it separate business entity that is focused just on that for this one they're kind of splitting the amount of resources in terms of how much you're going to spend toward it. By partnering with Uber, partnering with Hyundai, and then Uber Eats, GrubHub, and Rakudin, they're also leveraging their existing clientele as well as partnerships to be able to push out this product. So, in a sense, like they're not building their own cars, they are designing and utilizing their own stack to be able to build the the, you know, the autonomy that has to go with the robot, the delivery as well. uh but on top of that they're also utilizing the engineers from TOCA and that expertise to contribute toward a so in a way like they're all symbiotic in a sense where like all these businesses are contributing toward I would say the contribution of profitability to in the end feed into Nebius the core infrastructure business so it's kind of like an interesting flywheel per se like the more revenue that Nebius generates is the more money that they can invest toward Nebius but as well as the subsidiaries which is going to create more profitability which is going to feed back to Nebus and so on. But that also coupled in with this great demand in AI compute that uh Jensen Wong has talked about for years at Nauseium which Tavis knows about pretty deeply and also with the demand in future technology with AI not just in LLM but also with other software that are continuing to build an AI stack and develop uh develop um aentic AI uh revenue which includes Service Now and Microsoft and other companies as well. All this demand put together is really just going to drive the incentive to invest more in the company from a sovereign fund perspective like when Trump was visiting the Middle East last week uh to even private investors and even public investors like ourselves. Yeah. But here it's also important to note that the the subsidiaries itself will end up uh like vanishing not vanishing like literally but like they will deconolidate from the the core business at some point. For instance, Tolloka recently had an investment by uh Bezos Expedition Expeditions which is the investment arm of Jeff Bezos and the CTO of Choppify which became the chairman of Toka and essentially they're now uh they voluntarily like gave up on the most of voting rights because they don't want to spend their resources in like scaling to loca. They just want to have like uh they still have a majority economic stake but they like uh as I said voluntarily uh gave the most of voting rights and they will do the same uh with a just want to have and and to find a partner that can really scale this to another level and they said that they have been like actively discussing this with third party uh with third parties I would say like probably uh one of its partners maybe Yundai because Yundai has like other investments in other uh AV companies like motional and other companies. So I suspect it could be like Yundai or Uber or something like that. Uh so it you could see uh as Sam was saying that they all contribute to like a a closed loop system let's say but this will this is not the end goal. The end goal is to leverage these subsidiaries to be able to fund uh the core business of AI without um without uh like essentially minimizing shareholder dilution and also not having to incur in like huge depths levels like core for instance that has some depth attached to like GPUs and GPUs depreciate at like really high levels so it's not really a good depth to have on your balance sheet And today like during the earnings all was really really reinforced by the management team that uh they don't want to like dilute like crazy like other GPU clouds like are usually doing they really they will inevitably have to dilute like some offerings or something else in the in the way especially if they don't have um a good funding source for Avat let's say because as I was saying the peers I Even if you use the lower grade peers, obviously we won't compare Everride with Whimo or Tesla, but if you just look at the other companies in the field, you could say that Avride is worth like $5 billion and it would it could be even be a low-end range. So they could even uh get like one, two, three billion from um selling a a large stake in the company and like remaining uh attached to some uh economic upside. So I think that the and as Arcadi said the CEO this is the only um company in the space that apart from having the technology uh competitive advantages that we will see later uh they also have funding sources like non-dilutive funding sources that its peers uh don't have simply okay so a couple people ask about profitability remind me but I think that they guided for EBITDA profitability in the second half. Is that correct? Yeah. So, um they did guide for that. That was actually new to me. I don't think I've seen that before. I think they were saying actually like maybe at the end of 2025 uh or even 2026 they're going to reach adjusted EB profitability uh for the Nebus group. We're not talking about the core business. We're just talking about the group as a whole. Um but at the same time but they did mention the the in the sorry in the earnings they did mention that the core business itself like uh separating the subsidiaries will likely reach in uh the third quarter like a beat the profitability that's good yeah I mean also keep in mind you know that I mean I've invested in a lot of software companies um the adjusted metric is not a gap metric uh so by standard accounting practices or general accounting uh practices is it's not you're not going to see that uh gap profitability that you like Palanteer or Crowd Strike or other software companies that might reach a gap profitability versus non-GAAP or adjusted right they are taking a lot of metrics out of there in order to have that on there so you know that's something that's mindful for people who are looking to invest something like this that this is not a gap profitability so on so Wall Street might not take it that way um but it's certainly good to say they're heading in the right direction I think that's the takeaway we need to have from reaching positive or profitable on an adjusted IBIDA basis. They did say that in the medium term, which we don't know how far the medium-term is, that they're going to reach adjusted again with the adjusted operating margin of somewhere from 20 to 30%. Which is pretty good when you consider EBIT margins. Um, if that does translate into gap profitability, that would certainly be good. But, you know, this is data center expansion. This is not a cheap out. They're building uh an aggregated amount of 22,000 Blackwell GPUs in two different uh two different data centers. So, two massive clusters. And on top of that, we saw NVIDIA's uh Computex, I think that was the name of it, Computant last Monday where they were talking about NVLink Fusion coming out. Um, they're already talking about the Reuben models that are coming out next year. Um that's on top of the new models that are coming out 2026. So when I say next year, I'm talking about 2027 and so on like and clients who are using AI technology, you know, especially when talking about the mega caps, they want the latest and the greatest, right? They're not going to be like, you know what, those hopper GPUs are just fine, you know, even though yes, they can supplement, you know, the latest and greatest, which right now is the H200 for Blackwell uh with maybe Google TPUs using MV Fusion, uh or maybe AMD chips and so on. You're still going to want the best of the best. And I feel like that's what Nebius is delivering to to their clients being that they are a client of Nvidia. And when we think about this from like a weight list perspective, right? Like you can't just go out there and be like, you know, I want to buy a Blackwell GPU. Like no, you got to go on a wait list, buddy. Like think of it as like you're waiting online for a club, right? Teis, you know what I'm talking about. You're waiting online for a club. All right? And you you want to go in next like five minutes, 10 minutes. Like no, buddy. You got to wait in the back of the line. The bounce is going to stop you and so on. But Nebius has that $20 under its pocket or it's in with the bouncer or the owner of the club. They could jump to the front of the line and they could get those GPUs on a faster return. And that's I guess that's one of the benefits of a larger conglomerate who's building the best ships in the entire world having 4% stake in your company because they want to see you succeed, right? They don't want to lose out their money. So that's something that's actually very interesting uh as far as something that uh is with Nebius. But, you know, there's also Coreweave and I think later on we can touch on like what the difference between Coreweave and Nebius is. Uh, sorry, Nebius is because there's a very big difference between those two and I feel like a lot of people kind of confuse those two together. I I confuse the two. Yeah. No, we're right now. Well, before you do, it's it's kind of like a bridge to it. um talk about the software a little bit and and the stickiness of the product and the mode too because it's in part, you know, I'm asking this question because I like don't fully understand it and I'd like to u but I think that that's very relevant for this piece because right now the TAM is so immense and accelerating the demand is insatiable. They don't have to worry about stickiness and competition yet but it will come in the future. So yeah.

So um in my opinion the the main thing that enables Nebuse to has some differentiating factors is what I said like the engineering resources that they they have uh from uh Yandex essentially the team the talent because first that enables them to vertically integrate the business. They design everything like everything from like they have in hardware like from servers to racks to everything. They design they they uh send to an OEM on Taiwan and they build it they they build everything from scratch according to their design and this is the only nail cloud because uh the business model of these GPU clouds is called like no cloud. They are the only nail cloud that does this. This gives them a strong structural cost advantage against anyone. And then besides besides being uh more cost efficient, they are also more energy efficient as you're seeing like Sam is uh sharing. You see the power usage effectiveness is like on par with like Microsoft uh even better than uh um AWS is not there but I think it's like 1.15 as well. A AWS. So it's kind of bestin-class just like the hyperscalers. So with Navios you get hyperscaler level with more uh flexibility more reliability not not more reliability but um um more flexibility in terms of contracts more uh for instance I saw uh an interview with a client which was saying that they they changed to Nebuse from a different uh provider because they essentially were getting AWS performance with a a user experience like a UI that was beyond whatever he had seen in uh pre in like other hyperscalers even Azour uh Google cloud uh AWS. So I think um and this could be like uh difference uh from other perspectives but in my opinion the engineering resources allow them to be not only uh not only have like the structural cost advantage that I was saying but also to know what developers want to use and to have a customer centric approach that developers really really love and the most important thing that I saw in several interviews was not only reliability because they can get reliability in uh hyperscalers but as we can talk later like hyperscalers most of the times don't really want to work with those emerging AI native companies they want to work with large contracts they a lot of the the Nebus customers right now are not like open AI which is working with cororeweave and just wants bare metal because Nebuse has and this is another advantage they have a a tailor solution like a tailored solution to each customer. If you want to do something, you first talk with Nebu's team and they help you to uh have essentially the best solution for your needs. While for instance with corewave, they mostly offer you the GPU cloud and then I mean go ahead and make yourself what you want. You don't have the value added services of the software tools that they design. But maybe I'm um diverting a bit. I I would say that first hardware design they they are vertically integrated and they have uh not only uh the the racks but like all the equipment that makes a data center. They designed the data center from zero to 100. Then they have an AI cloud platform that they purposely built for uh AI native cloud uh computing environments. They literally took like between the the restructuring and uh the scaling of the business which started like three or four quarters ago. They took like 12 months rewriting the code like the engineers were like in a room for 12 months rewriting the code to develop the the AI cloud from scratch instead of using like the the general cloud purpose that they had from Yandex. They literally like get rid of got rid of that and uh rebuilt in during 12 months rewrote the entire code from scratch to be like specific to uh AI workloads and not like for instance AWS and Azour and uh Google Cloud. Uh of course they're amazing as well but they were not built from scratch to AI workloads. But then the only thing that I keep going back to whenever you say that because I know like I I agree 100% with you but it's the coreweave over uh Nebius conversation that I don't understand the differences between like first Cororee is not vertically integrated as I was saying they have uh they uh rents third party data centers mostly uh second they don't have most of the value added services that Nabius offers like in software terms and they develop tools for uh developers to have like a better experience and uh to have like access to uh for instance to every single language learning model that you have like Mislig a lot of models like uh Nebuse has built like an inference as a service platform to help developers um then the essentially because of not being vertically integrated they don't have the same flexibility that Nebuse has. For instance, uh if um Nvidia sends you Blackwell GPU like the most uh recent version, you will need to uh buy first the racks, the servers from let's say Dell, Super Micro, whatever. And Nebuse as like as soon as they get the GPUs, they can just put it on the data center and start to work with that. Um, I mean, Sam, do you want to start with coreweave like everything? Yeah. So, core Oh, go ahead. No, I was just going to say because we were talking about the power usage before and the overall performance, but if I'm not mistaken, coreweave outperforms Nebus there. So if you are just working on specific AI workloads, uh, and you just need GPU processing, GPU as a service, that's the key word here, G GPU compute as a service. Nebus doesn't offer just GPU compute as a service. It offers full stack. So if you want to run like a server or something in Nebas cloud, you can go ahead and do that. I haven't used corre personally, but it's not necessarily the option that uh let's say a smaller client would use uh to run all the workloads and servers and basically like an entire application on coreweave itself. It's something that you would go to Nebus for, especially as Manuel is tying into is that you can make a deal with Nebus with that. When it comes to core, if you're even one of the major hyperscalers, it's very difficult to do that. It's more like sign up for what you got. Like if you go to if you go if you ever use AWS before, you can't like negotiate terms or anything. You you you just go on there, you spin up an instance or an RDS instance, whatever it is, and you just use it. You get charged hourly rate or you spot instance, whatever. Nebus does deliver that as well. But core wave is just more on the GPU perspective like you're kind of you're renting out the compute like you're sending off your workloads to core to run it on there and then you're getting the data back. Nebas you're just running the entire application in their cloud. Yeah. So the the clients that want core essentially it's like open AI like we don't have capacity and we need like Sam Alman is always saying that they need more GPUs right that's why they now sign a contract with cororewave because they essentially don't want the the full stack that neb or they want but they nebuse can also just provide the bare metal just like cororeweave does but they want to integrate each customer in the the software layer that they have and the tool the AI tools and the application tools that they have because that that's what will drive what uh Roy was saying like the sticky customers and that's what they touched during today's earnings call that essentially all their customers are using the software uh services they now have been like um uh essentially offering those solutions to integrate the customers on the platform and then over time as customers get stickier uh there this is becoming a high margin much higher margin uh segment like the software part inside and the itself sorry and core sorry go ahead well I I was going to touch just not only on coreweave but also if you look at this if you look at this table on the right over here you have your bare metal suppliers over here where they do provide compute but on the bare metal perspective when I say bare metal I mean like literally you are using just one server that's on the rack when you think of cloud perspective like you you could be running the same application next to um let's say another small company like I don't know what example to use I think off the top of my head but let's say Cler is running a software piece of software on a server in in in a cloud right you could be running on that same exact bare metal hardware so you're not technically renting the bare metal server you're just sharing compute in a sense coreweave is basically running GPU clusters in a GPU in in a collocation or server they're renting the GPUs themselves and you could be running models on one of the GPUs or GPU clusters that a company next door is running the same models on. So then you come into the issue of security when it comes there in compliance. I mean working in it like that becomes a big issue for certain companies especially banks. You don't really have to worry about that as a startup is running marketing platforms and running running algorithms and running uh running uh data processing queries on marketing marketing demographics. But if you're running like proprietary data of customer addresses, phone numbers and so on, you don't want to mix that data. You want bare metal or you want to be on a separate uh let's say a separate private cloud versus a public cloud. And Nebius is able to offer basically the whole stack as you see here. Not only bare metal, not only GPU clouds, but also whatever the hyperscalers offer, they're able to offer all of that. Coreview is literally just one horizontal right here. GPU clouds, right? They don't really go outside of that space. So like that I think this this table right here kind of shows you visually as to what is the difference between Nebius and Cororee and that's not to say that Cororee can't you know can't certainly be uh successful in the business that they work in or the single vertical that they work in but at the same time Nebius is like literally all of this right so they have the optionality to focus on certain parts or maybe uh maybe excuse me scale in certain parts or scale in the bare metal if they don't see profitability if they don't see a lot of revenue there so they can focus more on the hyperscaler portion where they're offering the entire stack on there and so on. Sorry, some of these comments are like are like really funny here. Uh there was one guy up there that was saying like oh vertical integration that sounds very smart time to go full port Nebius like okay anyways just to kind of like give give the overview this because I feel like we're focusing too much on uh specific parts of it because I think the the pitch for Nebius is Nebius as a whole like the entire Nebus group not just the difference between Coreweave and Nebus and other hyperscalers and so on. I mean that that provides a cost advantage and keeps customer retention in but at the same time I think the question will really come in is if we start to see let's say they're missing their numbers that they're guiding on their uh on their ARR or maybe their customer retention is starting to decrease or maybe they're not going to reach profitability at a certain point. I'm not saying that those those topics are not important because they certainly are. But I think like to kind of take a step back and look at as a whole is where you put all these pieces together. So, uh, I have so many questions and I've been sort of jotting them down on the side, Sam, but since you were just on that point with regards to coreweave, I guess my first question here is with regards to other players in the space and especially hyperscalers that have way more resources. I mean let let's put aside the competitive advantage for now because I'm sure you guys are going to talk about it but does Nebius have a target audience a target market uh carved out to say well you know we'll let Coree take this particular part of the market and we want to focus on just this type of user just small businesses or just you know SAS companies or or whatever the case may be or like have they not really talked about customer segmentation what part of the market they want to really dominate in And why?

So um first um Nebus can uh like accept any kind of customer from small customers to huge customers like they as Sam said they really own the full stack and they just do whatever you want to. It's literally like literally that you once bare metal as he was saying you go to Nebuse and you can have bare metal. But to your point, right now um the first wave of customers for Nabuse are the AI native companies that are emerging right now. For instance, the the text to video uh websites that you see that are now appearing like Xfield AI, I don't know if you know it is a a huge client for Nebuse um and other uh emerging companies like those. Now um Nebuse is under like constructing a huge data center in New Jersey which will have up to like 300 megawatts capacity and after having that capacity the company already said that it will probably sign large contracts for instance those that coreweave was signing with openai Microsoft and something like that. Um and then after that there there will be like a third wave of customers which will be the enterprise ones because right now yeah AI is being implemented a lot but when you see like the percentage of corporate clients that you like you can have in two three four years as the adoption of AI is scaling is huge compared to now because like only a minor percentage of companies around the world are using AI or implementing AI to whatever you like needs they they might have. And here will be a huge advantage of Nebus which is as I said you just go there and you ask and you you you say whatever you want to do and Nebuse will have a tailored solution to you. It's not just uh you're going after a GPU cloud and you want the resources. When when the third wave of customers come, the enterprise ones, they will want a tailored solution for a specific problem for in just like Palanteer does, right? Like not like Palanteer does, but I know that they're really specific to the point to each customer and I think this is a huge advantage here. And then there's also a fourth wave of customers which is the the national AI initiatives. For example, last week uh Nebuse was selected by like the Israel Innovation Authority to build the country's national supercomput. So there was like four companies uh with like proposals to that and they chose Nebuse because it was essentially the most efficient one, the most uh not only efficient but like even cheaper because they can offer the the same reliability and this is something that customers and uh especially as I said now that they have mostly smaller customers like smaller AI native customers they always say that they offer the same reliability as hyperscaler. and corewave and etc but at the best performance per dollar possible and that's essentially because of the struct structural cost advantage that that I said Nebuse has because of the vertical integration and that's exactly why I was saying that uh in my opinion the the most important part here is the engineering resources that they have because that's what puts everything uh together and what enables everything and uh they not only have the ability to have more margins like higher margins but also to pass on the the savings to the customers and that's what's happening right now. they're sold out essentially now their capacity is not too large. That's why they're like building a huge amount of capacity in the short term short medium and long term. And that's why they're sold out just like as you uh were pointing out like hyperscalers hyperscalers sorry uh often uh don't don't um want to have a specific kind of customers first. They don't even have capacity for themselves. You're always hearing the hyperscaler saying that we don't have capacity enough. Next year we'll spend even more in capex. Oh, uh the Amazon co is always saying every time the AI cloud of AWS gets additional capacity, it immediately gets sold out. So like they don't have capacity for themselves, they don't even have capacity for more clients. They're always like trying to keep up with demand. Demand is literally exploding. So as the co of Nebu says like there will always be 30 to 40% of the GPU markets uh attached to Neoclouds which is essentially Nebu score with Lambda and uh some of the names you've seen um SAM sharing and within those let's say not even 40% maybe that's too bullish let's say 20 to 30% of the GPU clouds markets uh Nebuse being the leader and maybe you can count cororeweave as a leader but I even think that in the long term it might uh even uh overcome um core wave but even if it's just the second player of this market the the opportunity is like multibillion dollars in the short term and in the medium term even more so I don't think clients or demand will be a problem and I don't think nebuse is specific to a wave of customers right now they cannot um be uh offer solutions for instance to open AAI or to Microsoft as I was saying that they rely on core because of the huge capacity they have but as they're scaling uh the new data centers specifically this one in New Jersey which is very important and we'll have like all the newest versions of black wells and blackwell ultras and all that I think it's pretty possible and actually likely that they will start signing those large contracts that are attracting everyone to corweave and oh My god, Corweave signed the $4 billion contract with OpenAI and all that, but to be honest, I don't think they need that. But I think that will be a strong catalyst at least for attention. And I I would prefer like more uh stickier customers within the software layers to be honest. But I think that it will be really something that would validate the quality of Nebuse because I I think a lot of people doubt about Nebuse just because they don't have yet a customer like uh Core has like Microsoft and OpenAI but that also uh diminishes like the risk of customer concentration because Core right now like 60% of revenue comes from Microsoft and Microsoft is building capacity like crazy. Imagine like in three or four years they just say to Core with I mean screw you I don't want really I don't really need your capacity anymore I just or I just want like off of that and they just have an a huge hit on revenues and that's one of the things that I believe is much more stable on the business model of Nebuse uh rather than corewave because cororeweave just rushed to build capacity they got in depths like over $10 billion in depth just because they wanted to rush capacity for mic Microsoft and Nebuse is really taking a more pragmatic approach. It's first starting with the first wave of customers and has it scales up. it will reach every point of demand but on a a more like having a more pragmatic approach as I was saying scaling based on demand and with all the the engineering abilities they have they have like the the custom um hardware they can slowly uh adapt the the capacity as demand scales and not just like try to get $5 billion in depth and like build as much capacity in a year. No, they're like trying to be uh cautious with investors money and as I said try to take advantage of the other funding opportunities that they have and other companies don't have. Do we want to um get into some of the questions the chat has? I I've got a few questions um just in terms of um you know seeming like they're tackling every problem. I'm not saying that they're not, but like um I wonder why the hyperscalers aren't taking the route for the future of data centers and and also trying to vertically integrate their new data centers. Not not necessarily retrofitting the old ones, but essentially just building out a separate uh Nebus to to compete. I mean they probably there probably are but as I was saying the many customers don't uh not even once but also don't uh the hyperscalers themselves don't want to accept specific customers and the flexibility of neuse is really different from hyperscalers or coreweave for in if you go to an hyperscaler and this is something that I um that I learned from uh interviews to customers like Nebio's customers if you go to AWS or to Azour or to a hyperscaler let's say and you want to um to have some GPU capacity you uh most most of the times if not all the times you need to commit to a specific amount of capacity during that contract for instance you want a one-year contract and you you have to a clause that say you need to at least um use x amount of GPU's capacity or you'll pay like much larger fee or something like that. And Nebuse has something that very few clouds I'm not I don't even know if other clouds like minor clouds like lambda or so offer but corewave and um hyperscalers don't which is on demand GPUs like you can literally pay as you uh as you use the capacity let's say um sorry I remind me of your question because I why would or I'll just build on what you just said if If that is what those customers are looking for, why doesn't AWS enable that style of pricing? Exactly. You were asking about hyperscalers building the same thing. First, uh they don't have capacity for everyone. That's the first the first part. I mean you everyone can say that oh hyperscalers are like way better solutions and they will take all the markets but that's impossible. As I was saying, they don't even have capacity for themselves. like even more for the the other clients it's simply impossible and they are always prioritizing their themselves rather than uh other clients and that's for instance what Microsoft is doing right now uh they're now train training and and even Google they're training Gemini they they even need uh more capacity than they are building so the first point here I would say is um the the the the lack of capacity the supply demand imbalance. But then you can say okay five years they're all buying from the same suppliers. So the the cut like wouldn't it to me it would make more sense that if uh Nebius and uh you know Microsoft go to Nvidia and say hey we need some new GB300's that Microsoft gets their order filled and Nebius waits. Well, even even just building on that, like why can't the supplier themselves, like Nvidia in this case, spin off a portion of their supply and say, "Okay, for 90% we're going to be the provider and sell you the actual GPUs, but for 10% we're going to do Nebus' business model and essentially allow you to to to use it." And then all of a sudden you have a new competitor who is also your supplier. But which I think he's saying that they they are going to do which I agree but how are they going to grow in market share faster

than the existing hyperscalers. Well, I would argue that the market share as a whole, or the TAM as a whole, as it is this year, is growing much faster than being considered that there's enough for everyone to, or I would attribute it to like, there's enough pie for everyone to eat, and the pie is growing every single year.

Um, I'm not, I'm not suggesting that that's not the case. I'm just suggesting market share, like they're taking more of the pie than they did last year. Like, what was their revenue growth this year? Like something like 400%? What are you talking about? It's growing from a really low base. So, yeah, the comps are a lot lower. Uh, but ideally, what you want to track is not the trailing revenue. What you want to track is the annualized run rate. So, you're basically taking how much they earned in the month and you're just multiplying by 12. What they had for March was, uh, $250 million, and that's what they had in their earnings report. What they, they already gave commentary. They even had a slide on it. Uh, I'll bring it up in a second for April, that annualized run rate for just that month alone grew by about 24%. So, it's growing, and it's accelerating, in my opinion.

Yeah, but that's faster than the pie is growing. That's market share. That means then that they're taking market share. But there's a seasonality effect to all of this, right? So, if they grew at 24% every single month, they wouldn't say they'd only reach $1 billion by the end of the year. They're saying they're going to reach $1 billion by the end of the year because that means that as the compound monthly growth rate would be about 15.77%. So, that means that they overshot it in April. But I think what they're also including in there is seasonality, that you're not going to have all this drive for compute during every single part of the entire year. But that's the reason why stocks move, because of the future expectations. So, if they come by next quarter and they say like, "Hey, we're actually going to guide more closer toward the upper range." Last this quarter, they said that they're going to be, they reiterated the same range. But if you look at the fourth quarter, it was from $500 million to $1 billion. Last quarter, they guided for $750 million to $1 billion. This quarter, they put it in line. So, we're already in the first quarter of the four quarters, right? So, next quarter, that's when we'll get a better picture of what's happening. But I think a lot of that is attributed to the fact that, well, if you look at Google, out of all the hyperscalers, Google and AWS actually decelerated in terms of their growth. Microsoft accelerated, right? So, that is definitely going to be contributing toward CoreWeave when you think of that, because Microsoft, like Manuel was saying, was that 60% of CoreWeave's revenue is from Microsoft, right? So, if Microsoft's accelerating their CoreWeave, by proportional level, it would accelerate as well.

So, what happens if Microsoft comes out and they say, "Oh, our acceleration is lateral from last quarter. We still grew, but we didn't accelerate our growth." So, then now you've got to turn to Nebius, like, "Wait, hold on a second. So, then where did that market share grow?" If all the hyperscalers didn't accelerate, but Nebius accelerated, where does that put them in terms of the market share? So, it's kind of like a moving target here, right? But what I think, what we're both trying to get at, is that there's a reason why the stock doubled in the last, well, because the whole market was down, that made a huge difference. But there's a reason why the stock kind of doubled at a certain point. I know a lot of people in the audience are asking questions like, "Hey, it's trading so expensive, right? So, they just earned $54 million last quarter, and right now they're about a $7-8 billion company. I haven't really checked in the last couple of hours because they move so fast. But that means that they're trading at like, what? So, then that's 20, and then times, that means they're trading at like a hundred times, right? Which is ridiculously expensive. But what you need to look at is their annualized run rate that they're guiding for at the end of the year. They grew last quarter at nearly 400%. So, then what if they grow at 400% next quarter, which I'm not going to say they're going to, but they might be growing at, let's say, 300-something percent or maybe 200-something percent. So, then now their valuation gets a lot cheaper from that perspective, assuming that the stock stays stagnant, right?

But I think at the end of the day, what's really guiding the price with a lot of these companies and the valuation is not so much like what one company says. Because when I look at Nebius as a whole, I'm not just looking at Nebius. I'm looking at what all, what the commentary and all these other chip designers and fabless, and even the actual fabricators themselves, like TSMC, are saying is that demand is not, the supply is not meeting demand. So, therefore, they need to invest more capex and build more. And when you think of the amount of capex, like, let's say Amazon has an earnings call and they increase their capex, Nvidia is going to jump, right? Because guess who's going to be a large receiver of all that capex? It's going to be Nvidia. But then you've got to dissipate that capex across all the data center plays. And that's the reason, I don't know if anyone noticed, but Astera Labs, which is basically contributing to the designer and also they're, they're basically getting all their networking equipment done on the back end with TSMC and other companies as well. They were up like 5% today, right? So, I think that Nvidia, and this is my theory, right? Because I don't own Nvidia anymore. And of course, you know, I'm going to talk my book because my money's on it, right? I think there's a lot of opportunity in terms of opportunity cost when you think of these other companies, and not just Nebius, but also Marvell Technology and also Starcom, and also AMD, especially because they are mispriced in my opinion, and also they're not a $3 trillion company, right? So, they have a larger runway to go. When you think of Nvidia in terms of its growth rate and in terms of like, basically, you've got 50 million analysts looking at Nvidia every single day, running models, looking at its growth rate and so on, talking to all the clients and everything. I think it's really safe to assume at this point, whatever the street knows about Nvidia is pretty much already priced in, unless we hear new information from other companies. But Nebius, you only had like three analysts on this, right? So, it's kind of hard to tell like, "Oh, they missed their revenue or they missed their EPS and stuff." Well, because there's only three analysts covering the stock, right? Including five analysts here, I guess. But anyways, go ahead.

I understand the viewpoint that, like, okay, they're going to grow faster than Nvidia, because they're $8 billion and Nvidia's $3 trillion. I get that. But I also, like, I disagree with this comment. Like, just because Nvidia has a vested interest in Nebius, it doesn't necessarily mean that Nvidia can't disrupt them. Like, you better believe that if Nvidia saw value there to actually move into that space in a big way. Like, yes, there could still be a market for Nebius. Yes, they could still be growing. But at the end of the day, what comes to my mind, and it's a little bit different, it's not apples to apples. But what comes to my mind is the whole like Microsoft Teams Slack thing, where Microsoft Teams, even though they released an inferior product, because it's part of that wider ecosystem, they can bundle it with their other product offerings and still win out market share in that sense. And so this is, I think, one of the benefits that these hyperscalers provide, is because they have much deeper pockets from a resourcing perspective, and they have much deeper, you know, connections from or opportunities from a bundling perspective as well. So, I guess that's my wider question: Is there anything that's stopping, you know, what is the core differentiator that Nebius might have that you can say, like, "Oh, well, it doesn't matter, Nvidia can enter the space, but they'll never make a dent because X, Y, and Z?" Or are we always going to be waiting for the other shoe to drop for all of these other larger companies to essentially become the perfect competition from like, the monopolistic or oligopolistic competition that it is right now?

Well, I think when you think in terms of competition's coming, competition's already here. And I would argue that even with the competition, Nebius is doing quite well in terms of its growth, right? Uh, but when you think of, like, is Nvidia going to like change its mind and be like, "Hey, I'm going to go get into this business?" Which they're not going to, because Nvidia doesn't have some sort, they have a business where they rent out compute, some of the core compute does, but they don't necessarily build the entire stack like Nebius does. Um, if you're like waiting for that to happen, and you're assuming it's going to happen, I assume it's going to happen. It's probably going to happen. Not with Nvidia, but like, you know, if, like with Amazon HIMS, right? Um, Amazon is clearly seeing some benefit into getting into that sector and getting into the TAM that HIMS is getting into, and the competition is coming. Amazon is trying to flex on there. But you and I both know, as investors, and so does Roy, that Amazon probably will not catch up to HIMS in terms of innovation. And that's kind of where Nebius sits, being in charge of the entire stack, right? The hyperscalers can certainly, I would argue that the hyperscalers are ahead, but the thing is that they don't offer the certain competitive pricing, in addition to offering certain workloads and paying attention to the small business customer versus what AWS can provide. Like, you go on AWS, you see what you get, buddy. That's what you're going to get. You want to sign a hundred billion dollar contract? Okay, let's talk. Right? With Nebius, it's like, "Oh, okay. Uh, you want to sign like a $50 million contract? Okay, come on. Let's go. Hit me up next week. Let's go to bar and let's talk this out." Right? AWS won't do that. You're going to be like, "You get what you pay for." The competition is already here, right? And it's really the job and the trust and management that they're going to figure this out, which I would actually argue they are figuring it out. I mean, when they were running Yandex, they were basically running their entire data centers over there. So, they're bringing that expertise over here. When I say over here, I mean like other parts of the world besides Russia, and they're able to build the entire stack in an efficient way because they have that experience. That's something that other players besides the hyperscalers don't necessarily have. They don't have that experience. So, if you think about the expertise that hyperscalers have to actually build data centers, the relationships, and leveraging, let's say, sovereign funds to build their data centers and so on. I would argue that Nebius does have that capability, and that's something that, well, CoreWeave isn't in that business, but that's something these other businesses can't necessarily do. So, if you even look at the, sorry, I'm still talking, so I can't bring it up or anything. Nebius is very efficient. They're like the third, according to their investor day presentation from last October. They're the third, they're the third most efficient data center builder out there from the ones that they didn't also include.

Well, I mean, I think the most recent presentation has also AWS and other providers. Not CoreWeave. No, because that's a power of. Yeah, it's, that's the efficiency of the data center itself. It's not the, I think you're talking about the, the efficiency of the AI workload. That graph itself was about the energy efficiency of the data center itself. I don't think that matters at all. I think that it matters about the margin that you can get out on. Like, I think you're talking about that great system that put like CoreWeave at platinum and then Nebius at gold. No, right? Well, yeah, sure. You could. Yeah, sure. Because that's essentially what made CoreWeave be above Nebius was just the scale. Because right now, CoreWeave is at like a $5-6 billion scale. So, they've proved, they've proved that they can do like very efficient AI workloads at scale. And this is something that Nebius still needs to prove. And I mean, I, I sure hope they will prove it next year. Uh, so I think the difference is not that CoreWeave is more efficient than Nebius. I think the grading score of that, I'm not sure if that's what you're referring to, because but even in that, Microsoft was also below Google, I think was at the lowest tier. Um, you know, okay, so maybe it's not the grading I was referring to, because that was specific to nail clouds, like didn't include hyperscalers. But I wanted to touch like on your previous questions, because I think it, like, they weren't answered at all. Uh, you said about like hyperscalers, like what's preventing them from scaling even more than Nebius, for instance, and capturing even more market share? Uh, you see, like AWS is growing right now at like 100% year-over-year or something like that, and the same to the AI parts of the cloud of the other hyperscalers, 100-something percent. And they said that they are completely sold out, and that each capacity they had, they sell. You're saying AWS revenue is growing 100%? No, not the entire AWS, but the AI part of AWS is growing at 100%. And that was just commentary. Yeah, that was just commentary on the scale. It was not the presentation. And why aren't they growing even more? Because they use more GPUs for their own needs than they want to like provide to other clients. Essentially, they just want to have larger clients. For instance, most of the customers that Nebius has right now wouldn't be allowed to use AWS and wouldn't probably want to, because, as I was saying, like, a few minutes ago, like despite being very efficient, the customer support is not like anywhere close to Nebius, and that will end up affecting reliability. But even if you assume, like, because I know that's a big assumption for everyone, that hyperscalers can match everything that other tech companies do. Essentially, they're just not interested in like 40% of the demand. That's the main thing that you just need to understand to just get rid of the hyperscalers, like bare case. The thing is, is there's like 40% or even 50% of the market that the hyperscalers don't, I mean, can't really fulfill and don't even want to. That's why CoreWeave and Nebius and other nail clouds exist. The fight here is more with other nail clouds rather than hyperscalers itself. I would say it's like a niche within the AI GPUs cloud, because Nebius isn't really going after the demand that, I mean, maybe the demand that hyperscalers face and like fulfill, they could use Nebius, but they're using hyperscalers. Yeah, that could be the case. But most of the demand that CoreWeave and Nebius has is from customers that don't want to work with hyperscalers. And then to your point of like, why doesn't Nvidia just stop selling them GPUs or disrupt their business? It's not really because they invested in Nvidia. I mean, that's really for their value, like the Nvidia's value. It's just really a piece, a very, very tiny piece. So, I don't think that's because of that. But essentially, Yandex was like the largest non-American and non-Chinese customer of Nvidia for decades, and their teams have worked together for many years. They have like a long-standing partnership, like a very deep one. They work closely in product development, in software development. Their teams really work together every week. Like, I've seen interviews with members of both teams, like from Nvidia and Nebius. And for instance, Nebius, as I said, designed their own racks and servers for the data centers, like the fully vertical integration part. And they designed the racks and servers like in collaboration with Nvidia to know what, like, how will be the next generation of chips that, like, after, like, right now we had the Hopper generation, now we have the Blackwell, and we will have more, of course, in the future. And they do it in a way that allows them to, let's say, future-proof the servers and the racks and the equipment to be able to adapt the new generation of chips without incurring in the huge capex that other companies will need. Because, for instance, CoreWeave right now, imagine they have like a huge data center with Blackwells. For the next, like, as they are not vertically integrated, they have to buy the servers, racks from Dell, Supermicro, etc. Next generation of chips, they will have to buy from Supermicro the new version of racks, the new version of hardware, everything. And it is surprising though that Nebius is able to do that though. Exactly. And that's because of the huge team that they have from Yandex. Because I think people really underappreciate this advantage, because they are just a bunch of 800 geeks that really know everything about hardware. Like, it's to me, that's really, as I said, like three or four times already, like the big advantage of the company, and something that really most people underappreciate, is the power of engineering resources. Because this not only gives them the cost advantage, as I that I was saying, it gives them the efficiency, at the same time, better margins, it gives them the ability to know everything the customers want, and to have a customer-centric approach that really improves everything for the customer. They even built out their own LLM just to analyze the platform like on a weekly basis and say, "Oh, maybe you could improve this." Like, in the first quarter alone, they added 50 new, 550 new like features to the software stack and the software layer of solutions. They're like obsessed with improving the UI and everything in the applications, tools that they have, and like everything. They're really an obsessed team, and I think that's really a huge advantage here. And that speaks for itself that they ran away from their countries and are here. I've heard from like the vice president of sales that some of them are like, don't even see their families for like two years because they wanted to come here and like build a really differentiated technological company like they had in Yandex. Because most of them are really passionate, but they were like heartbroken from like the what happened in Ukraine, and they just didn't want to be part of that. Even the founding team, they were the founders of Yandex. The team that founded Nebius were the founders of Yandex like 30 years ago. Like Arkady and the other, they were the same. Yeah, they were the same. The same people that founded Yandex are the ones that like left the country. They changed nationalities and founded Nebius. So, really something underappreciated in the company, and what gives birth to like the other advantages. And yes, you can talk about the bare case of hyperscalers, but I, as I said, I think it's something that people focus too much, despite the fact that it's not really the same thing and not maybe direct competitors, but really not mutually exclusive, let's say.

Is there any, so let me just pick up on that, um, because a lot of mentions were made to the fact that they're an international company. Is there any macroeconomic pressures with everything we've been seeing around the US's stance, you know, um, with them not being a US company, interest rates, all this?

Um, I would say that the debt that they're financing, well, they got a lot of the proceeds from selling the Russian counterparts for Yandex, and also the capital raising through investors. Um, they haven't taken on any new debt lately. Um, but I'd imagine if they do take out debt, it would be in the form of 0% senior notes for 2030, much like a lot of other companies are doing as well, because they don't want to pay interest on it, of course, but also because they believe that the upside is going to be much higher than I would say liquidating shareholders in five years. Um, as far as the interest rates in the US goes, that would be negligible compared to raising 0% debt. Um, but also, let's remember that we don't need to worry about any type of, um, chip acts or anything, because they are headquartered in the Netherlands, and they're building data centers in, um, I forget if they're building a data center in Israel, but in Finland, in Paris, the US, of course. Um, one of the, there's, they have like a lot of collocation data centers, which means that they're basically renting out space in the data center and throwing in servers and racks in there. But the actual data center from top to bottom, it would be New Jersey, I believe. Paris, as well. And Finland is an existing one. So, Paris is the kitchen as well? Oh, Paris Colola. Okay. So, Finland is the one that is the one that they actually own. So, they have two actual data centers that they're building from scratch. Um, so they have a lot of optionality, especially working with, I think the company's called Equinix or Equinix, I forget the name. Not the gym. Not the gym guys. Um, but asking him to show the map. Yeah, I would show a map, but I don't think we really have time to do that. But, um, I certainly we do have, or Manuel has an article that he recently wrote. Is that the one that you shared me? Is that the most recent one that you want to share, Manuel? With the map, I think it has the map.

Yeah. Before you do, guys, I wanted to jump in a couple of times. First, just unrelated, but I find it hilarious that there's only one non-HIMS investor here. I know we're not talking about him, and it's Manuel, who was probably in at first because he appreciated so much in value, he was in so early. And that really pivots into the next question. So, Nebius, you're in this, you're very concentrated, and I respect that, Manuel. Sam, I think you're a little bit less concentrated. Me, it's a very small position, and frankly, from what we discussed today, I lean more towards it being a swing trade than a long-term position. I'm not convinced on the moat. You know, I see that there's a market there, but short-term, what I saw is that there is just insatiable demand that competition does not matter at all right now. Now, for the long-term thesis, I need to be convinced that they'll continue to iterate faster than competition, which there will be a lot of new competitors in this space, partly because, I mean, this kind of environment breeds competition. It does. And so, I think that what Tanner was saying, as far as his questions on hyperscalers and current competitors, not current competitors, future competitors pivoting this direction, I think it's really valued. But over the short term, like the demand is ridiculous, and it was undervalued. It's underpriced, mispriced. I'm up 60-something percent. So, my question, Manuel, like you entered, and for Sam too, like you guys entered this, you know, you're in it to make money. At what point are you like, "Okay, great long-term thesis. I love it, but it's gone up too much. It's time for me to take some profits or get out." Is there a price there?

So, for me, my average is $24, and it was a 1.5% position. So, probably close to a 3% position. And I mean, I'm a much different profile than Emanuel. I'm 37 years old. Sorry, I'm 21 years old. I'm 37 years old. I got a house. I got kids. You know, basically the same thing as you, Roy. And, you know, I'm not risk-averse where I want to put a large sum of my net worth in Nebius because I'm aware of the risks and I'm very vocal about the risk when it comes to this company. Would I buy here? Probably not. But seeing at the swings and the volatility in the market, this thing could be under $30 bucks in a couple of weeks. I'm not saying it will, I'm just would not be surprised. Is there a dollar amount where I'm targeting? For me, my fair value target is somewhere around $57. And it is a bit aggressive, but I would say in a year and a half. Which I think it seems a lot closer today than it does seem just a week ago. But at the same time, I would probably start selling some covered calls, maybe ran up a little more in that range, which would be like a month or two out. So, I'll be looking to capitalize some income off of the position as well. If the thing runs to $100, you know, I have for all my positions, including HIMS, I always have a core position that I never sell covered calls on in case the thing decides to run. With that being said, with Nebius, I would only sell half of it. The other half, I'd probably just let it run until whenever. And if it went to like $120, okay, I'll probably sell. Like that's like saying that when HIMS was $23, like where would I sell? I don't know. Maybe if it was like $80 or something. And then here we are today, right? So that's when I would start trimming HIMS, right? So, I think every single position portfolio is considered a swing trade per se, because like no one holds a position indefinitely unless you're holding the S&P 500.

Travis is 47 years old mentally or physically. I think you mean visually, and we'll leave it at that. He's physically 57. Way older. For you, what's your fair value? What's your point where you're like, "Okay, good enough. I'm out."

Okay. I don't think that point will be reached very soon. To be honest, I think it's still very mispriced, and I think there are too many catalysts for me to get out soon. I mean, unless it goes to $100 next week, of course. I mean, we all have limits, but I think it's still very cheap in my opinion. First, as I said, like, we still have, and the reason why I went so heavily and my average cost is like $25 per share, $25.67, I believe, is because I thought, and I still think, the downside, the downside, sorry, is very limited. We have the other subsidiaries supporting the valuation and are not enough right now. It's, I think almost at $10 billion, like 9-something. So, now we can't really say that you're buying the core business for free. But when I was like heavily buying it, it was really like the core business was being offered like on top of every other asset. And I think that if Nebius gets to even the lower end of the EBIT margins they guided for the midterm, and midterm is like a few years. So, I would suspect like maybe like 2027, 2028, and at that time, they will probably have like at least $5 billion in revenue, just based on the capacity they have. They are like right now expanding. This could be even 10. Just look at, we've went like from like $500 million to $10 billion in like two years or something like that. Those were terrible earnings, like the earnings were so bad and burning cash.

Percent higher, right? Like it could run in sentiment and we've seen it with HIMS, and I would even say HIMS isn't even overvalued today, you know, but in terms of market swinging in one direction further than it is the other. Sorry, I hit my fake plant, my real plant. But, um, I would say like this thing could run, and I'm not betting on it now, hoping that it'll run, but if it did, I'm gonna manage my risk, right? Like that's just what we do with almost any position that we should have. We should be risk managers looking at it. Investors, you know, different story. If I was in my mid-20s, like I would have a much bigger position than 2 point, than 1.5. I'm not even in my mid-20s. So.

Well, there you go. You have time. You have very little expenses and so on. But, oh, go ahead. Yeah. I gotta run soon. But yeah, just to finish what I was trying to say is like, I believe this could be like a $100 stock like next year. Like this is not my base case, but I believe it's pretty possible, and there's like the numbers support that upside. And just you can do like a quick sum of parts valuation. I did like a complete model for free, and it's like, you just, you put like clicks and clicks. I would say will probably IPO soon, but let's just stick with like the current valuation. But like, let's dilute a bit the 28% stake, because after this valuation round, it will get inevitably diluted. Let's say it's like $1.5 billion. Then we have Toloka, which I mean, I don't even assign that much of a value, but I think it could be worth significantly like in a few years with the CTO of Shopify and Bess was like investing in it and all that. I think it could really grow to a big company because from what I've heard, like it's really grabbing market share in the data labeling space, which is growing like crazy. Like they are more than doubling revenue for like three years, even though it's still a small base, but let's say like $300 to $500 million, which is quite low for a company like that. If it was trading in the private markets, it would be like in maybe the two, three billion. But I don't want to be like too optimistic. Then you have the Triple 10, which is like a net platform, which is just not really that big of a deal. I would assign like $100 million, really, really not that much. But then you have the Ride business, and I think this is one of the like largest short-term catalysts for the company. I think we will get this year a valuation round for Ride, and right now the markets doesn't know how to value the company. It's really, there's no, there's no benchmark. I mean, you have the benchmark, as I was saying, of the peers of the selected peers, and just by looking at them, you would say that it could be worth like four or five, you could even say $10 billion, because it's like a really, like a few years behind Wimo. But I, as always, I never took it like at $10 billion dollars. Even my initial valuation model was like $2 billion. And then I, when I saw that they were really in advanced phases of negotiations, I assumed that they could reach like $4 billion because I used like a selected period that in 2021 or 2022 got a $4.1 billion valuation from Hyundai. And just that, just those assets are like on the low end worth like $6 billion, let's say, or five even. So, you're buying the core business for like $3 to $4 billion right now, and they could be generating like next year $4 to $5 billion in revenue, and this is obviously a company like growing really fast. I mean, you know, the unit economics are really not certain in the future, especially with the next generation of GPUs. I know. And I'm really aware of the risks, and that's why I only went so heavily because I felt that the other subsidiaries were really limiting the downside. But essentially, I think that if they meet expectations in the midterm, this is still really a buy. Like this could be, just look at CoreWeave right now is worth like $43-44 billion, just based on that. Nebius should be double what it is right now. But I mean, CoreWeave is expensive. So, I don't, it's like those guys that say that HIMS should be at $1,000 because Palantir is at 200 times sales. I don't really base myself on CoreWeave, but I feel that even if CoreWeave was at the IPO price, like $40, right now he's at $90. Nebius' core business itself should be worth at least like $10 billion. You put in all the other subsidiaries, and you still get a 50, 60, 70% upside on the lower end case. And then you have the evolution of the business itself. So, essentially, I saw this as an asymmetric opportunity with really low downside and huge upside. I mean, now at $40, it doesn't have the same characteristics as an investment as it had like at $20, which went like my average cost initially was $31, $32, but I really went heavily at like $20. So, essentially that was it. I saw like a low downside, huge upside opportunity, and I still feel that the upside is not as near as it can be in like one to two years. And this is not even a five-year story. I think that as the market starts to see that at scale, this will get like good margins. I mean, hopefully, this is not certain, of course, it will always depend on the execution. I still feel that it will have major upside, and I don't plan to sell anywhere like close to these levels. So, yep.

Um, sorry, you guys can continue. I got to bounce. Um, but I did put comments in here for Manuel's Substack article, because he spends a lot of time focusing on these smaller companies, which we do not find enough content on, in my opinion. And we also have a Nebius X community that I started on Friday, which has like 500 members in it, which is awesome. Both are free, so, you know, but Manuel, of course, spends a lot of time on it, so, you know, make sure to take a look at that. He has a lot of beautiful content on there as well. I'm not getting a cut from it. I just, I'm trying to help a brother out, right? So, yeah, man, I think, I think, I think we can all wrap there, Sam. And then if there's interest, we can even do a follow-up episode, if there's like more detailed questions that we didn't get around to. But everybody, we were a bit like confusing with stuff coming from one side and the other side. I mean, look, man, I think all of these companies in the AI space are kind of, it takes a little bit of time, and definitely at least for me, even me who works in the tech space, like more than one read to fully wrap my head around them. There's a lot of concepts you have to understand before you can start to understand, okay, now what do these guys do again? Um, so I can understand if a lot of people in the audience are feeling that same way. But, um, yeah, really interesting. I see the value prop. I see the growth rate. It's small right now, but it has a potential to be much, much larger. The space is growing. Um, so, yeah, I think, I think more research is needed on my end. Um, Tanner, anything on your end, just before we wrap?

No, I thank you guys for coming on. Really appreciate it. Um, Sam, solid. I liked your comments less about Nvidia, but that's okay. I'll hold my tongue. Um, I mean, don't get me wrong, like, it's going to be an all-time high, and it's going to, it's a leader of AI. Like, it will always be relevant. Always relevant. Yeah, I'm not, I'm just more joking. Um, but yeah, thank you guys so much. I might take a position. I just wish it didn't run up so much on CoreWeave and Nebius, because I know that that is definitely my next play after playing the hardware side for artificial intelligence, that the software names will take over.

Well, those are still hardware plays. Sure enough. I mean, more annual recurring revenue. Oh, yeah. Yeah. Maling for tomorrow. So, yeah, and, um, yeah, thanks again, everybody, and thanks for being here. Reach out to these guys and join the Nebius community on X, and we will see you, and let us know actually what stock you want us to talk about next in the stock tank, since we're bringing it back. And we'll see you there. See you. Bye, guys.