These AI Companies Could Become the Next Cloud Giants

Did you miss the trade on Neoclouds? Many people think so when they see that these stocks are up like 500% in the last year. But what if I told you that they may still have room to run and that there's actually $1.7 trillion of capex pressure coming down the pipe with almost nowhere else to go. What's up everybody? It's LG here and welcome to Milk Road Stocks, the daily market show that isn't afraid to top last all-time highs because they might not be all-time highs for very long. Today is August 20th, 2026, recording on the 19th. Our leading AI analyst, Melvin, has been banging the Neocloud's drum since early February, buying Nebus way back at $95. He also owns Cororeweave, and just this week, he added one more Neocloud company that the market has not yet repriced like the others. And today, he's going to dive into the entire sector and explain how so much of that incoming capex is going to directly benefit these companies and whether SpaceX actually poses a real threat there or not. The numbers are absolutely staggering when you zoom out. And we're going to go through all of it today. And a reminder that all of our Milk Road Pro members get these trade calls in real time from Melvin and our four other analysts. These have been so good in the last few months that we are actually going to raise the membership price next week from $25 a month to 39. That's a 60% increase. So, make sure you lock in that original price before it's too late at the link below. And a reminder that our podcast today is free and that it wouldn't be possible without our partners at Saver. money, the stable coin payments platform built for Asia. Keep an air out for more information about them later in the show. >> Melvin, I feel like I'm in a dream or more so that I'm on a Neocloud every time we have a podcast because you're a Neocloud guy. Uh, and I always want to hear more bullish takes on Neoclouds, which is what we're going to do today. Yeah, I'm excited to be back here and I feel like we did an episode on Neoclouds like uh maybe two months ago and they have ran massively um up to even 270s 280s. If you look at um companies like Nebius, we had a bit of a pullback right now. I think Nebius is sitting around 230. But I think this um market is entirely long wrong because just if you just look at the capex alone in 2025, you can just put the first chart up LG um top seven hyperscalers Amazon Microsoft Meta Google Core spend 443 billion on capex um and then in 2026 that number goes to 911 billion and in 2027 it's going to $1.48 trillion. And where does that goes to? Where does that go to? That goes to training, inference, compute, and the companies building and operating the infrastructure, which who are these? Those are your Neoclouds of today. So that's what I wanted to talk to you guys about today. Numbers no one has ever even heard of until the last couple years. They're going to spend $2 trillion. This is something we've never heard of. And just for people listening, um, we've mentioned our campaign that's going on this week already by this point, but also Melvin has been our leader for Neoclouds. He bought a lot of them, Nebius and Corweave, months and months and months ago. We've seen an amazing runup for people who invested alongside him in Pro. Uh, and today, this morning, he bought another one, uh, which we'll only tease today, but you'll have to go milk pro to check it out. Uh, that's enough of me interrupting. Tell me, Melvin, like how of that massive capex that's coming every time we talk about it's another 1.5 1.7 trillion. How much of that is actually Neoclouds though? I think that that's that's a big question for us. >> Yeah, a lot of it actually. Um in fact um before I even if you're new here, let me just explain what Neocloud actually is because I think um a lot of people know what it is, a lot of people don't. And let me just give you some simple definitions, right? So Neocloud is pretty simple. you buy a bunch of Nvidia GPUs, put them inside, you know, data centers, and then rent that compute out to AI companies that need it. Um, that's the core business model. And where it gets more complicated is that not every Neocloud is renting raw GPUs out. Some companies like Nebius build their own infrastructure on top um top of the um GPUs making it like easier for customers to train models um run inference, manage workloads and like at a massive massive scale uh scale. And so the basic idea is basically a GPU rental business but better NeoClouds are trying to become the full AI infrastructure stack um building around those GPUs. So that's what that's what Neil clouds are essentially. Wow >> man. Yeah. I always get I always love hearing exactly kind of like that that good breakdown. I guess I guess I guess the question people really want to know Melvin too is is even if they're getting so much of that spend. How big can these companies actually become? Right. I think that that's like that's a real question right because even you know on the memory side you called Micron again months ago back in February when it was like you know two 300 bucks whatever it was. Um, and even we've heard that that's one of Jordi Viser's legendary calls. He called it years and years and years ago. But, but something like Neocloud is a bit more of a a new concept. And even though Micron went to a trillion dollars and there weren't trillion dollar companies until a few years ago, I think people wonder that also for Nebus and the core and the other companies is like, is that the trajectory? Is that the size that they're supposed to grow to? And and I guess I want to know from you is like how do you how do you measure that? How do you estimate that? Yeah, that is exactly the trajectory they are in right now. Because if you look at the chart that LG just put up, AWS right now currently, which is their cloud business, has a 42.2 billion in quarterly revenue. Um, a Azure, which is Microsoft has roughly 30.2 billion. Um, Google Cloud has 24.8 billion and then Coreweave um2.5 billion. and Nebius has $5582 million. So, Neo clouds are still a tiny fraction of the, you know, total cloud market. And the first instinct most investors have is they see a chart um they see this chart and they start to think that's a problem. It's not because that's where the opportunity is because look at where those hyperscalers have started from. If you actually look at the chart um next charts, this is the same NeoCloud revenue growth curve extended out to AWS and Azure um today. So AWS at quarter 81 was doing is doing $42 billion per quarter. Azure is around 30 like I said. The key thing is this chart shows what their trajectory trajectory looked like at quarter 20 to 25 which is where Nebius and Cororu are right now. So they looked exactly like this. They small but vertical. The curve looked like it's like it's hard to even tell apart because they're on the same trajectory uh for these um AWS and clouds of the world. >> It's it's amazing to visualize it this way. And I think also even just back on on the last slide, it's like I didn't even realize when I first saw this slide what I was looking at, which is just basically this kind of um like square diagram of a bunch of squares that shows how much of the cloud market is is or their how much revenue each of these major cloud companies make. Like you're saying, Azer and AWS and Google and everybody. Um and then how tiny Corwe and Nebius and Iron are. You know, it's it's like a blip but expected to to grow massively. and and you're showing them on this next chart where it's like listen that's what has happened to these other businesses like AWS and that's likely what what that trajectory could be for these companies. >> Correct. And the you know you might be wondering why is there why is why does Neo clouds even exist right like we have so much like um you know AWS we have Microsoft Asia and all that but the reason is um exists is because the hyperscalers cloud business are general purpose platform built to serve like millions and millions of different customers with different needs. That's that's the core idea behind their platforms and they want to serve it and use it in their own businesses. Um but their infrastructure is basically designed for flexibility, right? Whereas Neocloud is basically designed for maximum GPU performance uh for AI training and inference, which is what we're going through right now. This this entire AI boom. >> If you're not a pro member, what the hell are you doing? On August 26th, the price of Milk Road Pro is going up. Monthly goes from 25 bucks to 39 and annual goes from 250 to$ 299. And here's why. Pro used to be a paid newsletter with one portfolio attached. Now it's a full platform. Five analysts running real-time portfolios with every position, every trade, and piece of reasoning out in the open with live notifications to keep you in the know. And the results have been insane. Melvin's portfolio launched in February and it's up 41% since. If you put 10 grand behind his moves, then you'd be sitting on more than four grand in gains. 5 months of pro costs just 125 bucks over that same stretch. Is there a better ROI on any investment right now? If you join before midnight Eastern on August 25th, you keep today's price for as long as you're a member. After that, it's gone for good. The link is in the show notes. So that's I think I think you nailed my next question too because that was I was like wait listen if there's if there's already AWS doing 45 billion of revenue clearly people like using them why how could these income how could these new companies ever hope to compete with that but it's because they have advanced GPUs like they are specifically serving the hyperscalers and companies building out AI whereas AWS is like we use it for milk road for hosting our whatever we do you know >> exactly that's exactly the reason And let me show you like the demand side of this because if you go to the next chart um because I think most people are still underestimating how much demand there is. If you look at the chart, this is Anthropic's annualized revenue run rate for this year. So just to give you some context, Anthropic went roughly 1 billion in ARR in 2024 to 9 billion in 2025 and now is running at 65 billion as of July of this last month. That is a 65 65x increase in two years. And Anthropic is not alone. If you look at the chart, um, Anthrop OpenAI is generating about $45 billion in ARR, uh, by this year. Um, and they're, so if you combine both of them, they're roughly making $105 billion combined in the last two two years. Now, think about what that means for companies supplying the infrastructure underneath them. So every dollar of AI revenue ultimately requires more inference, more tokens, more uh more GPU time somewhere and that is the demand engine behind companies Neocloud's companies like Coreweave and Neocloud Coree and Nebus. They buy the GPUs, build the data center capacity and then rent that compute to companies like Anthropic that cannot simply get enough of it. And this is the same across the board um for all companies. There is there's no like there's it's so hard to find supply at the moment because if you go to the next chart, you know, next chart is very interesting because this chart shows the combined cloud revenue from Microsoft, Google and Amazon. Um and how much they have in backlog. um they just crossed 1.7 trillion in uh in Q2 2026 growing at a roughly 140 150ish percentage um year-over-year and 1.7 is already contracted cloud revenue that has to be delivered. So there's a massive massive backlog of how much compute and how much cloud you know we actually need. Uh because if you go to the next chart this is going to blow your mind. This is the total committed spend across um current leases, future leases and financing for Amazon, Google, Meta and Oracle. So, Amazon has 453 billion in total commitments, up 13% year-over-year. Google has 946. Meta at roughly 736 731. Microsoft 672. You get the point. Add all those up. Five companies have over 3.1 trillion in commitments. And guess what? Every dollar of those 3.1 trillion commitments eventually requires GPU, compute, data center. We had an episode on power yesterday. If you haven't seen that, you should go check that out. Cloud services deliver. And that is the total addressable market behind the NeoCloud trade. >> Oh man, that's insane. And those commitments, just to reiterate, those commitments are are through how many years? Like three, four years basically. Three, five, three to five yearsish. Yeah, three to five could be longer. It all depends on the contract that they're signing. It varies from place uh you know from contract to contract, >> right? Yeah, absolutely. Uh I had a question back on on on this slide as well, Melvin. Um this we've seen this we've actually we actually looked at this this these numbers that came out this week basically from Anthropic in terms of their revenue projections or their their quarterly revenue, right? And and we've heard the rumors of them uh going to IPO. Is this in your opinion and we've talked about we've talked about this with the other analysts as well this week and and I think it's worth reiterating with you who is our our leading portfolio analyst Melvin's portfolio is the leader of all our portfolios just so people know. So we want Melvin's opinion. Um does this confirm that AI makes money? You know what I mean? Like I think that that's been the big push back a bubble whatever. It's like what's it going to take for these companies valued at trillions of dollars that aren't public yet and we haven't seen their books. Does this confirm that it's like they are making money? They're going to continue to make money and it's going to continue to grow. Yes, absolutely. Um I think they're going to the you know that we've seen projections of them uh roughly anthropic hitting about 100 to 120 uh billion dollars in revenue by the end of this year. And there's projections out that that by next year they're supposed to hit 250 or even upwards of 400 billion in revenue. Um and I think that that is likely what's going to happen over time. Um and even what's interesting is Anthropic actually last quarter I believe they made a little bit of profit as well. Um you know they have they have huge commitments they they're spending massive on you know money on demand and you know uh signing massive deals with other company but they they their token prices they have approximately 80% margins as of now. So that that will be a huge huge um you know catalyst for them because the thing everyone's looking at is ROI is is their money are you making your money back and the question is the answer to that question is yes they are starting to make that money back and their models and everything is only going to get better over time. So I I completely agree that they're they will likely reach hundred billion dollars in revenue by the end of this year. >> No it's amazing. So you were talking about like every dollar spent here on uh from the hyperscalers. How do you how do you actually uh calculate that? Like what are the economics of that look like um when it comes to trickling down to neoclouds, right? Or rather I guess what what are the economics of Neoclouds? I think it's always worth reiterating that. We did an episode with you on Neoclouds a month or two ago that people found super informative. But I think now on this kind of bullish renewal, it's like it's good revisiting that, right? that it's like how how these companies stay profitable and continue to make money as as it comes down. >> Yeah. So the problem right now is they're not profitable these companies NeoClouds. The reason is simple because we've seen all the demand and prices for specifically memory. If you look at optical modules and if you look at pricing across the board for everything prices have basically doubled tripled um in the last few months. Um before it used to make uh it used to take 20 to $30 billion in um to build one gawatt out. Now that price is somewhere between I I've seen projections upwards of hundred billion to make one gawatt of power. But the problem is you have to spend so much money in capital and capital to build out these data centers. So they're not profitable right now, but the goal is to become profitable as you build out over time because once you build out the AI infrastructure, then it's all making money. It's all it's reoccurring revenue that you can rent and collect money from all these clouds. Because what's interesting is if you want to see a perfect example of this is just listen to coreweave what they said during their earnings. What they said was their A100 A100 chip which was a chip made in back in 2020 um so it's been about six years that chip contract they've signed till 2029 so that's 9 years worth of um appreciating you know asset that you can use to make revenue and I'm pretty sure they're going to go beyond that it's going to be longer than 9 to 10 years so that is the model that the NeoClouds are working under >> okay So we've seen basically like there you're showing us the revenue per active megawatt. I think I think the other question for the Neoclouds too and and especially when we're looking at um non the less AI focused ones like like AWS obviously um the big question for AI is is the megawatts right is like and we even did a roll up about this. We talked about energy yesterday of like what's the cap on the US making megawatts and it's it might not be that much right now but for these companies how much do they actually make in the grand scheme of things? I think that's the big question especially compared to to like a SpaceX which is also Neoclad but you know what I mean I don't think they fall in that category. They're they're also a much bigger company. >> Yeah. So like if you actually look at the previous chart um that LG put up you can like see how much these companies are making per megawatt of revenue. If you look at Coree they make roughly 9.8 million per megawatt annually. Um um if you look at Nebius they're about 9.4 Iron 10.4. You get the gist. as you go down, you make less down this chain. Um, and currently SpaceX roughly makes $50 million per megawatt. But what's interesting is Neoclouds are generating now more twice the revenue per megawatt of traditional data center operators. And what's even more interesting is that Nebias on their earnings call revealed that they're now signing deals 40 to50 billion dollar per megawatt which is 4x more than their baseline from start of the year because of the demand. they only have the demand or they they are the ones who have the supply and now they're starting to charge even more because the compute prices and everything has just come gone crazy and there so therefore they're capitalizing on these shortterm deals 3 to 6 months deals that they're signing and if you want to actually see this play out in the dental uh daily rental price of every um GPU right now um if you go to the next slide this right here is the daily rental price for every major GPU across non-hyperscaler cloud providers. So H100 is 2.74 per dollar per hour. H200 is roughly 3.28, B200 is 5.61 and even the older models A100 covering now is at 1.65. What this chart shows is that price for these GPU rentals GPUs bottomed in late 2025 across nearly every chip and h but has been climbing ever since. The H100 a chip that marketing has been on the market is is still commanding 2.74 per hour and it's up from $2. you know, it was I think it was priced around $2 last year and it's up to 2.74 now. >> Melvin, my question for you is is as these prices go up, there's also new compute coming online, right? And I know that there's a bottleneck and that there you need more compute, but how does that how does that justify it? You know what I mean? Like is there is there just so little new compute coming online? >> There is compute coming out, but supply is still tight. All these companies need all these companies need these infrastructure and because if you look at open AI or anthropic their their compute and how much money that they make uh is directly try tied to how much compute there is. So the more compute you have the more revenue you generate. So therefore there's always a demand for these GPU um rentals and GPUs because there's simply not enough to go around even even now. So that is the reason why >> yeah okay that makes sense. So are and another question on on that point are the customers actually like are they getting enough value from these chips because I know that they're being charged at a premium but is it is it worth it right? Like is it whoever bought that I forgot who signed that core weave A100 deal that they're like six-y old chips that they're going to sign through 2029. Is that actually worth the premium that they're going to pay or is it just like a desperation? it is worth it because not specifically A100 but the reason why these companies are signing short-term deals is because maybe if you look at companies like OpenAI or Anthropic maybe they have a new model that's coming out and they need more training inference or training just to get through that release phase so they're willing to sign like 3 to six month deals just to get through that. That's why compute becomes way more valuable. Let me show you um the chip trajectory u because I want to address something that comes up almost every bare case in this on this AI trade. Um if you look at the chart that LG just put up um chip performance per dollar is improving at roughly 49% um year-over-year. The GB 200 and GB300 are delivering multiple times the performance of H100 at a similar or lower cost per compute. So the bare argument goes like this. If the chip keeps getting better or cheaper and more powerful, won't AI companies just buy fewer of them? Won't the demand for cloud compute just flatten out? My answer has consistently been no. And the reason is the same dynamic which I talked about in that memory episode because cheaper compute lowers the cost of inference as you have seen which lower the cost of building AI products which means more companies build more AI products which means more inference volume which means more total de compute demand. This is the classic Jevans paradox because every time AI gets cheaper the more people use more of it and that's how our society has functioned for hundreds of hundreds of years and anthropic going from 9 billion to 64 billion in 8 months is that you know while the underlying like Nvidia hardware was available eight months ago is the proof of that because cheaper and faster chip um do not reduce like demand for cloud compute they expand the you know of what AI can do which expands the demand and this is the the case this is the bullish case for neocloud I would say you know one thing we've talked about a lot on this show is that crypto is quickly becoming a huge part of the global payments infrastructure and nowhere is that more obvious than in Asia but if you're actually running a remittance company or a payment business you know that the hard part isn't moving the stable coins it's dealing with local banking partners compliance liquidity and all of the operational headaches that come with sending money into places like India and Southeast Asia. That's why today's partner is Saber. They give payment companies stable coinpowered infrastructure to collect and make payouts across Asia without having to build all of that complexity themselves. They've already processed more than $3 billion in transactions across 40 different countries. So, this isn't just a concept. These guys are actually doing this for real. If you're building payment infrastructure or expanding into Asia, make sure you check out saber.money. >> Uh Melvin, can we actually just talk can we just get right to the companies, too? cuz I think, you know, this is all nice to hear and and I think if people dig into the the businesses, they'll eventually find a lot of these numbers and it's great to see a lot of the big firms like A6Z kind of projecting this, right? But I think inevitably uh people listening want to pick winners, right? And maybe they'll basket Cor and Nebius together, which isn't the worst play either, I'm assuming. But I think it's worth noting that they are they are still different companies with different P&Ls and um I think we want to learn a little bit more about them today. >> Of course, let's start with Cororeweave. So Corey uh posted the revenue like about a week and a half ago. Um results were exceptional. Revenue of $2.5 billion up 112% year-over-year. And if you actually go to the next slide, if you look at um Core Weeb's uh backlog, their backlog has exploded over the last year. It went from roughly 25.9 billion in Q1 2025 to um 104 uh billion in Q2 2026 as of now meaning they just 4xed their you know um cloud capac or not contracted RPO in five quarters and the biggest jump came in Q3 of 2025 which grew 84.7% um quarter over quarter Um and then in 2026 it jumped another 48%. And the one 1.4 billion that I mentioned does not even include the 25 billion in net new customer commitments that they added in the first few weeks of Q3. So the backlog is actually higher. And now to put that $1.4 billion in context, Corvy roughly has a market cap of $50 billion right now. meaning the backlog is roughly 2x the value of the company depending on you know um how the price moves up but obviously the backlog is not the same as like the profit right and that four the 104 billion will be recognized over multiple years but it just shows you how much AI demand core has already locked in >> how much how much confidence should we have in that 104 billion like I know it's I know and I guess it's a bigger question about all these these capex commitments, right? But it's like how much confidence are we supposed to have in that? Like is that guaranteed? >> Quite a bit because we have >> money's held in escrow somewhere. You know what I mean? Like it's definitely going to be paid out >> because earnings are going because all these um commitments are coming from hyperscalers. Their earnings are going up. You know, I have yet to see a somebody back out of a deal actually for these commitments. Um, if you look at companies like SpaceX and U, SpaceX specifically, they actually sh sign short-term deals in their contract where they have the ability to back back off these contracts, but these companies necessarily don't. They're not signing those type of deals at SpaceX. So I would say majority I would be shocked if these if all of this I I don't want to say all of it if most of it don't come through which what I think because I have not seen anyone back out of these deals just yet because we have no compute. How much how much do these guys have to spend to actually fulfill that though, right? Like is that something that's just waiting for their current facilities or they have to build more? >> The current and they already they already locked in all the supply and most of the supply that they need, but this is over a course of like 3 to 5 years. So the the buildout still needs to happen for them to you know obviously um have for this to happen. I'll give you an example. Um Nebius is a good example of this. They have a Vinland facility that they're building out and uh their commitment for this year was approximately like or one gaw approximately um of deliver power and villain is like 30% of it. They just got approved to build a phase two of it like last week. So yes, so the buildout still needs to happen for this to all work out. So the biggest bottleneck like like yesterday we talked about is power you know that that is what determines whether if all these gonna you know be able to essentially get built out. >> Yeah that makes sense. So actually I I don't know if you want to mention talk about this slide is kind of what I was asking about their spending but I do want to talk about Nebas. >> Yes absolutely. I do want to touch about the bare case here about Corv because their Q2 like as I mentioned before these companies are not yet profitable. Um because if you look at the chart, their Q2 2026 um revenue was 2.6. They spent 7.1 billion in capex. Um they also pay 640 million in quarterly interest expense, Corvy specifically with a net loss of 626 billion million. Um they also have a capex of uh 35 to $39 billion this year. So essentially what I'm trying to say is Cororev is roughly spending three to every dollar of it earns right now and the model works and the operating leverage is clearly emerging. But this is a huge problem because if something happens with rates, you know, if the rates goes up, that $626 million will be massively higher um if there's an interest hike. And then this in order for all this to happen we need the buildout to continue and that is also um you know up in the air sometime like depending on facility you're trying to build out and all the power you need. So um that is the that is the risk that you have with Corey right now. >> Mhm. Tell me about Nibbius man. Nibbius has been your your darling uh for months now. So how do they compare to this? They don't they don't make as much money right we saw that earlier in the slide. They make like 20% as much but uh clearly the market really likes them. Yes, the market loves them because um because if you look at this chart, this is one of the underappreciated um data point in the entire NeoCloud space right now because Nebius has 5.9 billion deferred and unear revenue on his balance sheet as of June. 5 billion of that is nonasset non-current, meaning customers have already paid Nebas for services that have not been delivered for more than 12 months. So, this is the same like thing that I, you know, talked to you guys about like LTA's like long-term agreements where customers are willing to give you money beforehand and then, you know, you do the build out and they're, you know, you can supply them and customers have already wired Nebius $5 billion for compute. Um, and this is essentially a interestfree $5 billion loan from customers who are not so scared of getting comput access in the future. So they're essentially paying now. So yeah, >> but wouldn't them getting paid in such an advance, wouldn't that make you bearish on the price though? That like wouldn't they already be reporting that in their earnings? Like that they already have the money. Isn't that Isn't that Isn't that bad for the numbers that it's like, well, they already got a lot of the money versus someone like Corewave, it's like that's not reflected yet, that payment. >> What Nebius does in their earnings call, they do something called sandbagging. This is um this is essentially what a lot of companies do. they purposefully lower expectations or lower you know how much revenue that they're gonna bring in because they want to keep on beating earnings you know you know over time this is this has been the core message or core theme for Nebius for like three four quarters now and everyone's like been calling them out on this and that that that is the reason I I think that's the reason why we're not like these numbers I believe should be much much higher they don't want to give you the actual numbers because A market likes consistency. Market don't want you growing let's say 500 800% whatever it is and then next quarter you the the growth fell off and your stock is going to get demolished 30 40% because of it. So you want consistent growth and I think that's what Nebius has shown um across the market right now and because they are the fastest growing as of right now. >> So with that in mind what how is the market pricing each of these like which one is it pricing more aggressively? It's definitely Nebius for sure because if you look at that that chart um that chart compares how expensive each stock is relative to its sales and how fast the revenue is growing. Nebius is in the upper right trading about 36.7 um and last 12 months of sales while revenue grew 454% year-over-year. Core trades at roughly 12.6x 6x um its revenue growth is lower at you know 112%. Um iron trades at 21.x while digital is applied digital is at 21.2x. Um so basically Nebius is the most expensive in the group but investors are willing to pay that premium because it's growing the fastest and if I if you ask me is this justified and I'd say the answer is yes. This is completely justified because of the growth trajectory. They have a killer team. They have actually if you don't know the story behind um Nebius, you should look that up. It's a great story. They actually have a you know they actually used to own a business in um Russia, Google of Russia actually. They you know disbanded during the war and all the engineers you know they had engineers with them for you know 10 decade long. Engineers came with them to build out Nebas. So they have a killer team. They also own like bunch of other businesses line. Um they're they have a ride sharing business. They have click house. They have all these different business line that Corv doesn't have. That's why the market is pricing it. Market likes to price companies that are ahead of its curve and Nebius is ahead of its curve. And if you want an example of that, when Kimmy K3 got released on day like within the first day, well few days after Nebius was one of the first ones to bring that out to the market and they Arrocotti which is like the CEO has said 100 like previously on um in an interview that when the Deep Deepseek moment happened back in you know 2020 like last year sometime essentially when Deepseek released the models that was like you know 100x cheaper, right? Everybody started going crazy, but that was when they had the best sales. Nebia actually had the best sales when Deepseek moment happened. And then Kimmy K3 moment happened where basically they caught up to the frontier models. And I bet you they're about to have an insane insane revenue growth because of Kimmy K3 demand and open source models, >> right? And I feel like there's one of these on this chart that is lagging and catch up soon, but I we'll save that for Milk Road Pro members for people that want to know more. I guess the last question, last thing we want to talk about here, Melvin, uh, is SpaceX as kind of like that's the elephant in the room. I know that's your old joke, but really like >> they're way bigger than any of these companies. So, how do they how do they fit into this? >> Yeah, good question. They I would say they would fit in the competition side of things, right? Yeah. So SpaceX is now generating about $2.6 billion per quarter in AI revenue almost matching exactly as Core's quarterly numbers and they're targeting actually 10 gawatt of capacity. So if you look at SpaceX, they signed $14.1 billion in cloud contract um paying, you know, Anthropic is paying SpaceX 1.25 billion. Google is paying them. There's a bunch a bunch of um you know companies that are also paying them. And if you look at SpaceX and their buildout, SpaceX brought Colossus one and two one and two data centers in 122 days and 91 days, faster than any construction timelines. So this is Elon's moat right there. And Morgan Stanley is now projecting that um SpaceX Neoclap revenue grows from roughly 20 billion in 2026 to 140 billion in 2030. And if that happens, you know, they become a formidable competition for Yelcloud, you know, but but the thing is I don't think this is a bare case. The reason why is because even if SpaceX takes let's say 10 to 15% of the NeoCloud market, the total addressable market is growing so fast that Corore and Nebius can still compound at high rates. >> Right? >> That's the entire thesis in one line. >> And that's it right there, right? is that even even with this competition I I mean basically what I've learned you know in my months during the show and I think this is even a good an even better way to put it like you just said is that it's like the demand is so strong >> too strong >> the demand and the need is so strong and and I think I definitely understand now too this week with something like anthropic being like yo we're actually making money like we we're making real money and we're going to continue to make money finally um really gives confidence in the endgame of all this that it's like AI is actually useful you know what I mean like I think that's always been the thing, right, is it's hard to see um it's hard to see the demand. It's hard to see the endgame demand for um the actually applica the applications that are AI um and especially because the hyperscalers have have some formats of that, but they've kind of blended it into a lot of their other products. So, it doesn't like stand out the same way versus the actual even just the LLM now showing some profit is like really bullish. And then we're just at the start of this this phase of companies making money with AI or new AI companies coming out or whatever you what have you. Um and as a result a lot of these bottleneck companies like these NeoClouds are set to continue to profit right because that demand is not slowing down. >> It's not because let me just give you one last data point. Um please >> um because so right now the top AI ramp AI index monthly spent per employee if you look at the top companies they're spending about roughly $650 per month on AI but if you look at the bottom like median of this um roughly the each company is spending $12 uh which is nothing right like but we've seen a actually acceleration in that they weren't at $12 actually end of this year they were actually at four or $5. So we're going to continue to see that acceleration into 2026 and 2027 because as models gets cheaper more there's more demand therefore more compute is needed and people will start to use it more. So we're going to need massive AI infrastructure buildout. >> And if you guys want to follow along you want to follow Melvin's calls as he's made them. He's had several calls that have done 100% 200% more gains. Uh and his portfolio has been up I think it's like been up 60% on some days since February which is crazy. Uh still up about 50 45 to 50% even today as we're talking. You got to get into Milk Road Pro. And this week we have a campaign running for Pro where we're actually going to raise the price. Uh not yet but at the end at mid next week on Wednesday we are raising the price finally. It's been 25 bucks a month for a really long time as people get all this amazing value including getting to talk to Melvin. You're very responsive in the Discord, which a lot of our members really love. And you get real-time notifications of Melvin's trades and and on our other four analysts as well. Um, and we're finally putting up the price, Melvin. It's finally it's we built something really great here. And you know, and you provided tons of value, and it's time that that's reflected in the price. So, it's gone up to 39 bucks a month. It's going to 299 a year. Uh, but for the next couple days, it's still just 25 bucks a month or 250 a year. And you lock in that rate forever. Okay? That's not just for one month. You get that. It's like prices might keep going up, man. We're doing a great job. You're doing a great job. We should keep raising price prices. So, anybody wants to get in at this original rate, um, now's the time. The link is in the bio. Make sure you smash it and go check it out and and go grill Melvin on the questions. And, of course, see the other Neocloud play that you just made today, Melvin. That's that's the nugget. That's what, you know, that's the extra little bit. We won't give it away. Um, but it's been mentioned a few times on this show. And if you guys know what it is, if you're still listening at this point in the show, drop it in the comments. Let us know if you know what it is. Uh, if you know what that company is. >> They're not going to guess it. They're not going. >> It's too obscure. It's not hot enough. It's not hot enough yet. Uh and that's why I think that's why Melvin's getting in getting us giving us another legendary call. Melvin, great been great to chat, man. Thank you for the education. Uh and I'll see you next week. Thanks for listening to Milk Road. If you enjoyed the show, make sure you like and subscribe. And if you're struggling to find winners in the market, that's exactly what Milkro Pro is built for. Our analysts have called some of the biggest winners early and Pro lets you see what they're buying next, every trade they make, and the research behind every position. Check out Milk Road Pro at the link below. Everything you hear on Milkroad is forformational purposes only. These are our personal opinions, not financial advice, and we may own some of the investments we talk about. Always do your own research and make the decisions that are right for you. See you next time.