My Two Biggest AI Stock Bets Just Reported Earnings

I have been extremely bullish on the AI infrastructure opportunity and over the past 18 months, Nebius and Cororeweave have been the two companies receiving the largest share of my new investment dollars. Both companies just reported earnings. So today, we're breaking down those results and whether my conviction has changed. Let's go ahead and get started. I want to talk about Cororeweave and Nebus earnings. Let's start off with Coree first. Ticker symbol CRWV. They ended up showing off revenue growth of 112% accelerating over the past two quarters. Not nearly as fast as whenever they were a much smaller company. But that's an important thing to highlight by the way whenever you compare it to the Nebuses of the world. I think that it's very hard to put on a ton of new active power each and every quarter. Mind you, Nebius barely even highlighted that number. So to keep up this level of growth at the numbers that Cororeweave is doing is exceptional. They grew essentially from 2 billion to 2.5 billion is essentially the size of Nebius. They grew a Nebius quarter over quarter. And so that's a really really big highlight. On top of that, where they're going going forward, 150% growth is the expectation. This was before Wall Street ended up doing their upgrade. So this is from yesterday. Still unbelievably high. 150% then 190%. I think that these are actually low. It's going to come in higher than this. And the reason is we got such an amazing amount of overall active power, way higher than what we initially assumed. The way to understand active power is active power is the revenue generator. You can't sell GPUs if they're not connected. I mean, you can, but you're not going to actually get payment that's going to go on the books just yet. So, the fact that they raised their fullear guidance active power from 1.7 gawatt to 1.85 85 means that Wall Street is now going to have to readjust how much revenue they're actually going to grow by and that's why the stock is up 19%. So large growth, but this is actually not without Wall Street readjusting their numbers and that's likely to come later. On top of that, they also added around $350 million of new adjusted IBIDA. This ended up showing off an increased amount of margin which was higher than what Wall Street had expected. So that was another really good sign that maybe we're entering an era where these products are more profitable. Now IBIDA earnings before interest taxation depreciation and amortization. There's a big letter there. That's the D depreciation. So that's an important point as we end up buying more GPUs doing over a six-year depreciation cycle. Adjusted IBIDA is taking out the most important factor of the profitability of the Neoclouds, but we'll address that here in a second. The interesting part is is that Wall Street like I said before we adjust for their new numbers is still expected to have increased amount of overall adjusted IBIDA margin rule of 40. So whenever you combine both their adjusted IBIDA and that revenue growth 171% this is growing faster than Palunteer. This is growing faster than Nvidia. This is one of the fastest growing businesses in the entire market at good rates on an adjusted EVA base at good margins on an adjusted EVA base. Cash also saw an increase 5.5 billion, but get ready here. Capex 6.4 billion and debt reached 51.6 billion. Martin Skreyelli came out yesterday and said that he is ready to short Cororeweave because the debt is just getting way out of hand. It's getting to unbelievable scary levels and you don't necessarily need to grow this fast. And that actually might be true to some degree that they could be a little bit more conservative in the way that they're growing, but you have to think about when this debt comes online. As the debt comes online, you don't see revenue from it for a while because you're setting up GPUs into the future. We didn't really see a tick up in revenue growth right now, going from 111 to 112% revenue growth. It's a good place to stay. But I think that the market and too many of these investors like Jim Chenos and Michael Bur and Martin Skrey now is attaching the debt levels of today for the revenue of today. But that's not the way to do this. This is for revenue of tomorrow. This is for the future. And I think that Nebius highlights this the best because they talked about their revenue per megawatt. But essentially, as many people who follow this industry know, a gigawatt data center can cost anywhere between 50 to 60 billion on Nvidia hardware, which is what Core Weave uses and what Nebus uses for the most part. And essentially 50 to$60 billion for one whole gigawatt. A gigawatt is a,000 megawatts. And if you end up putting a $40 million that that's the revenue that you're going to take in per megawatt, just scale that up by a,000, you're looking at $40 billion. And they even said 40 to 50. and sometimes even higher. So you could even look at that as 45 billion or $50 billion on short-term contracts for a gigawatt scale data center that is pointing towards nearly a one-year payback period. So, I think a lot of people are looking at this debt and expecting this to be extremely long-term whenever I actually think that a lot of this debt is very short-term and the amount of money that you're going to make from the data centers is going to immediately pay off the debt that you had previously. And then the new buildouts once you get to 8 gawatt, the next year is not going to go to immediately 25 as the way that they're scaling right now. It's going to go to 9 gawatt, 10 gawatt. And then the data centers that we have today are going to greatly pay off for the gigawatts of the future because it's not going to be leaping by 3x 5x like what they're doing right now for their active power. That being said though, still depreciation is a big core part of this company like what we were talking about with adjusted $1.4 billion is what they're writing off almost essentially every quarter. And that's likely to continue to increase as we have more active power on site. But that active power is revenue generating and as we'll show off here in the future that one it's getting better interests although interest payments are increasing the interest that we're paying on that debt is improving because we're using very creative types of debt financing and putting it against the actual GPU and token amounts. So then if you don't pay back the company that you're offering that debt to, they can take lease of that data center. But listen to this. We recently signed an A100 contract. This is the old 2020 Nvidia architecture that was before Hopper. That's before Blackwell. That's before Vera Rubin. Okay, this was a long time ago. These should have lasted back in the day between 3 to four years worth of the lifetime of those chips. However, they said that they signed an A100 contract that extended into 2029 at attractive prices. As a reminder, this skew was introduced in 2020. Clusters of prior generations of architecture offered installed energized production grade compute that's already running at scale is proven ROI for customers. What they're saying is that the lifetime value, how much they're depreciating by, not Michael Bur's assumed two to three years, not what Coree is telling you, which is six years, but that the chips are actually lasting as long as 9 years with higher than expected cost per hour on those GPUs. What they just showed off there is immense. It's not just for A100s. The assumption is that you're going to be able to do this exact same thing with Hopper, with Blackwell and Vera Rubin 9 years from now. Imagine signing a Vera Rubin contract between the times of 2026 all the way up to 2035. That would be the equivalent. But you depreciated those chips between now and 2032. So you're getting 3 years of just pure profit. That said, we increased the exit ARR annualized run rate of revenue number that we've provided in guidance for you folks right now at 18.5 billion to 19.5 billion for 2026. To put this into perspective, and we'll talk about Nebius here in a second, but they are on the sort of run rate of between 7 to9 billion. It's hard to put into perspective how just truly large Coreweave is and yet the market cap is still quite small compared to the size of the business that they're operating, but investors are worried about that debt. They said in June we brought more than 300 megawatts of active power online, which makes June itself the largest of any full quarter in our history. Makes June alone, that month, larger than any other quarter of active power. This is a mind-blowing stat and it just proves that they're getting faster at setting up these GPUs, knowing how to do these contracts and then delivering these are also at better economics and the lifetime value of each GPU is lasting longer. I mean like from top to bottom the contract values of what they're doing the GPU quality the data center lifetime the lifespan of a data center going from 15 years to 25 years the amounts that they're charging per GPU spot prices doing more short-term loans so they make more value unbelievable value for Coree they said turning to guidance as a result of continued strong execution we now expect to end the year with more than 1.95 gawatts of active power up from our previous guidance of more than 1.7 gawatt. That faster scale to more active power means they can guide higher in the future years to come. They're also launching brand new product lines that are immediately getting to scale. Think about this for a second. In the past few months since its launch, we booked ARR for our managed inference platform that has gone from $1 million to more than $100 million in a single quarter. We now expect to exit 2026 with at least $250 million of managed inference annual run rate of revenue. He said, "We really think about the fact that this is an incredible opportunity for us to offer the most bleeding edge compute that we have, but also a wonderful way for us to access and use GPUs that are coming off of contract in a way to extract maximum value for the company over time. The utilization of this company, they're trying to keep these GPUs running almost all the time, right? There's always some amount of bleed, so they can't say it's 100% utilization, but it's almost like the bacteria spray where they say it's 99.9% utilization. That's what Core Weave is essentially highlighting here. And that they are finding a real opportunity here because of the strength that they have in their technical expertise that they can offer a solution that no one else can, which is just truly bleeding edge latency, bleeding edge access to chips like being the first company to set up to production a Vera Rubin NVL72. So the AI labs that are willing to pay so much higher than anyone else goes to Cororeweave for that speed, for that access to clusters, for that time. So to highlight Roy that was on here yesterday showing off Corewave guidance raised across the board with its first ever tripleB raised ARR did better on fullear revenue, increased their operating income. They saw more active power than before. They are investing so much more than we initially had thought for capex, but that's because they see the opportunity there. So, while it's red, it's still dependent on how you see that. And then obviously, if you're going to increase that capex, then interest expenses are also going to have to climb. They did also say, and you can see this here, they noted that the weighted average cost of debt was reduced by 300 basis points, saving roughly $1.1 billion annually. As they scale, the loans are getting better. Even though the debt is really starting to couple on here, they're not really profitable just yet. They have a ton of debt on the balance sheet. So, you can look at them from an enterprise value point of view, but even then, it doesn't look the greatest until they start paying off that debt. But that price to sales on Cororeweave, it's hanging around 7.7 times. So although the price has been very fluctuating and we're up like 19 about 19% today, the price of this stock is actually some of the lowest areas that we've seen in its history because of that explosive 100% plus growth that we've seen. Now going to talk about Nebius Group, which I think was even better than Core Weeaves, which I couldn't even have imagined cuz Cororeweave just blew the hinges off the door here. Their growth ended up bringing on about $582 million. Let's put that into perspective. over the last 3 years. June of 2024, $14.5 million. Next year, one year later, $105 million. The year after that, $582 million. Pretty insane over a 2-year compounded annual growth rate to go from 14.5 to $882 million. A compounded annual growth rate of 533%. This was on a gross margin of roughly 77.1%. That's not nearly as important as adjusted IBIDA. Adjusted Ibida is climbing up about 41% for the full Nebius group, but on their AI cloud, they actually sported about a 50% growth rate. So that's why even as we scale, Wall Street is expecting and this is before upgrades this morning that we expect to end up growing the margin from 51 to 53% and bringing up adjusted Ebida margin just yearover-year from $236 million to 1.35 billion. That level of growth is hard to look at a valuation and say hey this company is too expensive based on this metric. It's like it won't even be the same company a year from now. depreciation about 260 million much smaller company than where Cororeweave is in terms of the scale but the actual depreciation as a percentage of revenue continues to fall lower and lower. So this can really show off that they are improving in their overall margins whenever you look at depreciation being one of the largest contributors to bringing in net losses. The rule of 40 on Nebia's group as a whole about 495% that includes the adjusted IBIDA that includes their growth rate last quarter was 715%. It's wild to see that growth compound and also by the way compound on the year-over-year comp was 600%. So on top of 600 then they just put up 495%. Just wild over a,000% over two years. They also said, Nebia said, most importantly, we could sell our entire 2027 capacity on the terms that we're offering today if we wanted to, but we're not doing this. We actually see higher value by retaining some capacity, not a lot, to serve shorter term and immediate client needs. They said that the shorter duration capacity, which is typically up to about 6 months for customers with immediate timebounded needs for high value requirements. They said they're ready to pay a significant premium. We're negotiating deals, like I said earlier, for $40 to $50 million per megawatt range and sometimes above that. It's so insane to put that into perspective. $8 billion of cash as you can see this. So, they're happy to continue to build out. Nothing looks bad there. If you look at the pricing that they ended up seeing for 2026, that was $12 million. That was the base for their active contract value per megawatt. So at a gigawatt scale, that's $12 billion per gigawatt. Then in Q2, the deals that they ended up signing, which was across four deals, an average of greater than $1 billion, a large amount of that is already self-funded by prepayments with names like Reflection and Coher, like names that are willing to spend a ton. But the short-term deals are as high as $40 million. So this payback period, they said, was deals closed in Q2. That's based on $20 million. based on Q3 short-term capacity deals, you're looking more towards one-year payback periods or even lower based on the total cost of the data centers that they're setting up. Like, that is so wild. It's hard for me to put into perspective. Imagine taking out a loan and then paying back that loan in one year based off the business that you end up setting up. What is year two? Year two is complete profit. Year three, year four, year five. And these data centers as Microsoft described last 25 years now. And the chips inside the data centers are lasting as core just said about nine years. So whenever you see these companies starting to take out debt, you really have to put into perspective just how much these contract values are growing. Total contract values growing 4x. This is a level of business development that I've never seen in my entire history of investing. And whenever those opportunities come, you have to snatch them up. You have to be willing to put your chips where those investments are. Betting on Nebus, by the way, $8 billion in cash and about10 billion in debt. So even they are going further into debt, but not nearly as much. Core weave about 5x more than where Nebius is. So a little bit different. I actually like the mix of Nebius's business. And by the way, Nebius is not just a pure play data center company. Although they want to be, they still have major investments in companies like Clickhouse that have not had a valuation increase, but they've continued to do well. AVide, Toka, all of these companies have been valued at billion plus dollar valuations just in the stakes that Nebus has. I think Nebus' stake in Click House is somewhere around $3.7 billion. So that's really really good to see. On top of that though, they continue to grow on the AI cloud side. adjusted IBIDA margins are higher than what their total mix is, which is inclusive of some of their other businesses that are not doing as well. They're still doing great. They're great businesses, but just the total business that they're seeing on the AI cloud side is phenomenal. Then they said, and we highlighted this earlier, but that they expect to keep up that run rate of about 7 to9 billion. They want to keep group revenue of about 3 to 3.4 4 billion group adjusted IBIDA margin of approximately 40% in capital expenditures between 20 to 25 billion a lot more orders for Nvidia. The SOFI members are going to really understand this one. After we announced our asset light model, we have had dozens of inquiries from potential partners who have significant capacity and enough capital but do not know how to build or how to sell it. With GPUs increasingly becoming an investable asset, look at the recent news. He's talking about Nvidia's $500 billion financing. We expect more and more companies who will want to come to the market, but will need our help to deliver this capacity to customers. They've already generated revenue in Q2 based on this asset light model. Now, the reason I brought up SoFi is because the loan platform business. They already know how to underwrite. They know how to do the business, but it would be greater if they could use other people's capital and other people's assets to then go out and find those customers for them. Nebus is doing this. They're saying, "Hey, we're so good at building data centers. We are so good at finding the customers that are required to then utilize those GPUs in those data centers." Nebia said, "Hey, let's go out. Let's find people that already have data centers like the smart bird AIs of the world that used to be a shoe company." It's like, "Okay, you have the money, you have the GPUs. Let's go ahead and find you some customers and we will take a flat free, no risk involved, no need to actually hold on to the chips. We'll put our software, you'll pay for this, and then we'll go out and find you customers. It's an unbelievable way to then accelerate your revenue even faster than the capital that you already have. Nebus is looking at this from every single angle and they're one of the most exciting companies doing it right now. On top of that, they also said that they launched their first capacity auction, which was extremely successful and cleared at the highest price that they've ever seen for Blackwell generation of chips, 15% above the highest price that they've ever charged before. They said that this gives them strong signal of the value of the capacity that they have in the market in real time. So all of the money that the Core Weeaves and Nebuses and AWS's and all these companies are making, it might even be at lower prices than what they could truly charge. And whenever you put it into an auction, you say, "Hey, bid." The same way that compute futures might be done. Bid on this amount of compute going forward. They ended up seeing prices raised and that people got even more competitive, bringing those prices up an additional 15%. Really, really great sign from Nebas. Now, this company comes at a little bit more of a premium than what you would catch a core at. obviously comes with much less debt and it comes with additional subsidiaries of the clickous and tas and AV rides and also an asset light business that might not show up on the books just today. So we're looking at a 47 times price to sales, but a company that could quickly become very profitable as we're seeing adjusted IBIDA really start to scale and the percentage of overall revenue that's going towards depreciation lower each and every quarter. Guys, I'm telling you, this was a blowout quarter from the NEO clouds across the board. And then we got to see the additional information from the Lumenums and SMCI seeing that this is demand across the board that there's just too many players here for Michael Bur to come out and say that this is an Enron level scam and a fraud. There's too many people. Someone would spill the beans and yet no one is. I'm a very excited investor as you guys can tell, but likely to just hold my position here and worry about where Nebius is going to be by 2029 2030. I'm not looking for this to be a trade. I'm looking for this to be a long-term hold.