Nvidia JUST Proved the AI Stock Bubble is about to Burst.

Well, the AI bubble is expanding. Holy smokes. I don't think what just happened is a coincidence. And I'm going to show you a few things that should get you to scratch your head a little bit. Think about this. In February of 2026, Google raised $32 billion in just 24 hours for AI infrastructure. And they included in that a 100-year bond offering in the United Kingdom, which was the first time we've seen a 100-year offering since the dot bubble. By June, Google had raised $80 billion. By June 12th, SpaceX came in and raised $75 billion, plus an 11 billion green shoe, plus a $20 billion bond refinance that SpaceX did for their one-year debt that was due next year. Uh we've got about $166 billion raised by just Google and SpaceX alone. Back when this money raise was happening at the beginning of June and earlier in the year, I said, "This is a massive sucking of capital. And guess who's going to win from that capital? The hardware companies. They're going to throw this money at hardware companies and we'll see a hardware rally 2.0." Is it a coincidence that when Google and SpaceX announced they raised $ 160 billion, more than that, it's like 186, right? But over $160 billion that all of a sudden Nvidia announces five minutes into well the earnings call started you know a little bit after two. But is it a coincidence that they announced an increase of their projected growth for next year from 48% growth to 70% growth coincidentally equaling a $160 billion increase in backlog. Think about what I just said. Just Google and SpaceX happen to raise enough to completely justify the $160 billion increase alone that Nvidia forecast in their guidance. Pretty remarkable. And the markets loved it. They ate it up. Why? Because it's more growth than we thought. And the more we can sort of expand the tale of the AI spending, the longer hardware rallies can actually continue. See, post the uh Iran ceasefire early April, we had our first hardware rally. Then we had the hardware sucketing in June and then that ended up with LEO fold at the end of July. Right? That's the money being raised and pulled out. But we know that money is just getting recycled right back into hardware like Nvidia. And so sure enough, here we are. Oh, $160 billion increase in backlog. And when that 70% increase was announced, guess what happened? The stock literally went from negative on the actual earnings release to, oh my gosh, what? You guys are forecasting a 70% increase in growth next year versus just the 48% we were expecting. Boom. The stock shot up. And when the stock shot up, where did it go? Well, you should already know the answer to this because you're a subscriber to the Meet Kevin YouTube channel. Hopefully, you're also a course member. Hopefully, you're also taking advantage of that Jackson Hole August 28th coupon code Jhole. We have actually got a picture of Jay. He took a picture with the coupon board. But anyway, that expires Friday. But you should already know where Nvidia went. Of course, it went straight to the 227 line. Now, you might be wondering, Kevin, did you just place that line there? No. And that's the beauty about these lines is those of you who have been here, you know, damn, look at that pre-market 227 reject. That is going to be the test for today. That is going to be the test to see if we can finally get Nvidia to rerate higher because honestly, it does justify a higher valuation. But I just want you to know where the money is coming from. Now, why does it matter where the money is coming from? You know, because people are going, "Oh, artificial intelligence is booming. This is this is amazing. The bubble's uh you know not a bubble because obviously Nvidia is doing so well. But remember a lot of that is coming from the circular financing and there's some red flags that we've actually got in the earnings call. Let's go identify those right now because that's what we like to do. Not trying to be biased in one direction or another. Just trying to give you facts to think about things with different perspective. So the sucketing led to an increase in hardware spend at Nvidia and it's likely going to increase hardware spend at other companies as well. Duh. We already expected that the 70% might actually be a floor because Nvidia sees this as a supply constrained outlook that there's so much of a backlog greater than $2 trillion that they could actually potentially grow by double if they had enough supply. Uh and they say that customer forecasts for you know Nvidia customers their forecasts point to Nvidia's growth potentially doubling next year. That would be 100% growth. And then I wrote Sam, Dario, and Elon deserve a cake. Although I should honestly include uh Google in there since um I included OpenAI, Anthropic, and Elon in the list. And Google deserves to be on that list, too, since they were part of the great sucking. The great sucking of OpenAI and Anthropic, those are still coming, which also continue to spend money on Nvidia. When they raise money, when Anthropic raises money, guess where you think that money's going? I'll show you in just a second. We'll go through some numbers. Uh we talked about Grog uh with the Nebius partnership already in a separate video. We made a separate fundamental analysis on Nebius. If you haven't seen that yet, go check that out on the channel. You may not have gotten a notification for it. You can also check a breakdown that we did on Dell. It's also on the channel. YouTube might not be recommending those stock breakdowns to you just because they get really niche uh and niche videos don't get as broad of viewership. Uh but you might enjoy watching those especially if you're watching this. Nearly 20 companies including cursor, SpaceX, Figma, and together AI now exceed 1 billion in annualized recurring annual uh well annualized run rate for for recurring revenue. This is really Nvidia trying to justify the spend that's going on. That's fine. None of this so far is a red flag. We got um biotech companies using artificial intelligence. Samsung is using them for computational lothography, uh drug buildouts. I mean, all of this is really cool. So, where are the red flags? Well, this is an interesting one. Enscale just signed a $45 billion contract for six years with Anthropic. Enscale was actually mentioned right here by Jensen. Now, why is that really interesting? Well, if you look right here, you could actually see uh the data for the Nscale deal. The Nscale deal was a 45 or is a $45 billion deal over 6 years. If you just divide 45 by 6, you get 7.5. Okay, we're going to walk through this math really quick, but I'm going to keep it simple for you. Okay, 7.5 billion. Okay, cool. In NScale press releases, they indicate that they have about 40 430,000 chips. It's pretty impressive. And they have total scale of 1.35 gawatt. This contract though is only for a fraction of that. It's for about 460 megaww. So, what we can really do is we could figure out, all right, let's divide these two into each other. This is how many megawws we get per GPU. The contract is for 460 megawws. So 460 * 318 GPUs per megawatt or 460 * 318 gets you about 146,000 chips. Okay, why does this matter? Why are we doing this math, Kevin? This is boring. It's because we could get a cost per hour for the GPUs of about $5.88. But it's not just this. That's not just the only thing that's interesting. We could also get a potential compute cost uh for a gigawatt basis. So the per hour cost of roughly $6 is relatively reasonable for like a Blackwell or a Vera Ruben ship. Nothing unusual there. It doesn't indicate that spot prices are going down. Doesn't indicate that they're going up. What's more interesting is this right here at uh 430 megawatts at 7.5 billion per year, you end up with about 17.4 billion per gawatt. This is a little bit problematic. Why is it problematic? Because Nebius is signing long-term contracts between 20 to $25 billion per gigawatt, also sometimes written as millions per megawatt. Okay, if Nebus is at 20 to 25, and Anthropic just signed this new contract at 17, which is below this long-term multi-year agreement here then uh well, this 17 is below the 20 right here, the 2025. Then why is Elon Musk still trying to justify revenue projections of4 to50 billion per gigawatt? That doesn't necessarily matter to Nvidia. Nobody at Nvidia cares how Elon justifies the spending. But Elon justifies spending on chips because he thinks the revenue potential is 40 to50 billion per gigawatt. And we just had a half gigawatt project on a long-term scale get rented out at $17. That's problematic because it increases the break even time frames from under a year to 3 four years assuming even those $17 contracts can last once all of this larger scale comes online for SpaceX. So, while Nvidia is the funnel collecting the money from the sucketing, you have to realize some of the economics that are underwriting future compute deals that are going into the spending are not lining up. The economics are already not lining up. That's not a red flag. Now, I don't want to sound bearish, baby. This is bullish. Hardware rally 2.0. We've been calling for a hardware rally 2.0. This is great. But let's not delude ourselves. It's gonna end up being a little bit of a bubbly. Okay. Now, it's not just that, but we have to understand what goes into Jensen's arguments. Jensen says the Frontier Labs have fantastic margins. Wait a second. The only way they have fantastic margins is if you go to Anthropic, and we'll see when we get the actual S1, but Anthropic adds back in their training costs. So all the training that they use for their models, they add back in and then they're like, "Wow, look how profitable we are. Damn, we make a lot of money." If you just exclude all of the money we just spent getting here, damn, we are rich and we print money. But come on, man. Those are some cooped up economics. You can't run a business sustainably like that. But that's okay. That's okay. We are bullish the bubble for now. Uh so I mean heck I mean we like I I'll put a bottom line on this in just a moment in terms of where my head is. But but look at this. Look at this. This is another one that's very interesting to know is remember the OpenAI facility that we talked about yesterday in the Nvidia video that I made where they're guaranteeing $15 billion. Okay, $105 billion is the Nvidia guarantee. Think about that for a moment. $ 105 billion guarantee to OpenAI, right? Okay, sounds good. So, if I've got a $ 105 billion guarantee to OpenAI, how much money is OpenAI going to spend on GPUs? Well, OpenAI expects to spend 1 uh or to buy 1.5 million Nvidia GPUs. They have that right here. So, they write it right there. If I take 1.5 million GPUs and I multiply it by about a cost of $55,000 per GPU, what do I get? I get $82.5 billion. That one deal that Nvidia guaranteed with $105 billion of Nvidia's balance sheet doesn't even show up on the balance sheet, right? But that $105 billion of guarantee basically got them $82.5 billion of data center revenue just in GPUs. Add in CPUs, maybe another 10%, add in NVLink, Spectrum, whatever. Okay, but wait a minute. That contract could actually expand. That contract could expand to be a $198 billion guarantee if the same economics are used in the future. So, what you're kind of seeing is this recycling of guarantees into data center revenue. Now, why does $82 billion of data center revenue seem interesting? Well, because if you looked at their last earnings report, they just reported $89 billion of data center revenue. Wait a minute. So, in one quarter, Nvidia reported $89 billion of data center revenue, but they also just guaranteed a project that expects to basically give them nearly a full quarter of revenue because of their guarantee. So they've almost engineered a full quarter of revenue, data center revenue from one customer, open AAI with their balance sheet might. Yeah. Now look, I'm not bearish Nvidia. I think Nvidia is going to profit a lot from SpaceX and Google and Anthropic throwing a lot of money at compute. But let's be clear, it's not obvious that the margins are good for the Frontier Labs to keep spending this sort of money. And it's also not obvious that $40 to $50 per gawatt of compute is realistic when Anthropic, literally the Frontier Lab that, you know, Elon wants to sell to is signing contracts for $17. So there's an insustainability here. There's an insustainability on the Frontier Lab margins. That'll get revealed once we actually see an S1 and we go public with Anthropic. And we are now generating with one deal with OpenAI about a quarter of the company's entire data center revenue that is completely guaranteed by Nvidia itself. So again, the money is going to go from money raised. This is suck. The giant suck. The giant suck goes to Nvidia uh and then all the downstream partners. So this is basically hardware, right? This keeps going. Nvidia keeps winning as long as these companies can keep sucking and they will. They don't care. They will keep sucking. What stops the sucking is probably an AI company like Anthropic going public and then their stock going down and investors saying that's it. we don't like you anymore. We're not going to support the suck anymore. That's when you get concerned about Nvidia. But until then, the bubble can keep on keeping on. And if you like that breakdown, join us over at mekevin.com. We do have a coupon code expiring tomorrow. And if you made it this far, you get a beautiful, beautiful picture of Jay, which has nothing to do with the coupon code Jhole, given that that is a reference to Jackson Hole, but there's Jay anyway. There he is. Go check that out over at mekevin.com. Email us if you have questions at staff at mekevin.com and we'll see you in the next video. Goodbye and good luck. >> Why not advertise these things that you told us here? I feel like nobody else knows about this. >> We'll we'll try a little advertising and see how it goes. >> Congratulations, man. You have done so much. People love you. People look up to you. >> Kevin Praath there, financial analyst and YouTuber. Meet Kevin. Always great to get your take.