Everyone Hates AI Right Now. Four Stocks That Are Bulletproof transcript

I want to thank today's sponsor, Copper One Resources. >> This was 50% of the way to a full-blown AI takeover. >> It is going to get out of control unless we do something. We need to slow down. >> I'm talking about there is a possibility this sort of thing could happen. >> AI that escapes its testing environment or blackmails developers to prevent itself from being turned off is AI that is too powerful. Those are four very big stories all hitting in the same week and they're all telling us to slow down.

And if you're holding Nvidia or a 401k that's full of it, then the question is pretty simple. If we decide to put the brakes on AI today, does that mean that the spending is also going to stop? From what we can see, not a single dollar has stopped flowing. Anthropic CEO published that essay telling the industry to slow down. And 17 days earlier, his company had signed a $45 billion worth of computing contracts. And exactly 2 days after the essay, the CEO of Broadcom was asked on TV whether any of it was going to change his numbers. >> Is there anything that's happened in this whole debate about the AI slowdown that would give you pause to that prediction? >> No, not in the least.

We we [clears throat] see the demand for compute infrastructure for AI development or AI frontier models as continuing to be very strong. The spending can't stop and the reason happens to be physical. So fewer than 4% of American data centers can even take a full rack of Nvidia's newest chips. So the old chips are going to stay in the data center and they're going to keep getting paid. A handful of years ago, I actually helped to build out a machine learning pricing model at Amazon.

And that was long before any of this was mainstream. And the habit that I kept from that job is the one that I'm using here. Find the constraint. Because the constraint is going to decide what the next big payoff is going to be. Before I go any further, I am not a financial adviser and I do this for educational purposes. Now, if you're getting any value from my research, then I would greatly appreciate it if you'd press the like button.

That simple act of gratitude truly means the world to me. Now, every big buildout in American history can be measured the same way, which is the money spent building in a year against the size of the whole economy that year as a percentage of the gross domestic product or GDP. Housing at the top of the 2005 boom took 6.6% of the economy. The railroads in the [music] 1870s ran around 5%. And AI, depending on who's counting it, is somewhere between 1 and 2% in 2026. and it's quickly heading to 3% by 2028 if Goldman Sachs is right.

This is a very real buildout and it's on the scale of the ones that your grandparents lived through and it is not the biggest one this country has ever done. So go ahead and keep that in the back of your mind every time that these negative news stories keep coming. Now there is a headwind going into this and here's what I mean. Amazon, Google, and Microsoft are all on track to spend about as much building data centers in 2026 as their entire cloud businesses bring in.

Google Cloud, Microsoft, and Amazon together, they have nearly 1.7 trillion of work that customers have already signed up for. About three years of the entire AI buildout at the 2026 pace. And every single dollar of that needs a data center, power, and chips before it even turns into revenue. And the amount of AI running inside those data centers is growing just as fast as the money. So Google alone is pushing about 330 times more AI through its data centers than it did just two years ago.

And Jensen Wang's own number is that one AI agent doing a job uses 15 to 100 times the computing that a person uses in doing the job. The investments have already been made and the bill for the buildout has already been written. So the question for your money is who is going to get paid for making these investments actually come to life? And it truly comes down to what's physically inside the data center. These are the four buckets that I'm going to be covering today.

So there's the processor that does all of the thinking. And then there's the memory that holds all the conversations. And then there's the light that moves it between all of them. And finally, there's the company that owns the finished machine and it rents it out by the hour. A model's working memory grows with every single word that you add. So a conversation as long as a 300page book that takes about 40 GB just to remember what was said.

And as a refresher, the chip that Nvidia shipped back in 2022, that holds 80 GB. And the AI model itself has to live on that chip as well. So a really big model could be bigger than the chip itself. So if that's the case, the conversation has nowhere to sit on the chip. Instead, it goes into a whole rack worth of separate memory that is nearby. And strangely enough, that is a purchase that data centers never really had to make before.

Micron told its investors back in June that its customers are doing exactly that. Which leads us to the memory group. And when the memory for your conversation sits all the way in another rack, well, the answer has to travel back and forth. And surprise, surprise, copper wire can't carry it fast enough. So, it rides on light through glass fiber, which leads us to the optics group. Then, of course, somebody has to own the finished computers, and they're going to rent them out to the companies that are answering all of your questions.

And an AI that works on a job for five days straight is renting the whole time. That is the inference group. Now, I'm not trying to guess which AI company is going to win. Instead, I want to own the parts that none of them can win without. And every one of those parts sits inside of a data center that already has power and a lease. So, you might be wondering, are those old chips going to get thrown out? Well, every year or two, Nvidia ships a new generation of chip.

And every time somebody on TV says the last one is completely worthless. In fact, I think Jensen Wang said it himself on stage in March of 2025. >> I said before that when Blackwell starts shipping in volume, you couldn't give hoppers away. >> Well, 8 months later, his own CFO told investors kind of the opposite where every chip that Nvidia has ever sold that is still plugged in somewhere is fully utilized. her words, all the way back to A100s that they shipped six years earlier.

Both of those statements are true, and the data center happens to be the reason why. A new rack of Nvidia's best chips pulls 140 kow. More than 20 of the old A100 servers put together, and it needs liquid cooling with pipes underneath the floor. That's why fewer than 4% of American data centers can actually take one. Satya Nadella said it plainly in November of 2025. Microsoft problem, in his words, is not a supply issue of chips.

It's actually the fact that I don't have warm shells to plug into. >> Right? It's not a supply issue of chips. It's actually uh the fact that I don't have warm shells to plug into. >> And if you're wondering, a warm shell is a finished data center with the power and the cooling that's already running. So in this case, a new rack goes into a new data center and the old data center keeps running what it already has because the alternative is an empty room with a lease that's on it.

And you can actually see it in the prices because Cororeweave rents these chips by the hour. And in August of 2026, it signed a contract for A100s. You know that old chip from 2020? Well, it's going to run until 2029. And it CEO said that pricing on the older chips is way above where it was just years ago. That happens to be a bottleneck in more than just one way. And I don't think a lot of people are talking about that because at the end of the day, those old data centers, they can't get more power to feed a new rack.

And a new data center can't get its power hooked up for years. So in the biggest markets, the wait for a grid connection is now somewhere around 4 years. And the equipment that can solve that problem is also years out as well. Every data center and every AI chip in this video, they all depend on copper. That brings us to our sponsor where this segment is disseminated on behalf of Copper One Resources Corp. where copper hit an all-time high back in August, above $6.70 a pound on the comics, up over 40% in just 12 months.

Billionaire Robert Freedelland says there is no rational price for something that you absolutely must have. And Stanley Denmiller called copper the tightest position he has ever studied. So with copper at record prices, copper 1 is down roughly 75% in 2026. And when the Iran war started in late February, junior miners began selling off hard. And Copper One was one of the few raising money right through all of it. And that dilution, well, that's why they hold the largest cash position and working capital in their history.

And it goes without saying that cash matters. Their market cap is about 14.8 million in Canadian and 10.4 million of that happens to be cash. If you go ahead and back out the cash and the market is pricing three copper projects at about 4 million Canadian, Majuba Hill in Nevada, a past producer, plus Red Roanda and Red Hill in British Columbia and Roanda is drilling right now. So, we're sitting with record copper prices, three projects, and cash that's in the bank in what is still a very early stage explorer with real risk.

Please do your own due diligence and check out the link down in the description on copper 1 Resources Corp. So, if the AI labs all completely just shut down tomorrow, none of that is going to change and the old ships are just going to keep earning in the meantime. So, that's the overall landscape. Now, we'll get into the money because a great story and a good entry are two very different things. Every name that I list in this video is going to run through my own investment system.

So, every night, it runs 20 million calculations across my whole watch list. And what comes out the other side is a set of signals that I look into. To put it into simple terms, it says buy a lot, buy, buy a little, hold, or sell. The signal starts with a simple question. What do I think the business is actually worth against what it costs today? So essentially, if I determine what my fair value is, and if the stock is trading below that, then it's telling me to buy.

And of course, I'm also watching where the price has been. So a stock that's climbing is a very different buy than one that's just been sliding. And me personally, I want to be buying into something that's turning up. But on the flip side, my system is also willing to tell me when it has a sell signal, which is a cue for me to look into it a little bit further and make a new decision. I don't know about you, but for me, knowing when to sell is almost just as important as when to buy.

Sometimes it's even more important. And my model's accuracy on giving me guidance on when to sell has been exceptionally strong in seven of the last eight years. And I can statistically prove it. So when the signal begins to show itself, I definitely dive in and I take it very seriously. And the system's very smart and it's always looking for problems because when it finds one, it changes the signal and it tells me to buy less.

And that's a very important part that I want you to pay attention to because it's going to happen on three of the seven names. And of course, I'm going to make sure and show those signals on the screen when we get there. Let's move on to the next group because every new data center that opens is going to need new chips. And the part of the new chip that's hardest to get right now is the memory that's bolted right on side of it.

And that's where our first name is going to be with Micron where Micron is one of only three companies in the world that makes this type of memory. And I think most of you know this that there simply isn't enough to go around. Now you can see what a shortage does to a business. Two years ago, Micron kept 11 cents of every dollar that it sold as operating profit. In its latest quarter, it kept 80 cents. And that is exactly what happens when everybody needs your product and only three companies can actually make it.

And Micron reports again on September 30th, which is the first real look at whether or not this shortage is going to continue to hold. So here is where my system is going to argue with itself. The stock has climbed all year, and it's still well above its 200 day moving average, even after slipping under the 50-day. On the price alone, the signal says buy and it says buy a lot. But in its own 10 years of history, Micron has been this expensive only about 2% of the time.

That's the problem. And it spotted it really quick and it pulls the signal all the way down to just buy a little. So in this case, I just buy a little. And my expectation is that the September report, it's going to tell me whether or not to buy the rest. For the most part, there's two kinds of memory within a data center. There's the fast kind that sits right next to the processor and that was Micron. And then there's the kind that stores the files and the same shortage just hit the storage side.

SanDisk makes the flash storage and it holds the data that a model reads from and it only became its own company in early 2025 when Western Digital spun it out. A year ago, SanDisk was barely even breaking even. I mean, it was treading water. In its latest quarter, it kept 78 cents of every dollar that it sold as operating profit. And for all of that, the price is still under eight times what it's expected to earn next year.

So in this case, I like to look at the PEG ratio, which is the price that you're going to pay divided by the growth that you're getting. And anything around a one or lower tells me that the growth is relatively cheap. And SanDisks is a 0.3. So in this case, the price and the trend both say by a lot here, too. But SanDisk has only been public company for 19 months. So there is no 10 years of history for me to check it against.

So when I can't run that kind of a check, well that happens to be a problem of its own. And the signal comes down to buy. So I buy and I don't lean into it. Now the exact same buildout that's paying those big memory names is also paying the optics names. But the latter reads them completely differently and leans more towards the cell. So, you're probably wondering, why on earth would I sell a part of a company whose sales grew 83% in a given year?

Well, let's start with a name. Lumenum makes the lasers that turn the electrical signals into light. So, the chips in a data center can talk to each other. And a new rack needs far more of these than an old one did. 2 years ago, and Lum was losing money on every dollar that it sold. Now, it keeps 27 cents of every dollar as operating profit. Now, if you go back and you look this one up, you're going to see a headline saying that Lumto completely missed it back in August.

And on paper, yeah, it actually did. But the thing is, the market didn't care. And the stock went up 13% the next day because the business underneath that just kept on doing what it's been doing. And honestly, I think the price may have gone further than the business deserves. So, right now, you're paying roughly 195 times last year's profit for a stock that has normally cost around 70 times. And I get it. The PEG ratio is under a one, which tells you just how fast this is growing.

And the price in my mind has run past it anyways. So in this case, the signal that I'm seeing today is telling me to possibly sell some. The business didn't do anything wrong. The price got ahead of itself. This next company designs the custom AI chips that the big cloud companies build for themselves instead of buying Nvidia's. And one of those customers already owns a piece of it. Marll is the one that cloud companies like to go to.

In August of 2026, Google took the right to buy about 6.7% of Marll and it earns that stake by buying roughly 120 billion of chips through 2033. So in this case, Google has already locked in the next 7 years worth of buying. The problem is that the price has already locked in the next 7 years worth of growth. So you're paying 80 times last year's profit for a stock that has normally cost 33. And I really hate to put my name to this, but this company is getting one of the strongest sell signals in this entire video.

The price just happened to run past the business. So, it's kind of interesting that a lot of the parts makers, they're getting paid upfront and they're getting blocked out business for years. I find it very interesting that the inference groups, the ones that are renting out the finished compute, those are the stocks that just got hammered recently. When it comes to the inference group, the first stock in that group is Cerebrris where they build the biggest chip in the world where a single chip is the size of a dinner plate.

Some customers buy the machine outright. Most of them sign a contract to use it for years and that's 70 cents of every dollar that Cereabris takes in. Now, Cabris did lose $450 million in its latest quarter. And that's the number that scares a whole lot of people off from the stock. But most of that loss never really even left the building because it was stock that was handed to employees when the company went public.

Right now, Cabris has $ 8.6 billion in the bank. And the number that I care about is the backlog. It has $25 billion of orders that are already signed against under $1 billion worth of sales that are happening this year. Unfortunately, right now, the stock is still sliding and on price alone, the signal is saying by a lot. But the problem here is actually bigger than SanDisks. Cerris has only been public for about 4 months.

So there's really no history at all for me to check against. And with that little to work from, my system can only build a fair value one particular way. And that's off of what Wall Street analysts think that it's actually worth. Now on every other name in this video, I'm able to check their numbers against my very own. But in this case, I just can't. So the signal gets notched down one. Like I said, what I am watching for is that backlog.

If it converts on schedule, then hey, the ladder's right with a buy. But then again, if a chunk of it gets pushed out, then that's the moment when I'm going to probably just sit on the sidelines. This next company's kind of interesting because they're betting on everyone else, including the old ones. And it's the company that signed that A100 contract into 2029. Corweave rents Nvidia chips by the hour, and they rent them to the labs and also the cloud companies that can't afford to build new data centers fast enough.

Two years ago, Core Weave sold 400 million in a quarter. Now it sells 2.6 billion in a quarter and customers have already signed for 4.2 gawatts of power when it only has 1 and a half running. It's comical because it's three times what it can actually deliver today. Now, this one definitely touches on risk. And this is where my system did something that I want you to see. Corowe borrow is to build ahead of demand. And over the last 12 months, it burned 13.7 billion in cash in order to do it against 7.6 6 billion in sales.

Sorry for the voice over. Unfortunately, a lot of the information changed between when I filmed it and when it's actually going to go live, but what you see on the screen is accurate. Originally, I was saying that it was on a different rung. But I wanted to make sure that I gave the most accurate information as possible. So, that's why my lips don't match what I'm saying right now. And now, we're going to land with the anchor.

And this is the one that I think every single one of you own. So, what exactly does the system say about Nvidia? Two years ago, Nvidia sold $30 billion in a single quarter. Now, it sells 96 billion in a quarter, and it keeps 66 cents of every dollar as operating profit. And just 2 years ago, it was 62. So, by its own estimate, it is only supplying about 70% of what customers are actually asking for. Now, here's the part that's a little cringey.

Nvidia has guaranteed up to $18 billion of its customers data center leases. So, if one of them can't pay, then Nvidia is the co-signer. They got to pay. And that happens to be a signal that I think the entire world is watching right now. As of today, we're paying 28 times last year's profit for Nvidia. Over the last 5 years, it has normally cost about 52. And the pay ratio while I'm filming this is at a 0.35. That's the cheapest growth on this list.

It's kind of funny because the company at the middle of every scary clip that we saw at the beginning is the only name on this list that's priced like nothing unusual is happening. And here's where my system completely stops arguing with itself. The price says by a lot and it's holding above its 200 day even after a very soft week. And for the very first time on this list, the history check comes back with no problems at all.

So the signal stays where the price is by a lot. But hey, I get it. If you look at the fact that the company is worth over $5 trillion, the obvious question, which I think that everyone is asking, is how much bigger can the company possibly get? Strangely enough, based on the demand, I think it can still get quite a bit bigger. I think one of the biggest takeaways is that dollar cost averaging, which is buying just a little bit every month, no matter what it is, is really going to be all of our friends right about now.

But if the price is sitting on that sales step, chances are everyone else is just overpaying. So, no need to jump into that group. This whole video started with the panic that's surrounding AI and it definitely had an impact on stock prices. For the most part, I feel very calm about what's happening. But here's the thing that I think would really change my mind. All of those signed orders at the very beginning, you know, that $1.7 trillion, if that number begins to shrink, then I think all of this begins to unravel and all of those buy signals completely get thrown out the window.

But until that happens, the labs can argue about slowing down all they want because the data centers themselves and the demand for what's already inside of them is only going to continue to grow. If you're curious, I do share as much of my investment system as I can with the community on Patreon. So, if you want to join and you want to see what we're all working on, feel free. The link is down in the description. As always, thank you so much for watching and we'll see you next