Transcript
0:00 A portion of this video is disseminated on behalf of Greenland Minds. >> In the year 2016, I delivered one pedlops and 128 gigabytes of memory to a small company called OpenAI. This is one pedlops and 128 GB. The the only difference of course is that I can hold this like this and DGX1 took two people to carry upstairs. And so this is the brand new RTX Spark PC. In a matter of 10 years, Jensen just mapped out how the same AI power went from a $129,000 machine to laptops that start at just over two grand. And that is exactly what's going to let AI move out of the warehouses and into the devices that we use every day. Now, something we also need to consider is that obviously in the middle of October, chip stocks had their longest losing streak since 2008.
0:49 about $2 trillion just completely disappeared, which interesting enough, that's about what Australia's entire economy produces in a single year. That sell-off came from a scare over the debt behind these AI data centers, which demand also kept climbing. And unfortunately, these swings have not stopped. And I think the sell-off has really turned into a phenomenal buying opportunity. Even better, Nvidia already gave us a road map for how to look at this. So, back in May, it broke its revenue report into two businesses. The first one is data center, the giant warehouse of chips that everyone's already panicking over right now. The second one is edge computing, which is AI living inside of a PC, a car, and soon to be robots. And that little change in their books means that we can finally watch the edge grow on its own quarter by quarter. Now, the easiest way to picture the two of these is with your own body. When you touch a hot pan, your hand is going to jerk away before your brain even knows that it's hot. That's because the reflex lives in the spinal cord, which is really close to the action. So it never has to wait for a message to go all the way to the brain.
1:49 So a self-driving car that needs to break is in the exact same spot. And so is a robot that's just about to drop a glass. Neither one can wait for a data center 1,000 m away. So the reflex needs to move closer to the device. But the brain is still where the learning is going to happen. So the reflex reports back and the brain works out what went wrong. So you're more careful around the stove next time. So in this video, the data center is the brain and the AI in your phone, your car or the robot is going to be the reflexes. So that full round trip of reporting back is why I think both sides are going to win because every device that gets its own reflexes sends more work back to the brain. The brain and the reflexes are running on very different clocks though.
2:31 So the brain is being built right now and it's where almost all of your money is going as investments and I think it's going to keep growing a lot longer than just another year or two. And on the flip side, the reflexes are barely even getting started. And my view is that they're going to become the bigger business sometime around in the early or mid 2030s. So in this video, I'm going to walk you through it in that order.
2:52 First, the data center and why its growth runs longer than the selloff is already assuming. and then edge computing and why it matters even though it's extremely small. Along the way, you're going to see 10 companies that are going to get paid at every single step. If you're new here, I'm Brian and unlike a lot of other YouTubers, I actually made my millions on my investments so that I could retire at 46. And I'm simply sharing that knowledge as my hobby, which means I'm not a financial adviser and I only do this for educational purposes. Let's go ahead and get started with the data centers that make up the brain because that's where the debt scare hit. But the biggest buyers, well, they never even blinked. So, Amazon, Google, and Meta alone plan to spend roughly $550 billion in 2026. And none of them has cut since the sell-off. Meta reported on October 21st, and it actually raised the low end of its spending plan. And you can see that demand in Nvidia's own numbers. Its data center business brought in about $89 billion in a single quarter, which is more than double a year earlier, while its edge computing business grew about 27%. So with AI moving more locally to your phone and your computer, is it going to kill the data center?
4:00 Well, right now, no. The data center is growing about four times faster than the edge. So all that demand comes with a cost. And for Nvidia, that cost is memory. It has promised suppliers about $279 billion. That's more than double what it was, mostly just to lock up memory. With memory prices climbing, Nvidia expects that over the next two quarters, it'll keep only about 71 or 72 cents of every dollar after making the chip. So, that's down from what it was at 75. Even with that memory squeeze, Nvidia's stock sells at about 23 times next year's earnings, which means you're paying about $23 for every $1 of profit that it's expected to make. For a company growing that fast, that's a pretty low price. So, the whales, the big funds that I track, were buying a little back in June, and we're going to get another look to see what they're doing in mid- November. My signal ladder measures where the price sits against what I think the business is worth. And it has Nvidia sitting right on the line for severely undervalued, the deepest of the undervalued tiers. My investment system, which takes that latter reading and adjusts it for risk, lands in the exact same place. I know that I say this quite a bit, but nobody knows where the bottom is. So, my way of combating that is to dollar cost average on the way down, but once again, that's just my way of approaching it. AI may seem like it's software, but it's grounded in a lot of physical ways, right down to minerals.
5:22 That brings us today's sponsor and this segment, which is disseminated on behalf of Greenland Mines Limited, one of the only handful of companies developing projects in Greenland. On October 14th, the current US administration shared that massive oil, gas, and mineral exploration is getting ready to begin in Greenland. That follows a September defense deal whose article X is built to keep China and Russia from controlling Greenland's mining. That matters because just one company mines platinum and palladium in the US. So we import about 89% of our platinum and nearly 3/4 of our palladium imports come from South Africa and Russia. Rare earths happen to be an even bigger problem because China processes roughly 90% of the metals in high-powered magnets and is restricting its exports. Greenland Mines keeps its resource operations squarely focused on Greenland. Scar Guard for platinum, palladium, and gold and Sarto for magnet rare earths. Neoperformance materials, which is a rare earth processor, holds non-binding rights to buy up to 60% of Sartok's future output. In late September, the company raised more than $59 million from existing investors, most of it at or above $12 a share.
6:26 Investors watch for this kind of setup. A deal that favors the West and a company that's already on the ground, but these projects are still in exploration, so the risk is real. As always, do your own due diligence and check the link down in the description to learn more about Greenland Mines. Now, the other company the brain runs on is Broadcom, which handles the jobs the reflexes can't. When your phone gets a request that's too big for the device, it sends that request to a data center.
6:51 If that data center belongs to Google, there's a good chance the chip answering it was co-designed by Broadcom, where Broadcom builds these custom AI chips for one customer at a time. Google happens to be a pretty famous one, and Meta and Anthropic are also on their list, too. Its AI chip sales hit about $17 billion in a single quarter, which is more than three times what it was just a year earlier. >> But when you want to productize inference, I see it continuing to be very, very strong. When he says productize inference, he means selling AI answers as a product, which is exactly the kind of work that Edge keeps sending back. Unfortunately, Broadcom's customers can't always pay upfront, and its CEO said capital might be the biggest constraint. That's why Broadcom has agreed to back or lend up to $71 billion to just one customer. And Reuters reports that they believe it's anthropic. So, look, the whales are not making a move either way on Broadcom.
7:46 and insiders have actually been selling, including a co-founder that's already on a set plan. Their current price sits at undervalued on my ladder, but my investment system pulls it back to slightly undervalued because the uncertainty around the financing is so high. That is where the brain stands today. And the bigger question is how long is it going to keep growing? Because the whole sell-off was a bet that this spending was really close to the peak. Nvidia, which obviously wants this to be true, says AI infrastructure is heading toward three to four trillion dollars a year by the end of the decade.
8:19 Deloro, an independent research firm, has data center spending passing $3 trillion a year by 2030. UBS expects the big cloud company's spending growth to slow way down by 2028. But I think that's too soon because in my view, big companies and even everyday people haven't even used about 30% of what AI is capable of. The best clue that this buildout runs longer than a year or two is in memory because the brain can't think without it. Is memory a cyclical boom and bust business anymore? Well, historically it has been. In late 2023, the biggest American memory maker was selling its chips for less than they cost to make. But in its latest quarter, it kept about 87 cents of every dollar after making the chip, which tells you just how fast that industry can swing.
9:05 >> This memory is what enables intelligence. It enhances intelligence. >> That company is Micron. What's different this time is that its customers are signing contracts years out, and they're putting real money down to lock in supply. Micron has more than 75% of next year's shipments already committed, and its long-term deals cover over a third of its revenue all the way through 2030. Micron stock sells for about five times next year's earnings. Now, if memory is still that old boom or bust business, then a price that low is a warning because peak earnings can disappear real fast. My own system doesn't have a valuation that it trusts on Micron because its earnings have swung so wildly in the past. So, Micron sits off my ladder for now. My answer to the memory question is that all the building blocks that say it's moved beyond being cyclical, they're there. Whether it's true or not, I don't know. But from everything that we're seeing in the buildout and the trajectory, if it's still cyclical, it's about a 10-year cycle. And that carries the brain's growth all the way into 2030. Micron's margins also gave me the biggest surprise in my research. Memory for phones and PCs, which is on the edge, now keeps about 90 cents of every dollar after making the chip. Its cloud memory for AI data centers keeps only about 83 cents. So dollar for dollar, the edge is already the more profitable business.
10:26 All right, now we're going to move on to the reflexes, which is Nvidia's edge computing business today. It's only about 7 and a half% of Nvidia's sales. And that's exactly why it's worth watching now before it truly gets big. Think back to that hand that was pulling away from that hot pan because the next few companies are each a piece of that reflex. And the first stop is the storage inside your phone and your laptop. SanDisk is the company that makes the flash storage. That's the memory that keeps your photos and your files when the power is off. And its biggest business is literally called Edge, where that Edge business brought in about five times what it did a year earlier, even though its CEO expects phone and PC sales to fall about -15%.
11:08 Interesting enough, the big funds have actually been taking money off the table. Now, my ladder puts SanDisk in its reassess the business tier because the price has dropped below even the worst case value that my system had for it. And that's my signal to go back to the fundamentals before doing really anything else. The next stop is the chips inside the devices. And this is where the obvious names live. The first one is Qualcomm, whose Snapdragon chips run a lot of Android phones, and it pitches the edge harder really than anyone else.
11:38 >> We've been talking about now that AI is the new UI. >> The problem is that memory got so expensive that phone makers are building fewer phones. Qualcomm's phone chips fell about 19% from a year earlier. Now the second obvious company is ARM which collects a royalty on almost every phone chip on the planet. Its data center royalties more than doubled. So it collects from both sides of the spectrum. But investors are paying about 100 times next year's earnings. And by the official accounting rules, it keeps only about 7 cents of operating profit on every dollar. Neither one of them comes in at undervalued on my signal ladder. Now, Qualcomm sits at slightly overvalued, and ARM keeps flipping between slightly overvalued and fairly valued depending on the price of the day. To me, that says the market already gave the obvious names credit for the edge. Now, the company that I'd actually call a pure edge play is a lot smaller.
12:29 It designs the chip that sits behind the camera lens and decides what the camera is going to actually look at. And all of that happens right on the device without sending the video to the cloud. That happens to be Amberella. and its chips go into security cameras, car cameras, drones, and even warehouse robots. But what worries me is that growth has slowed for six quarters in a row, down to about 13%. Sadly, the big funds have been heading for the exits, and about 14% of the shares are sold short, which means a lot of investors are betting that the price is going to fall more, which is going to give more opportunity for the rest of us. Right now, my system has Amberella as fairly valued, but the reading leans entirely on Wall Street analysts. So, I don't put much weight on it. I love what this business does, but I'd really like to see its growth start to tick back up before I really get interested. After the chips comes the robots and cars, which gives those reflexes arms, legs, and wheels. Is the robot army really going to happen?
13:25 >> Great. So, sort of aspirationally 10 million units a year versus a million units a year of Optimus Air, which is insanely difficult to scale production. >> Elon's goal is 10 million robots a year. So, for some perspective, the entire world shipped about 25,000 humanoid robots in the first half of 2026, and most of them came from China. Even Bank of America only expects the whole market to reach around 1.2 million a year by 2030. So, his goal is more than eight times what that is. Sadly, everything he says I take with a grain of salt because back in 2019, he did promise over 1 million robo taxis the next year. a little maybe a year, maybe a year and 3 months.
14:06 but next year for sure, we will have over a million robo taxis on the road. >> Last time I checked, mid-occtober, Tesla only had about 740 robo taxis registered in Texas. Goldman Sachs, on the other hand, which doesn't sell robo taxis, expects about 1 million worldwide by 2030 and about 6 million by 2035, which I think is a little more realistic. My answer to the robot question is yes, it is going to happen, but it's a lot further out than the people selling it really want you to believe. And of course, Tesla is betting the hardest on getting there, and it's paying data centerized bills to train the brain behind those cars and the robots. In its latest quarter, it spent roughly double what it did a year earlier on factories and computing. Tesla's stock costs about 200 times next year's earnings, and the whales have been leaving for three quarters in a row. Tesla's spot on my ladder is built only on outside estimates and the system has it at fairly valued. I'm definitely taking Elon at his word, or rather only about 10% of it. Now, for Tesla, there is a shining light because they do have one piece of the edge technology that's already getting paid for every month.
15:12 About 1.7 million people now pay for full self-driving and for the first time, half of them are on the $99 a month plan. That monthly bill is where the money is going to end up. Every one of those reflexes is still reporting back to a brain somewhere. And the companies that own both ends of that are what I call toll collectors because they're going to get paid whether it's in a phone, a car, or a smart speaker in your kitchen. Will people actually pay every month for AI the same way that they pay for Netflix? If you think your subscriptions for Apple TV, Amazon Prime, Netflix, and Disney Plus are out of control, well, what's it going to be when you're going to have all these different AI subscriptions? Now something to consider is the difference between streaming and AI subscription is that AIS actually have a return on investment. It gives you time whereas streaming just fills your time. The first toll collector is Alphabet which is Google's parent company and it collects on both sides. And then of course we have Apple which announced that its next AI models including the new Siri will be built on Google's Gemini. That means that Alphabet is going to get paid every time that an iPhone thinks. And on the data center side, Google Cloud grew about 82%. Its backlog of signed contracts more than tripled to about $514 billion in just three quarters. And then on the edge side, Google already sells AI subscriptions from about $5 a month up to about $200. The trade-off is that Alphabet is spending cash faster than it comes in. So, it is selling stock to cover that gap. And for Google, the whales are leaning in quietly. As of right now, my system has it as fairly valued. So, most of its good news is already in the price. Now, we'll move on to the second toll collector, which is my old employer, and I think it has the smartest play of the bunch because it gives the edge away to collect the toll somewhere else. Amazon already gives away Alexa assistant free with Prime.
17:04 And Amazon CEO says that people who try it sign up for Prime at nearly 25% higher rates. Amazon also owns the cloud where a lot of that AI work runs. AWS has about $496 billion of signed contracts that it still has to deliver, which is more than double what it was just nine months ago. And right now, the big money is split on Amazon. Some big funds added and Bill Aman's fund cut about a quarter of its stake. So, I read that as really no clear signal. My system just moved Amazon to slightly undervalued, but it reports this coming week, and I really want to see those numbers before I take action. All of this comes down to two clocks on two very different schedules. So today, my ladder finds the most value in the brain. And the reflexes are the bigger business that I'm going to be watching for in that early to mid 2030s. And when Nvidia reports in the middle of November, I'm going to go straight to that edge computing line because I think both sides are going to win. But right now, the brain is growing about four times faster than the reflexes. If that gap starts to close, that's the edge starting to wake up. And when people finally see what it's going to be, I think the impact is going to be a true tsunami. This is the whole data center and edge map all in one place with every company and where my ladder has it. So feel free to just pause and take a screenshot. Five of the companies in this video also report earnings over the next couple weeks and they're all listed here with their dates just as a heads up. If you want to go deeper on memory, my full memory stack map from June is linked down in the description. As always, thanks so much for watching and we'll see you next