If You Missed NVIDIA Stock, This is Way Bigger.

If you put $10,000 into Armstock when it went public 3 years ago, you'd have $40,000 today. If you put that money into Palantir's direct listing in 2020, you'd have almost $200,000. Well, SK Hynix makes the memory chips that every AI data center is racing to buy, and they listed on the Nasdaq earlier this summer. My name is Alex and I spent 8 years as an electrical engineer and AI researcher at MIT, which helped me find great stocks like NVIDIA, Micron, and TSMC years before the rest of the market. So let me show you what SK Hynix actually does and if they're worth investing in. Your time is valuable, so let's get right into it. I want to start this video in an interesting place, November 2022 in San Francisco. OpenAI just released ChatGPT and it instantly spread like wildfire, hitting 100 million users in just 2 months. It grew 4x faster than TikTok, 15x faster than Instagram, and 25x faster than YouTube. The wave that followed turned into the biggest technology investment cycle since the internet, and over the next 4 years, NVIDIA would grow from a $400 billion valuation to over $5 trillion because they designed the chips powering ChatGPT and every other generative AI application. Investors that caught it early made life-changing returns, and they caught it early by understanding the science behind the stocks. NVIDIA more than 12x'd in size because their GPUs were perfect for training and running AI models since they could make trillions of calculations per second. They're insanely fast and insanely power efficient, which meant the rest of the data center had trouble keeping up. Normal fans couldn't keep the hardware cool enough, normal networks couldn't move data fast enough, and normal memory couldn't feed the GPUs fast enough so the GPUs sat idle, burning cash instead of generating tokens. And the AI race quickly shifted from who had the fastest chips to who had the data center infrastructure fast enough to feed them, and in many ways, Wall Street still hasn't figured that out. This is why the hyperscalers, Amazon, Microsoft, Google, and meta platforms are on track to spend well over a trillion dollars on AI infrastructure in 2025 and 2026 alone. The GPUs, the cooling, the networks, and the memory. But while we could invest in the market leaders for data center GPUs, cooling, and networking, there was no easy way for most US investors to buy shares of the market leader for high bandwidth memory until this summer. There are three major players in the high bandwidth memory market, SK Hynix, Samsung, and Micron. Of the three, only Micron was listed on US exchanges, so that's the one I covered and invested in. But it turns out that SK Hynix dominates the HBM market with a 50% share, as much as Micron and Samsung put together. And on July 10th, they started trading on the NASDAQ with the ticker symbol SKHY, raising $26.5 billion in what became the largest US IPO by a foreign company in stock market history. According to their IPO filing, they'll use that money to fund new fabs and EUV lithography machines. But SK Hynix has been publicly listed on the Korea exchange for almost 30 years now. They listed on the NASDAQ as an American Depository Share, or ADR, the same way TSMC and Alibaba list on the New York Stock Exchange. Each ADR represents 1 tenth of a Korean common share. The problem is each ADR costs a lot more than that. Let me show you what I mean. If you google SKHY right now, you'll see that SK Hynix trades at around a $920 billion market cap. That number comes from its price on the Korea Stock Exchange, since that's where almost all of the shares actually trade. SK Hynix has 729 million shares outstanding, which works out to about $1260 per share or $126 per ADR. But the ADR currently goes for $175. Same company, same shares, but US investors are paying a 40% premium to buy it on the NASDAQ. Said another way, SK Hynix is worth $920 billion on the Korea exchange, but $1.3 trillion on the NASDAQ. I wanted to make sure I wasn't crazy for buying other ADRs, so I checked them out too. Alibaba trades at the same price in Hong Kong and in New York, since they can freely convert shares to ADRs and back. Taiwan Semiconductor trades at a 12% premium, because Taiwan will only issue new ADRs when the old ones get cancelled, so there's limited supply. And SK Hynix has an even tighter limit on their supply, since only 2.5% of the company can become ADRs in the first place, which is part of the reason it has such a high premium. If SK Hynix ever lets more ADRs exist, this premium should fall accordingly, and so should the price. But that raises the obvious question, is SK Hynix actually worth buying at a 40% premium? By the way, if you've ever tried comparing stocks like SK Hynix, Micron, and SanDisk head to head, you know it can eat your whole weekend, pulling all the numbers, building the spreadsheet, and tracking it after. Or you can hand it off to GenSpark, the all-in-one AI workspace sponsoring this video. I gave its super-agent a single prompt, compared these three companies on valuation, growth, margins, and HBM market share. Then, a few minutes later, it handed me a side-by-side breakdown that would've taken me hours to build myself. It even included all its sources. Then, I dropped that into AgentBase, and had it build a live dashboard that color-codes the winner for every metric so I can see who's ahead at a glance. But here's the best part, I saved that whole process as a skill. So now, it can rerun on its own, and it emails me the moment there's a clear winner with a link straight to the dashboard. And right now, you can try GenSpark with free credits when you sign up with my link below. Alright, in order to decide if SK Hynix's ADR is actually worth buying at a 40% premium, we need to understand the science behind this stock. The most common mistake investors make is assuming SK Hynix is a pure-play HBM company. SK Hynix splits their business into two major categories. The first is DRAM, the fast-working memory a computer uses to hold data that it's actively processing. The second is NAND Flash, the storage that holds your data even when the power is off. In the most recent quarter, SK Hynix's DRAM segment brought in about 73% of their total revenue, and NAND Flash brought in 27%, while non-memory products and foundry services accounted for less than 1%. High-bandwidth memory falls under DRAM. This is the specialized, stacked memory that sits next to every AI chip and feeds it data fast enough so that it doesn't sit idle. And this is what most investors focus on since it's where a lot of the growth and margins come from. But this segment also includes conventional DRAM, the server memory in data centers, the memory in your smartphone, and the memory on your graphics card. This part is priced much more like a commodity, rising and falling with the industry's normal supply and demand cycles. Then there's NAND Flash, which is basically data storage. SK Hynix's headline products here are their enterprise-grade solid-state drives for data centers, the kind of high-density storage that AI build-out can't get enough of. A big piece of that business comes from Solidime, the SSD business that SK Hynix bought from Intel for $9 billion. Solidime specializes in storage density. Their flagship drive holds 122 terabytes in a single unit, and AI data centers buy them by the rack to hold training data. As a result, SK Hynix is now the second-biggest enterprise SSD supplier in the world, only behind Samsung. Every model that gets trained, every data set that gets stored, and every system checkpoint gets saved somewhere on NAND Flash. That's why the enterprise SSD market more than doubled in size in a single quarter. SK Hynix's NAND revenue more than quadrupled year over year, thanks to higher prices and higher sales volumes. So these drives really are a second AI hardware business hiding in plain sight. As a result, their total revenues came in at 79 trillion yuan, or about $55 billion for the quarter, which is up 257% year over year. Here's how I've been thinking about SK Hynix after reading through their IPO filings and their latest earnings. Their NAND Flash business is 60% bigger than SanDisk's entire company, even though that's all SanDisk makes. And their DRAM business, which includes HBM, is about 25% bigger than Micron's. Like I said earlier, about half of every dollar spent on high-bandwidth memory goes to SK Hynix today. But you should know that that's down from a 64% market share this time last year, since Samsung was actually first to mass-produce HBM4, which caused Samsung's market share to jump from 21% to 33% in a single quarter. Either way, SK Hynix is the second biggest global supplier of DRAM and NAND storage overall, so investing in them means owning the entire memory layer of the AI revolution. Another common mistake investors make is assuming memory is still a cyclical, low-margin business. Before the AI era, memory makers didn't set their prices. The entire computing industry set them based on supply and demand. When demand outruns supply, prices and margins spike. But when supply outruns demand, they both collapse. So, memory makers lived and died by how well they could predict demand and how much of that demand they could fill without going over. Case in point, SK Hynix posted an operating loss of about $5.9 billion back in 2023, which was the last major down cycle for memory. But watch how fast that formula flips when demand grows much faster than supply, which is exactly what's happening in the AI era today. Last quarter, SK Hynix made $79 trillion in revenue, and $65 trillion of that was operating profit. In dollars, that's roughly $55 billion in revenue and $42 billion in operating profit, which works out to an operating margin of 76%, the highest in the company's history. Those are the kinds of numbers we usually see at Silicon Valley software startups, not 43-year-old hardware companies. Now let me tell you something that will put you ahead of almost every Wall Street analyst covering this stock. Conventional DRAM prices still move with the normal commodity cycle, but high bandwidth memory is different. It's built for a specific customer's specs and sold on pre-negotiated agreements, so the number of chips and the price get settled before the chips are ever even made. So, high bandwidth memory is priced less like a component and more like a major enterprise contract. Think about what that means for the risks we just talked about. A commodity company can't predict what their products will sell for for more than a year out. But this year, SK Hynix signed contracts with about 10 major customers that lock in volume and pricing terms for 5 years or more, and they all require deposits up front. As contracts make up more and more of their revenues, SK Hynix grows their sales, their operating margins, and they also become much more predictable, which shareholders love because it lowers their overall risks. But contracts can cut both ways. Last quarter, memory prices jumped between 30 and 60 percent, and SK Hynix had more of its revenue locked into fixed price contracts than Samsung or Micron, so they captured less of that growth as upside, which is why they missed Wall Street's expectations for operating profits and the stock fell close to 10 percent even though they reported record earnings. The same contracts that protect memory companies when prices fall can also cost them some of the upside when memory prices rise. That's the price for long-term predictability. But hyperscalers aren't just buying memory, they're knocking down the doors. On their latest earnings call, SK Hynix said that their HBM, DRAM, and NAND capacity will stay tight because new fabs take years to come online. Micron said the same thing on their earnings call, and Nvidia's earnings showed that their purchase commitments more than doubled, from $119 billion to $279 billion in a single quarter, primarily to lock in memory supplies. So let's talk about why memory supplies are so tight in the first place. Like I said earlier, modern AI processors are so fast that they end up spending a significant portion of their time waiting for data to arrive from memory. If the GPU is the engine of AI, then high bandwidth memory is the fuel line. It may not be obvious by looking at me, but I'm not actually a car guy. Oh, what's that? It's actually extremely obvious? Well, stick with me anyway. If the GPU is the engine, then HBM is the fuel line and data is the fuel. If the fuel line is too narrow, it doesn't matter how fast the engine can go, it will constantly be waiting for more fuel. So what determines a GPU's real performance, which is how much math it actually does per dollar or per watt when it's inside a real data center, is how fast you can feed it that data. That doesn't just mean higher memory bandwidth, it also means more power efficiency and better cooling. If we look at their roadmap, SK Hynix started shipping HBM4 in volume last quarter, and in June, they shipped samples of their next generation HBM4e memory to major customers. Each stack holds 48 gigabytes across 12 layers of memory, running at up to 16 gigabits per second per pin, which works out to about 4 terabytes per second of bandwidth per stack. Or said in English, that's enough bandwidth to transfer around 300 full-length 4K movies to the GPU every single second. If the GPU is the engine, then HBM4e is a massive fuel line. On top of that, it's over 20% more power efficient and has 17% better heat resistance than the previous generation, so it runs cleaner and gets less hot. See, maybe I'm a car guy after all. Here's why the supply constraints won't go away overnight. High bandwidth memory takes about three times the wafers to make compared to standard memory, process yields are lower, it takes 4-5 years to build the specialized fabs needed to make it in the first place, and unlike software, you can't just patch in new updates every few days. Another thing to consider is that SK Hynix is a fully integrated memory manufacturer. They own and operate their own factories, and they do their own wafer fabrication. They also do their own packaging and testing end-to-end. They even stack and bond the DRAM dies into finished high bandwidth memory cubes before TSMC mounts them next to NVIDIA's GPUs. And SK Hynix just broke ground on their first packaging plant in the United States, with production slated to start in 2029. So, memory will be supply constrained for years to come, because not only does each generation of HBM need to keep up with current GPU and network speeds, but the designs, the factories, and the fabrication process all need to be set up and coordinated years in advance. Speaking of years in advance, there's another part of their initial SEC filing that I want to highlight. On June 7th, NVIDIA and SK Hynix announced a multi-year partnership to co-develop next-generation memory built around NVIDIA's AI roadmap, spanning four of their next-generation platforms, VeraRubin AI supercomputers, the Vera CPUs, RTX Spark PCs, and even Jetson Thor for robotics. Six weeks later, they also signed a letter of intent on a partnership around memory supplies and AI factories. That partnership is valued at over $500 billion dollars. The first VeraRubin AI factory, powered by SK Hynix's HBM4, is planned for 2027. And even though NVIDIA qualified all three memory makers for VeraRubin, SK Hynix is expected to supply about two-thirds of the HBM4 for it this year. Here's a funny story from Computex 2026 in Taiwan. Jensen Huang walked up to an HBM4e wafer at the SK Hynix booth, wrote please make more on it, and signed it. Whenever I try to do that, I just get escorted out by security. The thing is, this isn't just a vendor filling purchase orders. Vendors can be swapped out on price. This is a long-term co-development partnership, where NVIDIA designs their chips around what next-generation memory can deliver, and SK Hynix's memory gets designed for NVIDIA's roadmap years before the chips even ship. So, SK Hynix isn't waiting for NVIDIA to hand them a spec sheet. They're helping write it. Alright, now that you understand the company, let's talk about the stock, the valuation, and the risks. And if you feel I've earned it, consider hitting the like button and subscribing to the channel. That really helps and it lets me know to make more deep dives like this. Thanks, now let's talk about SK Hynix stock. I think the best way to analyze their valuation is by comparing it to stocks we already know, like Micron for DRAM and SanDisk for storage. Here's a table summarizing everything I've covered. I built it by pulling numbers from each company's latest earnings, and I'm showing SK Hynix twice because the ADR price is significantly different from their price on the Korea exchange. I tried to make each row as apples-to-apples as I could, but all three companies have different fiscal years, different customers, and SK reports their numbers in the South Korean won, so the numbers might change a little based on the conversion rate. So, treat this as a good way to compare these companies, but not as official audited numbers. Let's start with their size. Micron is worth a trillion dollars and SanDisk is worth around a quarter of that. SK Hynix is worth 920 billion dollars on the Korea exchange and 1.3 trillion when you buy the ADR on the Nasdaq. All three companies are growing revenues extremely fast. SanDisk is up 372% year over year, Micron is up 346%, and SK Hynix is up 257%. SK Hynix isn't really growing slower than the other two. They're just growing from the biggest revenue base. And all three companies have operating margins that would embarrass most software companies. 80% for Micron, 78% for SanDisk, and 76% for SK Hynix. But here's where things get interesting. Back in 2018, SK Hynix put about $3.5 billion dollars into Kioxia, a Japanese NAND flash memory maker spun out of Toshiba. Kioxia's stock skyrocketed this past spring. So, SK Hynix had to report a $45 billion dollar gain, which inflates their net profits. I removed it since that's a one-time gain that has nothing to do with selling memory chips, which is why my table shows SK Hynix trading at around 12 times earnings and the ADR trading at around 16 times earnings. That's still less than Micron and SanDisk, which have price-to-earnings ratios above 20. And when it comes to their forward PE ratio, meaning what analysts expect them to earn over the next 12 months, SanDisk trades at a 7.3 and Micron trades at a 6.8. SK Hynix trades at a 4.2 on the Korea exchange and even with the 40% premium for the NASDAQ-listed ADR, it still trades at a 5.8. That's about a 15% discount to Micron, even after the ADR's premium. Alright, so is SK Hynix worth buying, even with the ADR trading at a big premium? SK Hynix is the clear leader when it comes to high bandwidth memory, and their enterprise storage business is a second stream of AI revenues that Wall Street is still ignoring. On top of that, they have 76% operating margins on hardware that will stay supply-constrained for years to come, and they're co-developing their next generation of memory with Nvidia. All while their stock trades cheaper than Micron and SanDisk, even after that 40% premium. In fact, the company itself is spending 40 trillion won, or 29 billion dollars, buying back its own shares after calling them undervalued in a recent SEC filing. Just to be clear, they were talking about the Korean shares, not the ADR, and that 40% premium is only because 2.5% of the company's shares can be ADRs at once. If that number grows, the premium shrinks, which means the price of the ADR would likely fall to close that gap. So, make sure you keep that in mind. In my opinion, if you have any way to buy the shares directly, that's the way to go. For example, Interactive Brokers has offered the stock itself since May. Otherwise, the ADR is actually still fairly cheap relative to the other memory makers, you just need to be ready for the price to drop if SK Hynix issues more ADRs. The rest of the risks are ones we've already covered. Samsung caught up on HBM4 production and is taking meaningful market share. Long-term contracts protect SK Hynix from falling memory prices but they also cap their upside when memory prices rise. And every memory stock on my table is priced around 75% margins, so if margins start to fall, prices might fall with them. But based on everything I've learned and everything I've covered in this video, I think SK Hynix is currently a great way to get rich without getting lucky. And if you want to see even more stocks I'm buying to get rich without getting lucky, check out this video next. Either way, thanks for watching and until next time, this is ticker symbol you. My name is Alex reminding you that the best investment you can make is in you.