“Sell the News” or Cycle Top? The Real Logic Behind the Sharp Drop in Samsung and SK Hynix Stock Prices
- Core Thesis: AI memory (HBM) is the most critical bottleneck in the current AI infrastructure. The three monopolists — Samsung, SK Hynix, and Micron — are seeing surging profits, but their stock prices are falling due to cycle panic. The market exhibits a cognitive divergence between short-term positioning and long-term fundamentals.
- Key Elements:
- Samsung's Q2 2026 profit of $58.5 billion surpassed Nvidia's $53 billion for the same period, with over 94% coming from the AI memory (HBM) division, which boasts a gross margin of 52%.
- The global HBM market is a monopoly of just three players: Samsung, SK Hynix (combined 60% share), and Micron. Each 1GB of HBM consumes 4 times the wafer capacity of standard DRAM, forcing Apple to raise prices across its entire PC lineup.
- Each AI inference requires reloading the entire model weights. New models have 10-20 times the memory demand of previous generations, creating a "memory black hole" effect.
- HBM prices have skyrocketed consecutively: up 90% in Q1, another 50-60% in Q2, with Q3 expected to rise another 20%. Samsung's annual profit has exceeded its total profit over the past 40 years combined.
- Despite record profits, Samsung's stock fell 9% and SK Hynix fell 15% on their earnings days, mainly due to a "sell the news" event combined with cycle panic, as the market fears a slowdown in capital expenditures.
- SK Hynix is listing on the Nasdaq via ADR, raising approximately $30 billion with 4x oversubscription. Leopold Aschenbrenner is a seed investor.
- Capacity from new wafer fabs is not expected to come online until 2030. Demand growth is 3-5 times faster than supply growth, keeping supply tight for years to come.
Collated & Compiled by: Odaily TechFlow
Guest: EJ, Co-host of the Limitless Podcast
Host: Josh, Limitless Podcast
Podcast Source: Limitless Podcast (formerly the Bankless channel)
Original Title: The AI Trade Everyone's Getting Wrong
Release Date: July 9, 2026
Key Takeaways
Samsung’s latest quarterly profit was $58.5 billion, surpassing NVIDIA’s $53 billion for the same period. Over 94% of that profit came from a single division: AI memory. Only three companies globally can produce HBM (High Bandwidth Memory): two in South Korea (Samsung and SK Hynix) and one in the US (Micron). Every AI inference requires re-reading the entire model weights, leading to 10 to 20 times the memory demand of consumer-grade products, a figure that doubles with each new generation. Meanwhile, 1 GB of HBM consumes the production capacity equivalent to 4 GB of standard DRAM, directly squeezing the memory supply for phones and computers, resulting in price increases across Apple's entire product line.
The paradox: memory companies are posting record profits and prices are still rising, yet their stock prices have collectively entered a bear market. Samsung fell 9% on its earnings day, SK Hynix fell 15%. The hosts believe this is a combination of a "sell the news" event and cyclical panic, with no change in fundamentals. SK Hynix listed on Nasdaq via an ADR, raising approximately $30 billion with a 4x oversubscription. Leopold Aschenbrenner participated as a seed investor.
Key Insights Summary
Memory Monopoly's Margins Trump Everything
"Samsung’s gross margin is 52%, SK Hynix’s is 72%. For every $100 worth of memory sold, $72 goes straight to the bottom line. For comparison, Apple's hardware gross margin is around 30%."
"Employees in Samsung's memory division received year-end bonuses equivalent to 6 times their annual salary. South Korea's luxury goods market has seen sales triple over the past 4 months."
AI is a Memory Black Hole
"AI is essentially a black hole for memory, more extreme than any consumer or enterprise technology wave we've ever seen."
"Every time you submit a prompt, whether to ChatGPT or Claude, it has to re-read the entire model weights. Every single time."
"Each new model demands 10 to 20 times more memory than the previous generation. With models at 15 trillion or 20 trillion parameters, this demand will only go up."
The Price Surge Isn't Over Yet
"Prices rose 90% in Q1, another 50 to 60% in Q2, and are expected to rise another 20% in Q3. Samsung earned more money in one year than it did in the previous 40 years combined."
Record Profits, Bear Market Stocks
"Samsung beat earnings expectations on its report day, posting a quarterly profit higher than NVIDIA. Then its stock dropped 9%. SK Hynix dropped 15%."
"Meta hinted at reining in AI capital expenditure. Honestly, they haven't created anything particularly useful either."
"If you bought a month ago, you look pretty foolish right now. But 6 to 24 months from now, these companies will still be very strong businesses."
Main Content
Samsung: The Misunderstood AI Profit King
Josh: Samsung just became the world's most profitable company, but almost no one knows exactly what business drives that profit. I used to think of Samsung in terms of smartphones and computers, but actually 96% of their profit comes from one division: memory. The memory market has only three major suppliers, and coincidentally, two of them are in South Korea. How did the two most profitable companies in the world end up in the same country? EJ, how did Samsung do it?
EJ: When I was younger, I only knew Samsung as a phone seller. Over the past year and a half, I've kept hearing its name tied to AI, and I got curious about what the company is actually doing. Look at the numbers. Samsung's Q2 profit was $58.5 billion, beating the analyst estimate of $55 billion by $3 billion. More importantly, they crushed NVIDIA, which had only $53 billion for the same period. A year ago, Samsung's Q2 2025 profit was just $3.4 billion. From $3.4 billion to $58.5 billion in one year – what happened?
The answer is HBM, High Bandwidth Memory. Samsung is currently the world's second-largest HBM supplier. Why is HBM so valuable? Because every GPU NVIDIA sells, every TPU Google makes, and any AI chip requires massive amounts of memory. Note the word "massive." The amount of memory needed for consumer devices (laptops, desktops) is predictable, but AI's memory demand is exponential.
Samsung makes $650 million a day, $27 million an hour, $7,500 a second. That's more than the world's most valuable companies. But Samsung isn't the world's most valuable company, which is quite interesting.
Memory Triopoly and HBM Monopoly
Josh: Computers need two things: a workbench (memory) and a file cabinet (storage). There are only three companies in the world that can make these.
EJ: Right. Three types of memory, each with its own use. DRAM is like the DDR5 in your computer – the workbench, temporary RAM that clears when you turn it off. NAND is your SSD, the flash storage in your iPhone – the file cabinet, slower but cheaper. Then there's HBM, High Bandwidth Memory, and this is everything in the AI era.
Manufacturing HBM is incredibly complex. DRAM chips are stacked 12 to 16 layers high like a skyscraper, using extremely precise processes. SK Hynix controls 60% of the entire HBM market. If any AI chip is made, it's likely using their memory. The triopoly is established, and profit margins are expanding as a result.
Josh: Here's a straightforward data point. 1 GB of HBM consumes the wafer capacity equivalent to 4 GB of standard DRAM. That means for every wafer shifted to AI memory, there's 4 times the phone and computer memory capacity disappearing. Apple implemented its first across-the-board price hike: MacBook Air went from $1,100 to $1,300, MacBook Pro from $1,700 to $2,000, Mac Studio from $4,000 to $5,300. Because people buying the Mac Studio want to run local inference, which has huge memory demands, and memory is simply not enough.
AI is a Memory Black Hole
EJ: If I had to summarize why people are so bullish on these companies: AI is essentially a black hole for memory, more extreme than any technology wave we've ever seen.
You write a prompt and submit it, whether to ChatGPT or Claude, it has to re-read the entire model weights every single time. What are model weights? They're the parameters that these companies spent billions of dollars training. Every inference re-reads them. And each new model demands 10 to 20 times more memory than the previous generation. With models at 15 trillion or 20 trillion parameters, demand will only continue to push higher.
But that's not even the biggest part. Your chatbot remembering what you said last time, remembering your information across conversations – that temporary storage falls under NAND flash, which also has massive demand. So, demand for all three types of memory is being pulled through the roof simultaneously.
Some people say this is a bubble that will burst. Historically, there's some truth to that; the memory industry has always been cyclical. Three years ago, SK Hynix was almost acquired by Micron – that was the absolute bottom of the cycle. They invested in HBM, not knowing if they could sell it, and Micron almost bought them. In the end, they didn't sell and doubled down on HBM. Now, they're the most valuable company in South Korea.
Price Surge: A Climbing Staircase from 90% to 20%
EJ: Let's look at the prices. SK Hynix and Samsung's pricing over the past 6 months: Q1 was up 90%. Suddenly, the memory in your phone or computer doubles in price because one component jumped almost 100%. Q2 was up another 50 to 60%. Samsung is raising prices another 20% in Q3.
Samsung earned more money in one year than it did in the previous 40 years combined, which is 19 times more than the same period last year.
In terms of gross margin, a grocery store makes $3 for every $100 sold, an automaker makes $7, Apple hardware makes $30. Samsung is at 52, SK Hynix is at 72. For every $100 of memory sold, $72 goes straight to the bottom line. Samsung memory division employees received year-end bonuses equivalent to 6 times their annual salary. South Korea's luxury goods market has seen sales triple over the past 4 months. People in Taiwan are taking out $60,000 high-interest loans from banks to buy TSMC stock. Some wild things are happening in the Asian AI market.
Josh: The price increases are brutal. Consumers are starting to feel the pain. A 32GB RAM stick is 2 to 3 times more expensive than last year. A third of the cost of building a computer is memory. These prices have already passed through to the real consumer market. The question is: how long can this last? Can they keep raising prices?
EJ: It depends on one variable: will the number of people using AI continue to grow? This is the only proxy indicator for whether memory demand will keep increasing. If you believe that in the future, everyone will run multiple AI agents and use AI for work and daily life, then memory demand is exponential.
What about the supply side? New fabs won't go into production until 2030. These factories are incredibly precise to design, you can't just flood the supply in the short term. Demand is outpacing supply by 3 to 5 times, and the next few years will see supply constraints. Over in China, CXMT is making similar DRAM and HBM, but their entire production capacity is being absorbed by China's domestic AI labs. Apple tried to find alternative suppliers in China, but there's no stock available.
Another argument is: what if a new model architecture emerges that doesn't need as much memory? I think the logic is the exact opposite. When memory becomes cheaper, more use cases become viable, more AI agents get deployed, and overall economic output is higher. The demand for memory actually becomes greater.
The Paradox: Record Profits, Bear Market Stocks
Josh: We've been very bullish on memory. Micron is up 150% since we recommended it at the end of last year. But recently, all memory stocks have dropped over 20% from their highs. A technical bear market. Samsung beat expectations on its earnings day, posting a quarterly profit higher than NVIDIA, and then dropped 9%. SK Hynix dropped 15%. Profits are at record highs, prices are still rising, but the market is saying "slow down."
What might have scared the market is Meta hinting at reining in AI capital expenditure. Though honestly, they haven't created anything particularly useful either. This is the strange point we're at: the company says "We're great, profits are at record highs, demand is strong," and the market says "Wait a minute, this has risen too much, there are unknown risks." What do you think?
EJ: My view is simple: this is just sell the news. The quarterly earnings season just ended. Global funds are heavily positioned in these stocks and are looking for a good price to re-enter. If you think I'm just talking my book, I understand. But if you're a long-term AI investor, memory is a fundamental necessity, and only these three companies can produce it. This structure won't change in the short term.
People are worried about the cycle. During the last memory super-cycle from 2017 to 2018, Micron's P/E ratio hit 4 to 5, and then the stock dropped 60%, even while profits were still rising. History does seem to be repeating itself, but the difference this time is: the last cycle was driven by mobile phones, where you could predict the demand ceiling. This time it's driven by AI, and the demand ceiling is invisible.
Here's another interesting historical detail. At the bottom of the last cycle, when memory prices were at their cheapest, who was pushing for the lowest prices and stockpiling memory? Apple. Apple had pricing power over Samsung and SK Hynix, forcing them to supply DRAM and HBM at the lowest possible costs. Now the tables have turned. Samsung and SK Hynix are doing what Apple did back then. In an efficient market, this is perfectly normal.
SK Hynix Lists on Nasdaq
Josh: Next up is a major test. SK Hynix is a South Korean company, but it's listing on Nasdaq via an ADR on July 10, raising approximately $30 billion. This is a critical moment for the memory industry. How the US market prices SK Hynix will determine the next phase of this memory trade. EJ, would you participate?
EJ: Short answer: yes. I am bullish on memory. I hold Micron, and I also hold a DRAM ETF (a basket of memory companies). If you're a US citizen, buying Korean stocks directly is inconvenient, but you can allocate through these basket products. I've been waiting for SK Hynix to come to the US for a long time. I will buy.
Currently, this IPO is rumored to be 4x oversubscribed. Institutions, pension funds, and retail investors are all vying for allocation. The scale of this offering suggests that institutions have done deep research on the memory sector and intend to hold it for the long term.
Josh: Guess who contributed $2 to $3 billion of that $30 billion? Leopold Aschenbrenner. He's back again. Last time we did an analysis of his portfolio, some people questioned his logic for shorting NVIDIA. Well, NVIDIA has dropped 20% since we recorded that episode. Now he's participating as a seed investor in SK Hynix's IPO. You can question his judgment, but he hasn't made a wrong call yet.
