Did you know that the entire market of affordable laptops under $500 might disappear in the next 2 years for the same reason iPhones are getting more expensive today? This is iPhone 13 Pro Max. I bought it in 2021 for 1200 US dollars at the time. Back then, this was the most expensive phone on the market. 5 years later, we hear everyone talking about memory crisis. Apple is raising prices on their products for the first time in a long time. Laptops are getting more expensive, which frankly feels like something that we should be expecting anyway because well, everything is getting more expensive. And if you paid $1200 for this phone 2 years ago, adding another 200 now, well, it's not like it's going to stop you from getting a shiny new iPhone when it comes out. But Apple raising prices is a small problem in the grand scheme of things. Memory crisis has far broader and bigger implications, and the most serious one is that the entire sub $500 market for devices can disappear. The market of students, schools, and first-time computer buyers. This video isn't just about the RAM crisis. It is about hidden economic consequences that it brings to the tech market and the world as a whole. We will get into the reasons, how and when it started, how it blew up, and what will unfold as it gets worse. Let's dive in.

For the longest time, RAM memory was a commodity. It was cyclical, it was low margin, everyone who needed it had it. It was boring, predictable, and very straightforward. The three main producers of RAM, Samsung, SK Hynix, and Micron, had all burned money in oversupply. All three had a shared trauma in 2019 bust what all three produced way too much, and they simply ran out of people to sell to. So, by 2022, none of them were itching to flood the market again. There's somewhat of an irony with this because when I interviewed for product management jobs in my career, this was a very common question in an interview. How do you figure out how much of product to produce so that you have enough supply during the year, but you want your warehouse to be near empty by the end of it? It is interesting how such a monumental product, such as RAM, had their supply demand predictions quite off. But, I digress. This cycle repeated for three decades. The prices on RAM would rise, manufacturers invest, the supply goes up, then the prices collapse, investment stops, and two years after the investment stops, supply tightens again, and prices recover. Then, the whole thing repeats. The memory business is predictably cyclical, which makes what happened next all the more extraordinary. 2017 and 2018 were bonanza years. Demand from smartphones, data centers, PCs, gaming consoles converged, and RAM prices went through the roof. Samsung's semiconductor division, which produces RAM, was generating 50% of the company's total profit. Micron had its record earnings at the time. The entire industry was flush, and in that flush state, all three companies did what they always do when the money's flowing. They spend aggressively on new capacity. Then, in 2019, the market reversed.

Cryptocurrency mining, which had been an anomalous source of memory demand, collapsed. Why did it collapse? Because it became simply unprofitable. Bitcoin peaked at $20,000 in December 2017 and then fell to $6,000 next year. Mining Ethereum with one GPU chip went from making $150 a month in 2017 to zero next year. And when mining becomes unprofitable, miners stop buying the chips. And when they exit the mining business and no longer need the chips, they start selling them. And when they start selling them, they flood the second-hand market with the chips that they had. And that creates a massive oversupply of memory. It's fascinating how a collapse in one market or one industry can drive such a drastic change in another one. A good example is what actually happened in Kazakhstan 2 years later. When China banned crypto mining in 2021, hundreds of Chinese miners packed their bags and relocated to Kazakhstan. Kazakhstan became the number two Bitcoin mining country in the world, and the country was so flooded with mining hardware that it consumed 8% of the country's entire electricity generation capacity. This crashed the national power grid, and so the government said, "No more mining." So, miners had to shut down and leave. And the ones who decided to exit the mining business started selling their chips. And so, high-end chips that had been sold for $700 in the past dropped to $300.

RAM went through its own version of that cycle. Years of overbuilding, then demand suddenly stops, and all of the extra supply hits the market that doesn't want it. The second reason why the market reversed is the change in the smartphone demand, which also went down because the market simply ran out of new buyers. China, which is the largest smartphone market on the planet, had already sold a phone to almost every single buyer they could have sold to. Just think about it. In 2018, China was selling fewer phones than in 2013. Data center customers that pre-ordered a ton of RAM in the boom years stopped purchasing new chips because they had warehouses full of the chips that they never touched. And so, the supply overpowered the demand. And the memory market was hit by multiple factors at once. And as a result, the prices fell 42%. The total memory market shrank by almost 25%. Samsung profits plummeted 60% in 2019 alone. SK Hynix prices dropped 32%. Server RAM was projected to fall 50% by end of year. Samsung, SK Hynix, and Micron hit the brakes hard. Samsung stopped building its new factory. SK Hynix cut RAM investment to bare minimum. And Micron slashed capital budget from $3 billion down to whatever number they could justify to their investors.

Then COVID hits and the PC and laptop demand goes up again. Everyone working from home, schools buying new Chromebooks, and this spike in demand refilled the memory market. But briefly. And it did so until 2022. The launch of ChatGPT and the beginning of AI era. OpenAI's GPT-4 launched with an 8,000 token context window. By next year, Google's Gemini came out with 2 million. And then Claude and every other frontier LLM. Every major AI lab was in a race to build models that could hold more memory. And every additional token in a context window translates directly into the need for more memory. The problem isn't just that AI needs a lot of memory. It's the type of memory that it needs. The type of memory that it needs takes four times as long to make as regular RAM. And therefore, the demand shifted towards high bandwidth memory, also known as HBM. HBM is essentially a premium form of RAM. The model race caused the data center race. The data center race led to the exponential demand in AI chips. And AI chips caused the demand in high bandwidth memory. And so, all the major AI chip manufacturers who needed the memory for AI chips, such as Nvidia, AMD, and Google's TPUs, started pre-paying for multi-year HBM allocation. And where there is demand, there is supply.

In 2024, the same three names, Samsung, SK Hynix, and Micron, shift production from consumer RAM to HBM. They did not kill RAM production entirely, but it was significantly chopped. Now, why did they do it? Because the margins on HBM versus standard RAM are substantially higher. RAM sits at about 80% margin, and HBM at 87%. In isolation, the 7% difference doesn't sound huge, but in manufacturing, such as semiconductors, small changes in price or margin translate into large changes in profit. A 1% improvement in price can drive 20% operating margin. And if you slice it per gigabyte, HBM sells for roughly $10 per one versus $3 for standard RAM. So, there's a very solid premium and a very solid reason as to why you might want to shift the production from RAM to HBM. Plus, the demand for high bandwidth memory is signed and pre-contracted for years in advance. Your main consumer is a hyperscaler, which is a gigantic B2B contract that is sticky by nature. They're not going to go anywhere. Your consumer is not price sensitive, and their own race for AI supremacy depends on you. So, when you are Samsung, SK Hynix, or Micron, it would not be reasonable not to go for this market, even if you have to change your entire business model or market positioning. And the sequence of these decisions brings us to 2025, when the numbers became public. SK Hynix started pumping out HBM. Samsung cut RAM production and also started pumping out HBM. Micron closed consumer business to switch to HBM. And the ratio of HBM production to RAM became 3:1, and this is what led to the prices. The consumers who need laptops, desktops, or phones were being served by whatever was left after the hyperscalers took their HBM.

Then, in March 2026, a completely unrelated event poured fuel on the fire. Iranian drone hits Ras Laffan Industrial City in Qatar. Ras Laffan is a home to the world's largest helium production facility. Now, why is helium important and what does it have to do with this problem? Because helium is an essential component in semiconductor manufacturing. It is used in lithography, it is used in leak detection, and you need close to 100% purity. So, you can't magically start producing more of it anywhere else in the world. All three giants need helium for memory manufacturing, and all three sourced helium from Qatar. The strikes in Qatar doubled prices on helium within weeks, and with helium being scarce, the same three fabs had to prioritize gas allocation and give all of it to HBM. If there is anything left for RAM, consumers are lucky. If there is nothing, nothing it is, which brings us to today. The pinnacle of the RAM crisis.

What did this crisis lead to? The domino effect. Perhaps the most noticeable one from the business standpoint is that the entire sub $500 laptop market is actively disappearing. See, the thing is, vendors and manufacturers have the ability to absorb costs, but when there is a sharp increase in component cost, a manufacturer gets to the point where they cannot absorb those costs anymore. It starts eating into the margins, and collapse in margins make the business non-viable. Now, on one hand, you may think, "Okay, when was the last time I bought any hardware device such as a laptop under $500?" On the other hand, a sub $500 laptop market has a massive implication on three groups of consumers. One, primary and secondary education that lives off of Chromebooks. Two, emerging markets. South Asia, Asia-Pacific, and even Japan school projects are heavy users of sub $500 hardware. And three, lower-income consumers in developed markets. Students, first-time PC buyers, households on fixed incomes, and so on. When that entire price tier disappears, it locks millions of people and schools out of the basic computing they need to participate in modern education and work.

Two, the entire PC market reprices. Lenovo, Dell, HP, and others all have warned about 15 to 20% price hikes starting H2 2026 by several hundred dollars. The same thing is happening with budget phones. Memory is about 10 to 20% of the phone production and low margin Android manufacturers, Xiaomi, Oppo, Vivo, no longer have a buffer. Now, it is important to note that these brands are not as widespread in North America, but this is very important in the global context. Budget smartphones in emerging markets or in the markets where Chinese exports are prevalent, go through this deepest price increase of any device category. Number four, Apple raising prices. They haven't raised their prices in years and this is the first time that they have publicly surrendered to a crisis in the component market. iPhone price increase on the new phone has not kicked in yet, but TechInsights ran the math and Apple previously spent $39 on RAM in a single device and $13 on flash per iPhone 17 Pro. At current memory prices, those figures go up to $145 and $51 respectively. Apple cannot absorb over $100 increase in RAM costs per device without either raising prices or accepting margin compression. So, to protect its gross margin of about 50%, they have to increase their prices and price iPhones 18 Pro at almost $1,400.

Number five, gaming. Sony raised prices on the PlayStation 5 and the sales crashed by almost 60%. Nintendo raised Switch and Switch 2 prices by up to 20% and their shares fell 10% in a single day. Xbox is expected to follow. The wider implications of gaming market getting hit from this is when hardware sales slow down, game sales will follow.

So, what will happen next and what should we do about this? The only thing that can get us out of this crisis is the increase in supply, which in our case means the expansion of capacity. The industry needs more production and more fabrication plants with a solid incentive to invest in RAM and not switch to HBM. These plants are indeed being built. SK Hynix plans to nearly double their capacity. Samsung resumed Pyeongtaek P5 construction at beast semiconductor factory and targets late 2027. Intel said without much sugarcoating that there won't be a relief until 2028. In the meantime, smaller Taiwanese factories are stepping in to serve regular consumers. Their chips are older and cheaper. They don't perform as well as the newest ones, but this is the reason why they can still afford to sell to budget laptops and PCs.

So, we spoke about Korean, American, and Taiwanese manufacturers, but where would we all be without China? China is rolling up its sleeves with CXMT, China's main RAM producer. CXMT is the youngest of them all. The company is only 10 years old and analysts say that it can provide the supply faster than anticipated. You may ask, "Why be a good Samaritan if everybody else pivoted to HBM and rightfully so?" The main reason is that CXMT does not have extreme ultraviolet machines to produce the HBM. So, they're not doing it not because they don't want to be doing it, but because they can't. Now, there is a chance that we might be seeing somewhat of a whipsaw. If CXMT aggressively expands capacity and the supply arrives in the second half 2027 and the supply surge coincides with any softening in AI spending, prices could collapse as fast as they rose. However, there is a significant constraint. US export controls have effectively capped CXMT's expansion. The whipsaw scenario is possible, but it has a big dependency. If China succeeds in localizing equipment by 2027, the expansion resumes. If not, we're screwed.

So, where does this leave us? Samsung, SK Hynix, and Micron made the right call for their shareholders and shifted to produce memory for AI. The consumers who got hurt on the way up will probably buy into the market on the way down. The companies that produce both will profit from both. This is the reality of the capital allocation in tech these days, and we're just watching how the scenario unfolds. But at least now you know who made this decision, how it was made, and exactly what it costs to everybody else. We hope this was helpful, and we'll see you in the next episode. Bye.