The Korea memory chip shortage is the reason your next laptop costs more than your last one. However, almost nobody outside the industry can name the companies responsible. Two of them sit about forty kilometres apart, in Korean industrial towns most foreign readers have never heard of. Between them, Samsung Electronics and SK hynix decide how much memory the planet gets — and what the rest of us pay for it.
Meanwhile, the effects have landed everywhere at once. Xbox consoles jumped 60 percent in a single announcement. Apple raised prices on nearly every Mac and iPad it sells. Even the Raspberry Pi, the hobbyist board built to stay cheap forever, quietly went up by sixty dollars. In each case, the underlying cause traces back to the same place: a supply decision made in Pyeongtaek and Icheon.
This is the story of how a Korean industrial duopoly repriced global computing, why it happened so fast, and when — if ever — it ends.
The $500 Laptop Quietly Died
For roughly fifteen years, the budget laptop was a fixed point in consumer electronics. You could always find something serviceable for under five hundred dollars. That category has now effectively disappeared.
The reason is arithmetic rather than strategy. Memory used to account for somewhere between 15 and 18 percent of a PC’s bill of materials. By mid-2026, industry estimates put that figure closer to 35 percent. In other words, the single most commoditised component in a computer became its most expensive constraint.
Manufacturers responded the only way they could. IDC’s analysis of the shortage shows Lenovo, Dell and HP all confirming contract price hikes of 15 to 20 percent. As a result, IDC now expects global PC shipments to fall between 4.9 and 8.9 percent in 2026, with average selling prices climbing 4 to 8 percent depending on how bad things get.
Smartphones face a milder version of the same squeeze. Memory represents 15 to 20 percent of a mid-range phone’s component cost, and around 10 to 15 percent for flagships. Consequently, IDC models a shipment decline of 2.9 to 5.2 percent for the year.
None of this reflects weak demand. Rather, it reflects a market where the product simply costs more to build than it did eighteen months ago.
How Two Korean Companies Came to Price the World’s Memory
DRAM is the working memory inside every computer, phone, server and console on Earth. Three companies make almost all of it. Two of them are Korean.
According to Counterpoint Research’s Q2 2026 rankings, Samsung held 39 percent of the DRAM market, SK hynix 26 percent, and America’s Micron 25 percent. Together, the three control roughly nine-tenths of global supply. Korea alone accounts for around 65 percent.
That concentration did not happen by accident. It was built over four decades of deliberately brutal capital spending, in which Korean firms kept investing through downturns precisely when competitors retreated. Japanese and European rivals exited one by one. What remained was a duopoly with extraordinary pricing power — a structure Seoulz has explored before in the context of the HBM chip war and the cash piles these companies now sit on.
Interestingly, the rankings themselves have been scrambled by AI. SK hynix held 39 percent of DRAM revenue in Q2 2025; a year later it held 26 percent. The company did not shrink. On the contrary, it posted record earnings. Samsung simply expanded faster, and Micron grew its revenue roughly fivefold over the same period.
The Chicken Game That Built the Duopoly
Foreign readers often assume Korea’s memory dominance is a recent, AI-era phenomenon. In reality, it was won during a series of price wars so vicious that Korean analysts still call them the chicken game.
The pattern repeated across the 1990s and 2000s. Whenever demand softened, memory prices collapsed below production cost. Most manufacturers responded sensibly, cutting output and conserving cash. Korean firms did the opposite. Backed by cheap domestic credit and conglomerate cross-support, they expanded into the downturn, pushed prices lower still, and waited for weaker rivals to bleed out.
The casualty list is long. Germany’s Qimonda filed for insolvency in 2009. Japan’s Elpida, itself a merger of survivors, followed in 2012. Taiwan’s producers retreated into niche roles. By the mid-2010s, a market that once had more than twenty serious players had three.
That history matters now for one reason. The same instinct that made Korean firms invest through pain also makes them extremely disciplined about supply. They have spent thirty years learning that flooding the market destroys value. Therefore, when AI arrived offering premium prices for premium memory, the strategic response was almost reflexive.
Critics call it restraint that looks a lot like restriction. Executives call it capital discipline. Either way, the behaviour is entirely consistent with how these companies have always operated.
Inside the Korea Memory Chip Shortage: The HBM Trade-Off
Here is the mechanism that broke the consumer market, and it is simpler than most coverage suggests.
AI accelerators need High Bandwidth Memory, or HBM — stacks of DRAM dies bonded vertically and wired directly to the processor. HBM sells for many times the price of ordinary memory. Crucially, it also consumes far more manufacturing capacity per gigabyte, because stacking requires more dies, more wafers and more packaging steps. Industry estimates put the wafer consumption at roughly three times that of conventional DRAM for the same delivered capacity.
Every fab has a fixed number of wafer starts per month. Therefore, each wafer diverted to HBM is a wafer that never becomes a DDR5 module for a gaming PC. Deutsche Bank analysts have described the situation bluntly as a zero-sum game.
Given the price gap, the allocation decision was never in doubt. Hyperscalers were committing hundreds of billions of dollars to AI data center buildouts, and they were willing to pay almost anything for memory bandwidth. Consumer PC makers were not. Naturally, capacity followed the money.
The result is a strange kind of shortage. There is no shortage of factories, no natural disaster and no export ban. Instead, there is a shortage of willingness to make the cheap version of the product.
The Numbers Behind the Korea Memory Chip Shortage
The price data is genuinely difficult to believe.
DRAM contract prices rose roughly 172 percent across 2025. Then the second quarter of 2026 delivered another jump of about 60 percent. Research from SigmaIntel shows what that looked like at module level: a 16GB DDR4 stick went from $137 to $207, a 51 percent rise, while a 96Gb LPDDR5X module climbed 89 percent, from $77.10 to $145.90.
DDR5 fared worse still. Modules that retailed around $90 to $100 before the squeeze now command two to three times that.
Third-quarter forecasts suggest the pace is finally cooling, though not reversing. Analysts expect DRAM contract prices to rise a further 13 to 18 percent quarter-on-quarter, with NAND up 10 to 15 percent. Jefferies, by contrast, sees a harsher path: 30 to 40 percent in the fourth quarter, followed by another 40 to 45 percent through 2027.
Corporate commentary tells the same story from the buyer’s side. HP finance chief Kevin Parkhill told investors that memory prices were “roughly doubling versus the prior quarter,” and warned that operating profit would fall below the company’s long-term range for the rest of the year.
Meanwhile, IDC expects 2026 DRAM supply growth of just 16 percent year-on-year — below historical norms, in the middle of the largest demand shock the industry has ever seen.
What the Korean DRAM Price Surge Did to Your Shopping Cart
Abstract percentages become concrete very quickly once you look at retail shelves.
Gaming consoles took the worst of it
Consoles are sold at thin margins and stuffed with memory, which makes them uniquely exposed. In April 2026, Sony raised the PS5 Disc Edition from $499.99 to $649.99, and the PS5 Pro from $699.99 to $899.99 — both 30 percent increases.
Microsoft went further. On 1 August 2026, the Xbox Series X moved from $499.99 to $799.99, a 60 percent jump. The Series S rose from $349.99 to $599.99, or 71 percent. For a console positioned as the affordable entry point, that is close to a repositioning of the entire product.
Valve’s Steam Deck followed suit, with the 512GB OLED model climbing from $549.99 to $789.99. Nintendo, whose Switch 2 uses comparatively modest memory, escaped with a $50 increase, from $449.99 to $499.99.
Even the Raspberry Pi could not hold the line
The Raspberry Pi Foundation built its reputation on price stability. Nevertheless, The Register reported a second increase in two months during early 2026. The 16GB Pi 5 went from $145 to $205. The 8GB model rose from $95 to $125.
Notably, the 1GB variants did not change price at all. That detail is the whole story in miniature: the cost lives entirely in the RAM.
Apple broke its own pricing rule
Apple almost never raises prices on existing hardware. In June 2026, however, it did exactly that across Macs, iPads, the HomePod, Apple TV and Vision Pro. Fortune reported the MacBook Pro moving from $1,699 to $1,999, with some iPad models up by as much as $300.
When the most vertically integrated hardware company in the world cannot absorb a component cost, the squeeze is real.
The enterprise bill nobody talks about
Consumer prices get the headlines. Nevertheless, the larger transfer of wealth is happening quietly in corporate procurement.
Server DRAM has risen even faster than the consumer variety, because hyperscalers bid against each other for guaranteed allocation rather than shopping on price. Cloud providers absorb those costs first, then pass them through in instance pricing and storage tiers. In practice, that means small companies renting a few virtual machines are paying an AI premium on infrastructure they use for payroll software.
Enterprise buyers have also lost flexibility. Multi-year supply agreements that once locked in falling prices now lock in rising ones, since manufacturers have little incentive to commit volume cheaply. As a result, procurement teams that used to negotiate discounts are negotiating for availability instead.
Why Samsung and SK Hynix Cannot Simply Build More
The obvious question is why Korean manufacturers have not simply added capacity. In fact, they are — at enormous scale. The problem is that memory fabs run on geological timescales.
SK hynix has committed to a domestic investment programme reported at roughly $712.5 billion, spanning the Yongin cluster and expansions in Cheongju. Even so, the first Yongin fab is scheduled to begin operations only in May 2027, and DRAM fabs typically take a year to eighteen months to ramp fully. Analysts therefore expect meaningful market impact in 2028 and 2029. Cheongju’s M17 fab starts construction next year and may come online in 2029.
Samsung’s timeline runs similarly long. Mass production at its P5 megafab is not expected until late 2028.
There is a subtler constraint too. As the JoongAng Daily has reported, much of today’s capital spending goes into cleanrooms and infrastructure rather than immediate wafer output. Industry sources describe cleanroom space as the binding constraint across the sector. Consequently, new Korean capacity is unlikely to arrive before the second half of 2027, and each maker is expected to add roughly 150,000 wafers per month when it does.
In the meantime, manufacturers are squeezing efficiency wherever they can. Samsung has reportedly weighed moving legacy memory back-end work to Vietnam in order to free up domestic capacity for HBM. Micron, for its part, exited the consumer Crucial brand altogether.
The Antitrust Shadow Over the Memory Shortage
Rapid price increases in a three-firm market attract lawyers. Predictably, they arrived.
On 25 June 2026, seventeen plaintiffs — fourteen individuals and three small PC businesses — filed Garciaguirre v. Samsung Electronics in the U.S. District Court for the Northern District of California. As Tom’s Hardware detailed, the complaint alleges violations of Section 1 of the Sherman Act by Samsung, SK hynix and Micron.
The central claim is pointed. Plaintiffs argue that the coordinated pivot toward HBM served as cover for curtailing DDR3 and DDR4 production, and that memory costs rose roughly 700 percent over four years as a result. They are seeking class certification, an injunction and treble damages.
History gives the allegation weight. SK hynix pleaded guilty to criminal DRAM price fixing in 2005 and paid a $185 million fine.
Whether the case succeeds is another matter. The defendants can point to a straightforward economic explanation: AI buyers outbid everyone else. Still, the litigation guarantees that the Korea memory chip shortage will be argued in a courtroom for years, regardless of how the market resolves.
China Is Coming — Slowly
One further variable deserves attention. Chinese memory makers are growing at rates that look almost fictional.
Counterpoint recorded CXMT revenue up 716 percent year-on-year in Q2 2026, with Taiwan’s Nanya up 690 percent. Those numbers start from small bases, admittedly. Even so, they signal something important: the commodity DRAM that Korean firms abandoned is exactly the segment Chinese producers are now filling.
For Samsung and SK hynix, this creates a genuine strategic dilemma. Vacating the low end is profitable today. However, it hands a rival the volume, the yield learning and the customer relationships that eventually make higher-end competition possible. Korea has watched this movie before, from the other side, in displays and shipbuilding.
When Does the Korea Memory Chip Shortage End?
Forecasts diverge sharply, and the disagreement is itself informative.
Micron’s chief executive has suggested the shortage runs through 2027. SK hynix’s leadership has floated 2030. Between those two dates sits a difference of hundreds of billions of dollars in consumer electronics pricing.
The optimistic case is straightforward. New Korean fabs come online from late 2027, Chinese supply absorbs the commodity tier, AI capital spending normalises, and prices drift back toward trend by 2029.
The pessimistic case is equally coherent. AI demand keeps compounding, HBM4 and its successors consume even more wafer capacity per gigabyte, and every increment of new supply gets absorbed before it reaches consumers. In that scenario, today’s prices are not a spike. They are the new baseline.
For now, one signal favours the pessimists. Global chip market revenue is projected to reach around $975 billion in 2026, with memory potentially exceeding 40 percent of it — an all-time high. Markets that profitable rarely rush to correct themselves.
What This Means for Korea
Domestically, the shortage looks less like a crisis and more like a windfall.
Korea’s July 2026 exports reached $98.89 billion, up 62.8 percent year-on-year and the second-highest monthly figure on record, behind only June’s $102.2 billion. Semiconductors alone contributed $41 billion, a 179 percent surge, topping $40 billion for a second consecutive month. The trade surplus came in at $30.32 billion.
Yet dependence of this magnitude carries obvious risk. When two companies drive national export performance, national economic data starts behaving like a sector chart. Investors saw what that looks like during the KOSPI crash earlier this year, when a memory-led rally reversed violently in a matter of weeks.
The distributional politics are also getting complicated. Record chip profits have not translated evenly into household income, which has fuelled a domestic argument Seoulz covered in the Korea AI wealth gap. Semiconductor tax receipts are on track for record highs, but the boom is concentrated in two companies and a handful of supplier towns.
Those towns are worth understanding, because they explain the political weight of the industry. Pyeongtaek, Icheon, Cheongju and now Yongin function less like industrial parks and more like company cities. Housing prices, school enrolment and local tax bases all move with fab construction schedules. Consequently, capacity decisions made for global market reasons land as local economic policy, and Korean governments treat them accordingly.
There is a hedge underway as well. Seoul has been funding a domestic push into AI processors, an effort Seoulz examined in Korea’s AI chip startups. The logic is defensive: selling memory into someone else’s AI boom is lucrative, yet it leaves the country dependent on foreign chip designers to set the pace.
What It Means for You
A few practical conclusions follow from all this.
First, buying decisions have changed. If you need a machine, buying now with maximum RAM is likely cheaper than upgrading later, since aftermarket module prices have risen faster than whole-system prices.
Second, specifications will quietly shrink. Rather than raise headline prices further, vendors are shipping less memory at the same price point. Watch the spec sheet, not the sticker.
Third, used and refurbished hardware has become unusually good value. Machines built during the cheap-memory era carry component costs that cannot be replicated today.
Finally, expect the phrase “AI tax” to keep spreading. Consumers who never touch a chatbot are now subsidising data center construction through the price of a laptop. That is an uncomfortable framing for the industry, but it is difficult to argue with the arithmetic.
The Bottom Line
The Korea memory chip shortage is not really a shortage in the traditional sense. Factories are running. Wafers are flowing. Profits are at records.
What changed is allocation. Two Korean companies, plus one American rival, decided that AI infrastructure was worth more than consumer electronics — and the global price of computing adjusted accordingly. That decision was rational, legal in all likelihood, and enormously profitable.
It also made the world’s computers more expensive for everyone else. Whether that resolves in 2027 or 2030 depends on choices being made right now, in fabs in Pyeongtaek, Icheon and Yongin, by people most consumers will never hear of.
