Samsung Electronics and SK Hynix, two of the largest makers of memory chips on the planet, will invest a combined $590 billion alongside the South Korean government to build a sprawling new manufacturing complex in an effort to relieve the AI-driven shortage that has sent the price of computer memory soaring worldwide.
South Korean President Lee Jae Myung, standing alongside Samsung Chairman Lee Jae-yong and SK Group Chairman Chey Tae-won, unveiled the plan on Monday at the Blue House in Seoul. The president announced plans to strengthen the country's leadership in semiconductors and expand its AI infrastructure through large-scale investments and new manufacturing capacity, saying South Korea must move faster than rivals to secure the technologies underpinning the AI era, according to CNBC.
The headline figure — that $590 billion investment — is part of what the South Korean government calls its "Three Mega Projects for the Great Leap Forward." The package spans new memory fabs in the country's southwest, an advanced packaging cluster in the central Chungcheong region, and a supply-chain hub in the southeast. At its core, Samsung and SK Hynix will each build two chipmaking plants — four in total — for 800 trillion won (~$516 billion), the country's industry minister said, with the aim of doubling South Korea's DRAM production capacity within five years.
Thanks for subscribing!
Read the best of Moneywise in 5 minutes or less.
By signing up, you accept Moneywise Terms of Use, Subscription Agreement, and Privacy Policy.
DRAM (dynamic random access memory), for those who don’t know, is the short-term memory inside nearly every laptop, phone, and server. It's what lets a device juggle the programs you're actively using, and it gets wiped each time you power down. For most of the past two decades, it got cheaper and more plentiful, almost predictably so. Then AI arrived, and the math broke.
How "RAMageddon" throttled the world of consumer electronics
The shortage known as "RAMageddon," as Moneywise has explained in its coverage of the crisis, has a structural cause. AI data centers consume massive amounts of what’s called high-bandwidth memory, or HBM — and making it eats up far more factory capacity per chip than ordinary memory does.
In basic terms: When a manufacturer shifts wafer capacity toward HBM for AI servers, it pulls a disproportionate amount of ordinary memory out of the global supply pool. There’s an obvious profit motive here — HBM creates far richer margins than commodity memory — so the three companies that dominate the market, Samsung, SK Hynix, and Micron, have steered their best lines toward AI customers, which has led to everyone else fighting over what remains.
The price data is staggering. TrendForce, the Taipei research firm that’s been tracking the memory market closely, sharply upgraded its forecast in February, projecting that conventional DRAM contract prices would jump 90–95% in the first quarter of 2026 alone — the steepest quarterly surge on record. It then forecast a further 58–63% rise in the second quarter.
Memory, which used to be a rounding error on a device's bill of materials, has become one of its costliest components. As Moneywise previously reported, some vendors even began selling pre-built PCs without any RAM at all — a workaround that would have seemed absurd two years ago.
Then came the moment the crisis stopped being a problem for spec-sheet obsessives and started showing up on ordinary receipts. Apple raised prices on MacBooks and iPads on June 25 — its first formal move to pass higher memory and storage costs to consumers. In its own statement, the company said the consumer electronics industry was facing an unprecedented challenge, pointing to the rapid expansion of AI data centers and the extraordinary surge in demand for memory and storage that came with it.
CEO Tim Cook, who had warned for months that memory costs were coming for Apple's margins, told The Wall Street Journal he had never seen anything like it in over 40 years. Apple's stock fell more than 6% that day, its worst session since April 2025. Hours later, Microsoft said it would raise Xbox prices starting August 1.
The severity of the shortage is why Seoul is committing more than half a trillion dollars to solving it. The hope is that flooding the market with new capacity eventually will bring prices back to earth — for the AI giants, yes, but really for the rest of us.
The circle of life: Booms, busts, and long lead times
If the plan works, it won't work anytime soon, unfortunately.
The memory industry has always run in cycles. You get periods of heavy investment in new fabs, which tends to lead to oversupply a few years later, which then leads to collapsing prices, which leads to manufacturers pulling back, which leads to another shortage, and the cycle continues.
The market saw downturns around 2012–13 and 2018–19, then recovered, but this time it’s different because this isn’t about consumers upgrading into new devices, but rather a once-in-a-generation infrastructure buildout for these AI data centers, where Silicon Valley’s most cash-rich companies are buying up as much memory as possible in an effort to own the “next big thing.”
To be clear, the memory chipmakers are reaping the benefits. SK Hynix, the dominant supplier of HBM to Nvidia, passed Samsung this year to become South Korea's most valuable company, with shares up sharply in 2026 and a market capitalization above $1 trillion. The company is now capitalizing on that run with a U.S. listing. In a registration statement filed with the SEC on June 24, SK Hynix laid out plans to issue up to 17.79 million new shares as American depositary receipts on the Nasdaq — an offering, valued at roughly $29.4 billion, that would rank among the largest share sales in history, eclipsing both Alibaba's 2014 listing and Saudi Aramco's 2019 IPO. Trading is expected to begin July 10.
But building working fabs is another story altogether. SK Hynix's Chey said the project would require a ton of real estate along with massive sources of power, water, and skilled workers, noting it took nine years for the company to establish its major manufacturing cluster in Gyeonggi Province. A modern memory plant takes years to design, construct, and ramp to full yield.
As Fortune reported, most of the projects announced so far won't deliver supply until 2027 at the earliest, and Deutsche Bank expects DRAM to stay tight beyond 2028.
Of course, as with previous memory chip cycles, there’s a chance the same cyclical nature of supply and demand that’s led to the current shortage we’re in could also lead to a surplus. Economists and industry experts have warned of a possible oversupply issue if the current extremely high demand for AI memory weakens over the long term. If you build too much capacity for a boom, and then that boom fades, chipmakers could find themselves staring at warehouses they can't empty.
You May Also Like
- JP Morgan sees gold hitting $6,000/oz before 2027 — and a Gold IRA lets you hold the physical metal while deferring the tax bill. Get your free guide from Priority Gold
- Dave Ramsey warns nearly 50% of Americans are making 1 big Social Security mistake — here’s what it is and the simple steps to fix it ASAP
- Thanks to Jeff Bezos, you can now become a landlord for as little as $100 — and no, you don't have to deal with tenants or fix freezers. Here's how
- Millionaires under 43 are reshaping investing — just 25% of their portfolios are in stocks. Here’s where their money is going
Dave Smith is the VP of Content at Wise Publishing and Editor-in-Chief at Moneywise and Money.ca. His work has also been published in Fortune, Business Insider, Newsweek, ABC News, and USA Today.
