SK Hynix’s CEO Warns the RAM Shortage Will Outlast Its Record IPO

SK Hynix pulled in $26.5 billion in the largest Nasdaq listing ever completed by a foreign company, and within days its chief executive told Reuters the RAM shortage squeezing prices worldwide has not even peaked yet. Kwak Noh-jung said 2027, not this year, will be the worst year for memory supply in the industry’s history, with real relief unlikely before 2030.

The timing lands awkwardly. Days after the South Korean chipmaker’s Wall Street triumph, a federal class action accused SK Hynix, Samsung and Micron of engineering the very shortage that just made their shareholders rich.

A Record IPO Lands With a Grim Warning Attached

Kwak made the comments in his first English language interview, according to Bloomberg, following the record breaking US stock offering. He told Reuters plainly: “We forecast that next year will be the worst year in the industry’s history from the supply perspective.”

He did not stop there. “We still forecast that customer demand will remain higher than our supply capacity even beyond 2030,” Kwak said. “But we are doing our best to solve the problem.”

The listing itself was earmarked largely for HBM (high bandwidth memory, the dense chip stacks AI data centers depend on) production capacity. It is not the first time someone at SK Hynix’s parent has floated this exact timeline. SK Group chairman Chey Tae-won previously predicted the shortage might not clear until 2030, a call Nvidia’s chief executive has since echoed.

Customers are not waiting to find out who is right. Kwak told Bloomberg that buyers are signing long term supply contracts because they believe the shortage will last for longer, locking in years of elevated pricing rather than betting on a near term fix.

Why AI’s Memory Habit Is Starving Everyone Else

The mechanism is straightforward even if the fallout isn’t. Hyperscalers building AI data centers need enormous volumes of HBM, and manufacturing it consumes wafer capacity that used to go toward ordinary DDR5 and mobile RAM.

Matteo Rinaldi, a professor of electrical and computer engineering at Northeastern University’s analysis of the shortage’s causes, put the scale in blunt terms: a single AI server can use as much advanced memory as a dozen or even hundreds of traditional laptops. When hyperscalers build thousands of those servers at once, they absorb a large fraction of global memory output before a single consumer laptop gets its share.

Nada Sanders, a Northeastern supply chain professor, described the tradeoff more starkly. If RAM is fuel, she said, most of it is now being consumed by corporations running AI data centers while ordinary buyers are left out in the cold.

The Price Tags Even Analysts Call Shocking

What a Single Gigabyte Costs Now

Research firm SigmaIntel tracked consumer memory prices through the second quarter of 2026, and the swings were brutal even against an already inflated first quarter.

Gartner projects a 130% year over year jump in DRAM pricing for 2026. Ranjit Atwal, the firm’s research director, said it plainly: “The speed at which the memory pricing has increased has shocked everybody.”

A separate IDC analysis of the smartphone and PC impact found Lenovo, Dell, HP, Acer and ASUS confirming 15 to 20 percent device price hikes and contract resets across their lineups.

The Devices Already Priced In

The component math has already shown up on store shelves.

• Apple raised MacBook Pro prices by $300 and set its entry level MacBook Neo at $699, then followed with a second round of hikes worth up to $200 across the wider lineup.
• Microsoft added as much as $500 to Surface pricing and pushed Xbox console pricing from $499 toward $799.
• Meta raised the Quest 3 headset by $100 and pointed directly at memory costs.
• Raspberry Pi increased board prices by as much as $150.

None of these are outliers. They are what happens when memory, once a predictable line item, turns into the single most volatile cost in a device’s bill of materials.

Consumers Take the Big Three to Court

On June 25, 2026, a federal class action landed in the US District Court for the Northern District of California. It names Samsung, SK Hynix and Micron, the three companies that together control roughly 90 percent of the global DRAM market.

The plaintiffs, 14 individuals and three small PC businesses, allege the companies crushed conventional RAM production “in defiance of all economic and business logic,” producing what they estimate is a 700 percent price increase over four years. The suit calls the pattern “concerted anticompetitive behavior” and seeks both damages and an injunction.

It is not the industry’s first brush with this accusation. In 2005, SK Hynix and Samsung were fined $180 million and $300 million respectively after pleading guilty to fixing DRAM prices, and plaintiffs point to a similar pattern between 2016 and 2018 that never reached a legal ruling.

Micron has rejected the current claims outright. “Micron disagrees with the allegations contained in the complaint,” the company said in a statement. “We compete vigorously, fairly and in compliance with all applicable laws wherever we do business. We will defend ourselves against these claims.”

Do Samsung and Micron Actually Agree With SK Hynix?

Mostly, yes, on the worst part. Samsung and Micron both point to 2028 as the year major shortages start easing, a year later than Kwak’s own math implies, but all three companies agree 2027 will be the low point for supply and the peak for DRAM prices.

Micron’s own numbers back up the grim middle years. Chief executive Sanjay Mehrotra told investors during the fiscal third quarter earnings call that the company continues to expect supply and demand for both DRAM and NAND to stay tight beyond calendar 2026, and that Micron can fulfill only “50% to two-thirds” of some customers’ demand. That quarter alone brought in $41.46 billion in revenue, up 345.7 percent year over year, with guidance calling for $50 billion in the following quarter.

Where the story splits is on how much comfort to take from any of it.

• SK Hynix (Kwak Noh-jung) says demand will keep outrunning supply capacity even past 2030.
• Micron (Sanjay Mehrotra) calls the current squeeze only the “first innings” of a multi-year cycle.
• Samsung and Micron, in joint industry messaging, target 2028 for shortages to begin easing, while still calling 2027 the worst year for prices.
• Bloomberg Intelligence analyst Shuli Ren argues the shortage already peaked in the second quarter of 2026.

The Contrarian Bet Against the Shortage Narrative

Not everyone on Wall Street believes the apocalypse story. Ren’s research forecasts conditions easing through the back half of 2026 and into 2027, with a possible swing into outright oversupply by 2028 once new capacity in the US and South Korea comes online.

OpenAI Chief Operating Officer Brad Lightcap has called memory, not GPUs, the AI industry’s real bottleneck. That framing cuts both ways: it explains why hyperscalers keep signing multi-year contracts, and why some investors think the resulting capital spending spree is setting up its own hangover.

Michael Burry, known for his early bet against the 2008 housing market, has been linked to a wager against the memory trade continuing, according to a 24/7 Wall St analysis of his recent positioning. Investor Thierry Borgeat made the cyclical case even more bluntly in a post detailing memory’s historical boom-bust swings, noting Micron shares fell 90 percent after the 2000 cycle topped, 80 percent after 2008, and roughly 50 percent in both 2018 and 2022.

The best commodity business is still a commodity business. Wonderful at the bottom of the cycle. Deadly at the top.

That was Borgeat’s summary, not SK Hynix’s. But it is the exact tension sitting underneath Kwak’s warning: a company celebrating its best quarter in history while telling the world the good times are structural, not cyclical.

When It Actually Makes Sense to Buy

Nada Sanders’ advice to ordinary buyers is simple and a little grim. She said she would be buying new devices now, because they are only going to get more expensive until the underlying problem is addressed.

Buyers whose current laptop or phone can realistically stretch another two or three years have the easier call to make. Everyone else is choosing between paying today’s price or a higher one next year.

Kwak’s own numbers say that wait could run past 2030.

Frequently Asked Questions

When Will RAM Prices Actually Go Down?

Estimates vary sharply by source. Jefferies expects average selling prices to only begin trending downward in 2028, when new fab capacity finally reaches volume production, while Bloomberg Intelligence’s Shuli Ren sees a possible oversupply forming that same year if AI demand growth moderates.

What Is the “Legacy Price Inversion” Analysts Keep Mentioning?

Counterpoint Research has reported DDR4 spot prices reaching roughly $2.10 per gigabit, higher than the roughly $1.70 per gigabit price of HBM3e, the advanced memory used in AI accelerators. Older, simpler memory is now pricier than the cutting edge chip it was supposed to be cheaper than.

Could Car Buyers Feel the RAM Shortage Too?

Yes. A Bloomsbury Intelligence and Security Institute report on the shortage’s market outlook notes automakers still rely on DDR4 and LPDDR4 in most vehicles, and the industry’s three to five year design cycle cannot easily absorb a fast memory transition, raising the risk of delays or spec changes for 2028 and 2029 model year vehicles.

How Much Is Micron Spending to Fix Its Own Supply?

Micron has committed roughly $200 billion to US manufacturing expansion, but meaningful production from that investment is not expected before mid-2027, with its new Idaho and Singapore capacity not shipping real volume until mid-2027 and 2028.

Will Chinese Memory Makers Like CXMT Change the Timeline?

Not soon. Jefferies assessed Chinese manufacturer CXMT’s DRAM technology as 1.5 to 2 generations behind global leaders, making it unlikely to reshape the market through 2027, though large scale overseas shipments after 2028 could make it a bigger factor later.

Could Tariffs Make an Already Bad Situation Worse?

Potentially. Clarkston Consulting’s supply chain analysis of the shortage points to the CHIPS Act’s 25 percent federal tax credit for domestic chip manufacturing as one policy pulling in the opposite direction, even as broader US tariff plans targeting a 15 percent rate on imports threaten to push memory and device prices higher still.