SK Hynix Approved $38 Billion of New Memory Fabs That Won't Produce a Chip Before December 2028

SK Hynix (NASDAQ: SKHY) is putting a historic amount of money into new capacity. On Aug. 7, the memory maker's board signed off on 54.3 trillion won of spending -- about $38 billion -- split across two new plants. Some 35.2 trillion won goes to the Y2 fab in Yongin, which will make high-bandwidth memory (HBM) and next-generation DRAM. The remaining 19.1 trillion won funds the M17 facility in Cheongju, which will make NAND flash memory.

The timelines are the striking part. M17's groundbreaking comes in February 2027, and its first cleanroom doesn't open until December 2028. Y2 waits until July 2027 to break ground, with its first cleanroom opening in June 2029. And a cleanroom opening comes before the equipment goes in, let alone volume shipments. The soonest either plant opens its first cleanroom is nearly two and a half years away.

For a memory market where prices have been surging on short supply, the gap between approving capacity and producing from it may be the most important number in the announcement.

Image source: Getty Images.

Two fabs, two clocks

The projects serve different halves of the memory market, and they run on different schedules. M17, the NAND plant, is the nearer one, with construction starting in early 2027 and the investment running through April 2031.

Y2 is the bigger commitment at 35.2 trillion won, and the later one, with spending planned through October 2031. It's aimed at HBM and next-generation DRAM -- the chips feeding artificial intelligence (AI) processors, and the products behind the company's record second quarter.

So M17's first cleanroom opens about six months ahead of Y2's. Either way, the wait is measured in years.

The company framed the spending as keeping pace with demand. "This investment is a decision made to seize opportunities in line with the market's growth speed," SK Hynix said in the announcement.

Demand is already outrunning supply

The urgency shows in the numbers SK Hynix reported a week earlier. Second-quarter revenue came in at 79.3 trillion won, up 51% from the first quarter and 257% from a year earlier, while operating profit reached 60.5 trillion won, good for an operating margin that expanded to 76% -- both all-time records. DRAM and NAND flash prices both rose significantly from the first quarter, and first-half revenue crossed 100 trillion won (about $70 billion) for the first time in the company's history. The company also began mass shipments of HBM4, its newest high-bandwidth memory in mass production, during the quarter, and said it has finalized long-term agreements with about 10 key customers.

Even that wasn't enough to fill orders. Customer demand exceeds the company's supply capabilities, management said in its second-quarter update.

The supply SK Hynix can actually add before 2029 comes from projects already in motion (its M15X fab and the first phase of the Yongin cluster, whose cleanroom opens in early 2027), not from the two plants just approved.

Years of tight supply?

Memory prices have been climbing because supply is short while cloud computing giants and other AI infrastructure builders keep buying. The new fabs are arguably the clearest signal yet of how long SK Hynix itself expects that to last. After all, a company doesn't commit $38 billion to plants whose cleanrooms open in December 2028 and June 2029 unless it expects demand to still be there when the doors open.

But the same timeline says something about pricing between now and then. No supply from either plant reaches the market before December 2028 at the earliest. Whatever loosens memory pricing before that will have to come from fabs already under construction, from technology upgrades inside existing plants, or from demand cooling. That is a short list.

Shares closed at about $166 on Thursday, Aug. 13, up 7% for the day, and the stock's high since its July debut, $194.80, sits about 18% above that.

The stock trades at about 8 times earnings, and the multiple falls under 4 against analysts' forecasts for the coming year -- a valuation bracing for the boom to fade fast. To me, the construction schedule argues against that timeline. Supply this slow to arrive could keep pricing tight for longer than a multiple under 4 assumes.

Of course, memory cycles have punished confident capacity plans before, and it is usually demand, not construction, that delivers the surprise. But for at least the next two years, the relief isn't coming from these fabs.

Daniel Sparks and his clients do not have positions in any of the stocks mentioned. The Motley Fool has no position in any of the stocks mentioned. The Motley Fool has a disclosure policy.