$38 Billion Approved, First Cleanroom in 2029 - Read the Calendar, Not the Headline

The Money

SK hynix's board signed off roughly 54 trillion won, about US$38.1 billion, on August 7:

• 35.2 trillion won for the Y2 DRAM fab at Yongin
• 19.1 trillion won for the M17 NAND fab at Cheongju

Both sit inside a longer plan of around 1.1 quadrillion won, with roughly 600 trillion won earmarked for Yongin and 100 trillion for Cheongju.

TSMC did the same thing four days later. On August 11 it approved another US$29.44 billion, its third authorization of 2026, taking the year to close to US$105.7 billion.

The Calendar

Now the dates, which is where this story actually lives.

| Fab | Construction starts | First cleanroom | |---|---|---| | Y2 (DRAM, Yongin) | July 2027 | June 2029 | | M17 (NAND, Cheongju) | February 2027 | December 2028 |

A cleanroom opening isn't production. It's the point where tools start moving in, and meaningful output lands well after that.

So the honest reading of a 54-trillion-won approval isn't that relief is coming. It's that the people closest to this market looked at demand and concluded it justifies capacity that can't ship a part for about three years. Nobody spends that against a spike. They spend it against a plateau.

Why More Wafers Won't Fix HBM

DRAM is expected to stay tight through 2027. HBM bit shipments should grow 50-60% year over year in 2027 and still fall short of demand, because HBM eats considerably more wafer input per bit than conventional DRAM. Adding wafer starts doesn't add bits proportionally.

That's arithmetic rather than sentiment, and it's why supplier pricing power is expected to hold right through next year.

What It Means for Your 2027 Plan

Nothing announced this month changes what's available in 2026 or 2027. If a capacity headline has been quietly softening your risk assessment, take it back out.