SK Group's chairman expects the memory chip shortage to persist until 2030, with wafer (the silicon discs chips are cut from) supply currently running 20% behind demand.

That's the verdict from the top of one of the world's largest memory producers. SK Group controls SK Hynix, the world's second-largest memory maker and a leading supplier of high-bandwidth memory (HBM, the stacked DRAM used in AI accelerators). When a supplier of that scale puts a 2030 date on scarcity, buyers plan around it.

How big is the gap?

The headline number is 20%. Wafer supply — the raw capacity that ultimately caps how much memory the industry can ship — trails demand by that margin, according to the chairman's assessment. It's not a marginal shortfall. In a capital-intensive industry where a new fab takes years and billions of dollars to build, a 20% structural deficit doesn't close in a quarter or two.

That's the core of the 2030 call. The chairman's timeline implies the industry needs the better part of a decade to bring enough new capacity online to rebalance supply and demand.

What does this change for buyers?

The practical question for anyone sourcing memory: what does a shortage stretching to 2030 mean for procurement? Mostly, longer planning horizons and tighter contract discipline.

• Pricing pressure stays. Supply trailing demand by 20% keeps upward pressure on DRAM and NAND contract prices.
• Allocation becomes the norm. Hyperscalers and AI developers that can commit to long-term volume get priority; smaller buyers face the squeeze.
• Capacity announcements gain weight. Every new fab or expansion plan from the major suppliers now reads as a down payment on closing that 20% gap.

Why the date matters

A 2030 horizon is unusually long for a shortage forecast. Memory markets are cyclical — oversupply, price collapse, underinvestment, shortage, repeat. Historically, the cycle turns within two to three years.

A chairman calling for scarcity through 2030 signals he sees something structural rather than cyclical. The demand side, driven by AI datacenter buildouts and the HBM they require, is growing faster than the supply side can realistically respond. Wafers, equipment, cleanroom space, and skilled labor all scale slowly.

It also carries weight because of who's saying it. SK Group's memory arm competes directly with Samsung and Micron. A public shortage forecast from a rival's chairman is effectively a statement about the whole industry's capacity trajectory, not just one company's order book.

The bottom line

Two numbers define the story: 20% and 2030. Supply trails demand by the first; the chairman expects the imbalance to last until the second.

For the trade, the takeaway is planning discipline. If the chairman is right, memory stays a seller's market for years, and buyers who treated recent price strength as a temporary spike may need to reprice their assumptions. If he's wrong, and the cycle turns early, the 20% gap closes faster than forecast.

Either way, the forecast sets the reference point the industry will now argue against.