SK Hynix now expects the memory shortage to last until 2030 - PCWorld

SK Hynix anchoring the shortage to 2030 is not a forecast—it is a pricing-power declaration. Echoing the 2017-18 DRAM supercycle, the signal to NVIDIA and hyperscalers is unambiguous: lock HBM long-term agreements at premium rates for five-plus years, or lose allocation priority. Technically, HBM4's 16-high stacking with hybrid bonding consumes three to four times the wafer area of DDR5. Every HBM wafer pulled from the line permanently erases three to four conventional DRAM wafers. This is not a cyclical dip; it is a structural siphon of the entire DRAM ecosystem toward AI compute, systematically squeezing consumer and PC OEMs out of the priority supply queue. In the competitive arena, Samsung's HBM3E qualification delays with NVIDIA hand SK Hynix a 12-to-18-month near-monopoly window on cutting-edge HBM. Micron holds capacity but lacks packaging scale to become a credible third pole before 2026. Expect Samsung to over-invest in P4-line expansion while SK Hynix sustains its scarcity-premium strategy. The 12-to-24-month tail: HBM4 mass production in H2 2025 will trigger a second wave of conventional DRAM price spikes. Server OEMs should brace for 40-to-60 percent memory BOM cost inflation. The 2030 narrative will also pressure governments to classify memory as strategic infrastructure, accelerating subsidy disbursement under CHIPS Act Phase 2 and EU chip frameworks.