Samsung and SK Hynix executives say they expect global memory supply shortages will persist until 2027 due to AI demand, after the companies reported earnings
Kim Jaewon /Nikkei Asia:
Context & Ripple Effects
This earnings-season warning follows reports that major memory makers were already operating near capacity and that 2026 output had been largely allocated, tying the shortage directly to AI-led demand rather than a short-lived inventory swing.
Later coverage reinforces the duration of that constraint: Samsung said customers were placing orders for 2027 memory supply, while SK Hynix’s chairman said added capacity was unlikely to close the demand gap until around 2030.
First-order effects
- Samsung and SK Hynix gain clearer demand visibility and remain constrained in how much memory they can supply through 2027.
- AI infrastructure customers face continued difficulty securing memory volumes, particularly as production slots are already tight.
Second-order effects
- Customers are pushed to reserve supply earlier, a pattern reflected in orders being placed a year ahead, which can widen the gap between committed demand and available output.
- Rival memory suppliers and the wafer supply chain face pressure to add capacity, but the reported capacity lag limits how quickly the market can respond.
Third-order effects
- If AI demand remains durable, memory availability becomes a binding constraint on infrastructure build-outs rather than a routine component-procurement issue.
- The cycle points to a more reservation-driven memory market, where long lead times and capacity planning carry greater weight for both chipmakers and AI customers.
The trend: This is part of an AI infrastructure supercycle in which memory capacity is struggling to scale as quickly as demand for AI systems.