SMIC CEO Zhao Haijun says the industry is “a bit panicked” about the memory chip shortage, and more supply may arrive in nine months; prices are up 80%+ in 2026
More supply could come to market in nine months, executive says — The chief executive of China's largest contract chip maker …
Context & Ripple Effects
The shortage narrative has intensified from warnings of tight 2026 production slots to forecasts that data centers will absorb more than 70% of 2026 high-end memory output. Zhao Haijun's nine-month supply timeline gives that constraint a nearer-term boundary, while underscoring that relief is not immediate.
The reported price increase follows earlier signs that constrained memory availability could raise device costs, including Xiaomi's warning about higher mobile-device prices. It also arrives as CXMT and YMTC pursue their largest planned expansions, positioning Chinese suppliers to respond to the supply gap.
First-order effects
- Memory buyers face continued tight availability and elevated pricing until additional supply reaches the market, based on SMIC's stated timeline.
- SMIC and its customers must plan around a shortage that the company's CEO says has created industry-wide anxiety, rather than assume prompt spot-market relief.
Second-order effects
- Device makers may have to pass higher memory costs into product pricing or adjust component procurement, extending the pressure Xiaomi previously flagged.
- The shortage strengthens the commercial rationale for planned capacity additions by Chinese memory producers, though those additions may not ease the immediate market.
Third-order effects
- If AI-oriented data-center demand continues to claim a disproportionate share of high-end output, memory becomes a broader constraint on hardware pricing and product planning, not just a component-level cycle.
- The episode points to a capacity-lag market structure: supply investments can improve regional supplier positioning, but long build-out timelines can preserve pricing power through the interim.
The trend: This is another marker of an AI-driven memory supercycle in which demand shifts faster than manufacturing capacity can respond.