Bernstein: Surge in Chinese Semiconductor Capacity Ignites Wafer Equipment Supercycle

The global AI computing power competition and the ongoing escalation of Sino-U.S. technological games are pushing the Chinese semiconductor equipment market into a new super-expansion cycle. Bernstein significantly increased its forecast for the Chinese wafer fabrication equipment (WFE) market in its latest report, pointing out that the order growth rate of Chinese equipment manufacturers far exceeds market expectations.

Bernstein raised its forecast for the Chinese WFE market size for 2026 to 2028 from $55B/$59B/$61B to $58B/$67B/$77B, with the largest revision for 2028, increasing by $26B, a 26% increase. The core driver for this adjustment is the unexpected acceleration of capital expenditures in memory.

CXMT and Yangtze River Storage synchronize multi-factory expansion

Bernstein's channel survey shows that CXMT and Yangtze River Storage have significantly accelerated their expansion plans: both companies will build a new wafer factory each in 2026, add two more each in 2027, and release even more capacity in 2028.

A key structural advantage deserves attention: Chinese memory companies only need about one year to build clean rooms, while global peers such as Samsung, SK Hynix, and Micron usually require two to three years. This means that in the next two years, the pace of capital expenditure release for Chinese memory will far exceed global competitors. At the same time, both companies are expected to generate tens of billions of dollars in cash flow in 2026, with IPO fundraising, ample capital ammunition, and the financial constraints for expansion have been basically eliminated.

AI localization drives demand positive feedback

The report reveals a logic chain that has been less systematically elaborated before: the continued tightening of global DRAM and NAND supply has greatly increased the willingness of domestic customers to adopt local storage chips, especially DRAM, which has been unable to scale due to technology gaps.

As Nvidia continues to be blocked by both China and the U.S., the localization process of China's AI computing power accelerates, and the autonomous demand for HBM and high-end solid-state storage climbs simultaneously, further driving DRAM and NAND capacity investment, forming a superimposed demand effect. Bernstein labels this logic as an additional upside risk beyond current storage forecasts.

Localization rate welcomes structural leap, 2027 is the key turning point

Bernstein provides a clear localization rate timetable: the self-sufficiency rate is about 21% in 2025, expected to rise to 26% in 2026, jump to 34% in 2027, and reach 43% in 2028. Among them, the localization of DRAM equipment in 2027 is considered the most important structural turning point.

Before 2025, the localization process mainly focused on advanced processes and NAND, with mature processes and DRAM progressing slowly. However, the latest feedback from channel surveys shows that the performance of domestic mature node equipment has reached a level comparable to foreign competitors, and the price is about 20% lower. This marks a shift in the driving force of localization from "policy compulsion" to "economic spontaneity," making the substitution process more sustainable.

Import data decline is a "false signal"

From the beginning of 2026 to now, China's WFE import has declined by 13% year-on-year, and some market participants have doubts about the equipment's prosperity based on this. Bernstein clearly points out that this is a misreading of the data. It usually takes about a year for Chinese equipment manufacturers to confirm income from orders, while foreign manufacturers such as ASML and Applied Materials can confirm upon shipment; the decline in import data reflects the mismatch of the shipment time window, not a real decrease in demand.

The guidance for the decline in China's revenue share from the five major global equipment manufacturers (ASML, Applied Materials, Lam, KLA, and Tokyo Electron) is fundamentally due to the higher demand for advanced processes and DRAM abroad, increasing the overall denominator, rather than a contraction of China's absolute demand. In terms of dollar absolute value, Bernstein expects China's import value to remain stable or slightly increase on the historical high basis.