Memory stocks do not necessarily determine the fate of semiconductor equipment stocks

Overview

This report addresses investors' core question about semiconductor equipment stocks: if memory stocks stop rising, can semiconductor equipment stocks still continue to perform? Bernstein believes the answer is yes. Through stock-price correlation, rolling return, and cycle-stage analysis from 2011 to 2026, the report argues that memory and WFE share broad semiconductor market factors, but their relative returns are more determined by the drivers of their respective sub-industries.

Core views

The core views are: first, the historical correlation between memory and WFE is not high and has been quite volatile, so one cannot simply assume they rise and fall together; second, WFE has a more stable correlation with the broader SOX semiconductor index, indicating equipment stocks are more closely tied to the broader semiconductor capex cycle; third, history has seen multiple periods when WFE outperformed even as memory weakened or correlations were high; fourth, memory's premium relative to WFE has now stretched significantly, and if memory prices normalize in 2027, WFE may not necessarily fall with them; fifth, AI capital expenditures, HBM, advanced packaging, logic/foundry investment, and memory capacity expansion continue to support equipment demand.

Analysis framework

The report combines historical market data with fundamental catalysts: it first compares 12-month rolling correlations among memory, WFE, and SOX, then uses 3-month and 24-month windows to observe changes in short- and long-term correlations, followed by using a cycle framework to break down relative performance across different stages such as memory up/down cycles, the COVID chip shortage, HBM, and AI capex, and finally forms its investment conclusion by combining capital expenditure announcements, earnings revisions, and valuation judgments.

Methodology notes

• Market correlationRolling correlation analysis

  Use 12-month, 3-month, and 24-month rolling correlations of daily returns from 2011-2026 to compare the degree of co-movement among Memory, WFE, and SOX.

  This method is used to distinguish common semiconductor factor exposure from sub-industry-specific drivers. The report argues that correlation can explain only part of shared rises and falls and cannot directly infer cumulative return differences.

• Cycle analysisSemiconductor cycle-stage framework

  Divide stages based on real supply-demand backgrounds such as memory up/down cycles, the COVID chip shortage, HBM, and AI capital expenditures.

  This framework is used to test whether WFE can perform independently of memory under different supply-demand environments. The report finds at least two stages when the two moved in completely different directions.

• Relative performanceMean reversion perspective

  Observe memory's long-term cumulative performance relative to WFE and the degree of current premium deviation.

  The report points out that memory has significantly outperformed WFE since June 2025, and the recent premium has reached a historically rare level; if mean reversion occurs, it could be more favorable for WFE.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• WFE/semiconductor equipment stocks

  The report's most important bullish direction, representing the semiconductor capital expenditure chain.

  Strengths

  Driven by AI capital expenditures, memory capacity expansion, advanced packaging, logic/foundry investment, and technology migration; historically able to outperform when memory weakens.

  Weaknesses

  Still exposed to the semiconductor cycle, capex downgrades, and valuation volatility.

  Comparison

  Compared with memory stocks, WFE has a more stable correlation with SOX, but relative returns are still determined by equipment demand and the capex cycle.

  Risks

  If memory volatility expands into a broader industry demand downgrade, WFE fundamentals may also be damaged.

• Memory stocks

  A sub-industry exposure related to WFE but not fully synchronized with it.

  Strengths

  HBM shortages, upcoming HBM price increases, and tight conventional DRAM supply may still bring earnings upgrades.

  Weaknesses

  The current premium relative to WFE is extremely high, and memory price normalization in 2027 may bring volatility.

  Comparison

  Memory has recently significantly outperformed WFE, but the report believes part of this may reflect a reversal after long-term underperformance and an extreme cyclical premium.

  Risks

  Price normalization, supply release, long-term competitive threats from China, and valuation pressure.

• Samsung Electronics, SK hynix, Micron

  The main memory companies the report continues to favor.

  Strengths

  HBM and memory shortages bring the potential for earnings upgrades, and the report remains positive after the recent pullback.

  Weaknesses

  Highly affected by the memory price cycle and supply-demand changes.

  Comparison

  The report is more positive on them than on KIOXIA.

  Risks

  Memory price normalization, changes in the capex pace, and industry competition.

• KIOXIA

  The negatively rated name in the report.

  Strengths

  Has exposure to the NAND memory industry.

  Weaknesses

  Valuation pressure and long-term threats from China.

  Comparison

  Compared with Samsung, SK hynix, and Micron, the report is more cautious on KIOXIA.

  Risks

  Valuation pullback, the NAND cycle, and intensifying long-term competition.

Key data

• 2012-2018 memory vs. SPE/WFE correlationAbout 0.4The report says historical correlation was not high.
• Memory vs. SPE/WFE correlation since 2019About 0.6Higher than in the earlier period, but still below WFE's correlation with SOX.
• SPE/WFE vs. SOX correlationAbout 0.8-0.9Shows equipment stocks have a more stable linkage with the overall semiconductor index.
• January 2015 to December 2016WFE rose 21.9%, memory fell 16.2%WFE outperformed by about 38.2 percentage points over the two years.
• January 2021 to December 2022WFE rose 15.3%, memory fell 34.0%WFE outperformed by about 49 percentage points over the two years.
• Additional SK hynix investmentKRW 100tn, about USD 67bnFor the new Cheongju fab, reflecting accelerating memory capital expenditures.
• Memory premium relative to WFEAbout +2σ above the historical meanThe report believes the current memory premium is at a historically rare stretched level.
• Coverage rating sampleTEL, Advantest, Disco, Lasertec, ASML, Besi, AMAT, LRCX, KLAC, etc. are rated Outperform; Screen is rated Market-Perform; KIOXIA is rated UnderperformPlease refer to the report tables for the specific target prices.

Impact & implications

The investment implication is that investors should not avoid semiconductor equipment stocks merely because they worry about a correction in memory stocks. If memory volatility mainly comes from price normalization rather than a downgrade in broad semiconductor capital expenditures, WFE may still benefit from AI servers, HBM expansion, advanced packaging, logic/foundry investment, and technology migration. At the portfolio level, the report suggests viewing WFE and memory as related but not fully homogeneous exposures, with the two able to provide some diversification in return sources.

Risks

• Memory price normalization in 2027 exceeds expectations and spreads into a downgrade in broad semiconductor demand.
• AI server, HBM, or advanced packaging capital expenditures come in below expectations.
• A slowdown in logic/foundry investment weakens the momentum for WFE orders and earnings upgrades.
• Equipment stock valuations are already relatively high; if earnings upgrades fail to materialize, valuation compression may occur.
• Competition and substitution risks from China create long-term pressure on some memory and equipment-chain companies.
• The report covers multiple companies, and individual target prices and ratings need to be judged separately in combination with each company's own disclosures, valuation methodology, and risk factors.

What to watch

• Whether memory price normalization in 2027 affects only memory stocks or also spreads to WFE demand.
• The implementation pace of SK hynix's Cheongju new fab investment and Korean government support for Samsung and SK hynix capacity expansion.
• Whether HBM price increases, tight HBM supply-demand conditions, and rising memory content in AI servers continue.
• Whether consensus expectations for the WFE market and equipment company EPS through 2028 continue to be revised upward.
• Whether Memory-WFE relative performance mean reverts from the current extreme premium.
• Orders, capex guidance, and advanced-packaging-related commentary from equipment companies such as AMAT, LRCX, KLAC, ASML, and TEL.