Bernstein Says Semiconductor Equipment Does Not Have to Fall with Memory Stocks

Overview

This report addresses whether WFE and memory sectors must move together after recent increased volatility in memory stocks and market concerns that semiconductor equipment stocks might be dragged down in tandem. Bernstein's core conclusion is that the two share a common semiconductor beta, but historical correlation is not enough to determine relative returns, and WFE has outperformed memory in multiple cycles when memory was weak.

Core views

The report argues that the investment question should not be simplified to whether memory and WFE are correlated; instead, it should assess whether current memory volatility will broadly weaken WFE demand. The authors view the current situation as an unusually strong memory cycle driven by HBM and traditional DRAM shortages, and if the memory premium mean reverts, it can coexist with strong WFE fundamentals. Over the long run, Bernstein's WFE bullish case is based on memory capex expansion, higher memory content in AI servers, leading-edge logic/foundry investment, advanced packaging, and technology transition demand.

Analysis framework

The report uses 2011-2026 rolling daily return correlations, monthly returns, cycle scenario tables, and relative performance analysis to compare Memory, WFE, and SOX, combining statistical correlation with real semiconductor supply-demand cycles to separate common beta from sub-industry drivers.

Methodology notes

• Historical correlation analysis12-month, 3-month, and 24-month rolling correlations

  Comparing daily return correlations of Memory, WFE, and SOX

  By observing short-term and long-term rolling correlations, the report shows that WFE has a more stable relationship with SOX, while memory is more sub-industry specific.

• Cycle scenario analysisSemiconductor cycle grouping

  Grouping by memory up/down cycles, COVID chip shortages, HBM, and AI capex context

  The report does not rely only on statistical breakpoints; it places different periods back into real supply-demand contexts to identify business reasons behind relative outperformance differences between WFE and memory.

• Mean reversion analysisMemory premium relative to WFE

  Observing whether memory’s long-run excess performance versus WFE has become excessively stretched

  The report states that memory’s recent premium versus WFE has risen far above its historical average, and if mean reversion occurs, WFE could benefit.

Asset mapping & comparison

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

• WFE / semiconductor equipment

  Primary bullish theme of the report

  Strengths

  Supported by memory capacity expansion, AI capex, leading-edge logic/foundry investment, advanced packaging, and technology transition demand; highly correlated with SOX and has historically outperformed when memory was weak.

  Weaknesses

  Still exposed to semiconductor cycles, capex pace, and valuation volatility.

  Comparison

  Relative to Memory, WFE has recently lagged; if the memory premium versus WFE mean reverts, WFE could benefit.

  Risks

  If memory volatility spreads to broad semiconductor demand weakness, WFE demand and upward earnings revisions could come under pressure.

• Memory

  Related to WFE, but not the sole determinant of WFE performance

  Strengths

  HBM and traditional DRAM tightness, rising memory content in AI servers, and potential near-term HBM price hikes could drive upward earnings revisions.

  Weaknesses

  Its recent excess performance versus WFE is stretched, and memory prices may normalize in 2027.

  Comparison

  Memory has significantly outperformed WFE recently, but history shows the two can diverge materially.

  Risks

  If memory price normalization occurs faster than expected, memory stock valuation and earnings expectations may decline.

• Samsung Electronics / SK hynix / Micron

  Bernstein remains constructive on selected memory names

  Strengths

  Supported by shortages, rising HBM prices, and potential upward earnings revisions; still favored even after recent pullbacks.

  Weaknesses

  Affected by memory price cycles and capex discipline.

  Comparison

  Compared with KIOXIA, the report is noticeably more constructive on these major memory names.

  Risks

  If LTA support is limited, prices normalize, or supply expansion comes in faster than expected, earnings upgrades could slow.

• KIOXIA

  Negative-rated name in the report

  Strengths

  Located within the memory chain and may benefit from industry cyclicality.

  Weaknesses

  Bernstein assigns Underperform due to valuation and long-term threats from China.

  Comparison

  The report is significantly more cautious on KIOXIA than on Samsung, SK hynix, and Micron.

  Risks

  Valuation pressure and long-term competitive threats could limit returns.

Key data

• 2012-2018 correlation between Memory and SPEabout 0.4The report says historical correlation is not high.
• 2019 onward correlation between Memory and SPEabout 0.6Correlation has risen but remains below WFE versus SOX.
• SPE and SOX correlationabout 0.8-0.9The report says the two have been highly correlated over the long term.
• WFE relative performance versus memory from Jan 2015 to Dec 2016WFE +21.9%, Memory -16.2%, relative outperformance around 38.2 percentage pointsShows the two can move in opposite directions.
• WFE relative performance versus memory from Jan 2021 to Dec 2022WFE +15.3%, Memory -34%, relative outperformance around 49 percentage pointsShows WFE can significantly outperform during shortage and capex cycles.
• SK hynix incremental investmentKRW 100tn, about 67bn USDFor the new Cheongju wafer fab, supporting the view of accelerated memory capex.
• Recent status of Memory versus WFE relative premiumabove +2σThe report says the memory premium relative to WFE is significantly above its historical average.

Impact & implications

The implication for investors is that semiconductor equipment should not be mechanically avoided solely because of memory stock volatility. If AI capex, advanced process investment, advanced packaging, and memory capex expansion continue to support WFE demand, WFE can remain relatively resilient even if memory prices normalize or memory stocks pull back. At the same time, the memory segment may still be supported by shortages and HBM price increases, which can drive upward earnings revisions.

Risks

• If memory prices normalize quickly in 2027 or earlier, memory stock earnings expectations and sentiment could be affected.
• If memory volatility spreads into broad industry demand weakness rather than a sub-industry-only mean reversion, WFE fundamentals may be pressured.
• WFE-related companies trade at relatively high valuations, and if expected earnings upgrades do not materialize, valuation compression is possible.
• Chinese substitution and long-term competitive risk may affect some equipment or memory companies, with the report specifically noting KIOXIA faces long-term pressure.
• If AI capex, HBM demand, advanced packaging, and logic/foundry investment slow, the WFE constructive case would weaken.

What to watch

• Whether memory premium versus WFE rolls down from extreme historical levels.
• HBM price increases, traditional DRAM supply-demand tightness, and the magnitude of memory earnings revisions.
• Execution of capacity expansion plans by Korean players such as SK hynix and Samsung and implementation of supporting Korean government measures.
• Whether WFE market and major equipment-company EPS consensus expectations for 2028 continue to be revised upward.
• Whether AI server memory content increases, advanced packaging investment, and leading-edge logic/foundry capex remain sustained.
• Whether memory stock volatility remains confined to the sub-industry or evolves into broader semiconductor demand weakness.