Bernstein: Memory Correction Unlikely to Drag Down Semiconductor Equipment Sector
Taylor Wilson
Bernstein argued on July 13 that historical data contradicts the market assumption that a memory downturn must hurt equipment stocks, and that equipment offers a better risk-reward profile today — the key question is whether this memory correction stays cyclical or turns systemic.
Does a memory slump always drag equipment stocks down?
Market instinct says yes, but Bernstein's data says otherwise: from 2012 to 2018, the stock-price correlation between memory and wafer fabrication equipment (WFE) was only about 0.4; it rose to roughly 0.6 after 2019 — still well below 1.
By contrast, WFE's correlation with the Philadelphia Semiconductor Index (SOX) has stayed at 0.8–0.9 over the long term. This means → equipment stocks track the broader chip-industry cycle, not the memory sub-cycle alone.
In plain terms = memory is just one branch of semiconductors; equipment makers fill orders from the whole industry, not from memory alone.
Have equipment stocks actually held up during past memory downturns?
Bernstein reviewed seven semiconductor cycles since 2012 and found equipment stocks posted positive returns during multiple memory downturns.
Two clear examples: in 2015–2016, memory logged negative returns while equipment delivered double-digit gains; in 2021–2022, memory losses deepened yet equipment still finished in positive territory.
This reflects a deeper point — correlation alone cannot predict relative returns. What actually drives performance is each sub-sector's own fundamentals, not short-term price co-movement.
What makes this AI cycle different?
Driven by high-bandwidth memory — HBM, a type of high-speed memory designed for AI chips — and tight conventional DRAM supply, memory has vastly outperformed equipment over the past year-plus, pushing the cumulative return gap to a record high.
This means → memory's valuation premium over equipment is now at a historically elevated level — the market's optimism on memory is already "priced in."
Before this rally began, memory had actually trailed equipment for years, only catching up in cumulative gains earlier this year before pulling sharply ahead.
If memory prices normalize, will equipment demand weaken too?
Bernstein is explicit: memory prices returning to normal does not mean equipment demand deteriorates in tandem.
Four pillars still support equipment spending: AI infrastructure buildout + advanced logic nodes + advanced packaging + ongoing technology upgrades.
In plain terms = chipmakers buy equipment for far more than memory production — AI and leading-edge logic are both queuing up to place orders.
What is Bernstein's bottom line?
Bernstein expects the global WFE market to keep growing over the next several years, with room for industry earnings forecasts to be revised upward.
Compared with memory, where optimistic expectations are already fully reflected, equipment stocks offer a more attractive risk-reward profile.
The key validation point: whether this memory correction is merely an intra-industry cyclical reset or escalates into a systemic risk that cuts wafer-fab capital expenditure — the former favors equipment stocks; only the latter poses a real threat.
Content is for reference only, not financial advice.