Written by: Rita
The semiconductor capital equipment industry continues to break through previous supply ceilings, but market interest in further upside is cooling.
On September 7, Morgan Stanley released its SPE industry report, outlining seven key debates. Morgan Stanley believes the market is facing a critical divergence: WFE forecasts continue to be raised, yet investors are increasingly reluctant to pay for additional upside without clearer evidence of cycle sustainability. The firm recommends names priced by the market for peak demand in 2027, including AEIS, MKS, and ONTO. Morgan Stanley maintains an “Equal-Weight” rating on the U.S. SPE industry.
Why SPE Shares Aren't Rising: Two Frameworks for Explanation
Morgan Stanley noted that market hesitation toward SPE shares stems from two interrelated but logically distinct frameworks.
The first is the AI capital return framework. SPE is part of the investment logic for AI infrastructure, where expansion in semiconductor capacity directly maps to gigawatt-scale deployments by hyperscalers. If investors lack confidence in the pace of gigawatt deployments—such as delays in data center construction or stress in hyperscaler bonds or credit default swaps—then stocks related to AI infrastructure become difficult to buy. Morgan Stanley believes the market’s assessment of fundamentals has not changed, but the way these fundamentals are being interpreted has. Since Morgan Stanley is uncomfortable with further upward revisions to WFE, it is applying cyclical trough valuation multiples to EPS for 2027 or 2028.
The second is the fully expected framework. Market expectations for WFE in 2027 and 2028 are nearing saturation, with buy-side expectations heard by Morgan Stanley at approximately $230 billion and over $300 billion, compared to Morgan Stanley’s forecasts of $223 billion and $254 billion. The market believes there is limited upside potential for further revisions, and thus applies valuation multiples consistent with the cyclical trough.
How is WFE converted into computing power?
Morgan Stanley has updated its estimate of WFE demand per GW of computing power. Although the specifications for Rubin Ultra have not been finalized, Morgan Stanley estimates WFE demand at approximately $3.4 billion per GW, or about $7 billion in WFE demand for every $100 billion in AI capital expenditures. This estimate is lower than Lam Research's range of $9 billion to $10 billion, as Morgan Stanley's analysis focuses solely on NVIDIA and does not include Google's TPU or Amazon's Trainium.
Based on this framework, Morgan Stanley’s Internet team forecasts that hyperscaler capital expenditures will reach $1.2 trillion by 2027, corresponding to 29 GW of computing power. This translates to an incremental WFE demand of approximately $87 billion to $125 billion per year starting in 2025. Assuming non-AI end markets do not receive additional capacity, this implies that WFE in 2027 will range from $204 billion to $242 billion, with Morgan Stanley’s forecast of $223 billion positioned in the middle of this range.
The previous bull market scenario has largely been realized.
Several bull market catalysts previously forecasted by Morgan Stanley have largely been realized. Intel guided in its recent earnings report that capital expenditures in 2027 will be significantly higher than in 2026 and raised $20 billion in equity. Kioxia’s K3 announcement and its collaboration with Solidigm to expand production indicate that new NAND capacity projects have gained market recognition. Memory manufacturers are accelerating capacity deployment as early as possible, and Morgan Stanley expects Lam Research, Applied Materials, and Tokyo Electron to record DRAM equipment shipments exceeding $6 billion in the December quarter.
New incremental demand may come from three sources. On the Terafab side, Morgan Stanley expects WFE contributions of approximately $2 billion per year from 2027 to 2028, increasing to $6 billion in 2029. Regarding mature logic, Morgan Stanley anticipates a 2% decline in 2026 followed by an 18% increase in 2027. Morgan Stanley believes the market has underestimated the resilience of the mature node recovery.
DRAM supply and HBM de-specification
DRAM bit supply has increased by 36% year-to-date in 2026, with TrendForce forecasting a full-year 2026 growth of 31%. Morgan Stanley’s model projects bit supply growth of 32% in 2026 and 38% in 2027, with HBM bit growth at 55% and 57%, respectively.
The transition of HBM from 12-layer to 8-layer has drawn market attention. Morgan Stanley noted that moving from 12 layers to 8 layers does not imply a proportional reduction in all process steps. SK Hynix previously stated that a 12-layer stacked chip is 40% thinner and has 13% narrower gaps than an 8-layer chip, significantly increasing control complexity. Morgan Stanley believes that if the HBM roadmap cannot evolve toward 16 layers or more, demand for hybrid bonding will weaken, and the intensity of process control will struggle to further improve.
Intel's market share may face a realignment
Intel has historically been a key customer of Applied Materials and Tokyo Electron, with both companies traditionally holding nearly equal shares of Intel’s capital expenditures. However, Morgan Stanley believes that Intel’s manufacturing strategy has shifted since Pat Gelsinger became CEO and Naga Chandrasekaran took over the foundry business in June 2024, potentially leading to a reallocation of supplier shares.
A clear example is the strength in process control equipment. Morgan Stanley estimates that process control's share of Intel's WFE spending has risen from approximately 6% in 2023 to around 10% in 2026, and KLA and Onto Innovation have already begun to benefit from this trend.
Misalignment between subsystem and OEM valuations
Morgan Stanley noted that the inventory dynamics in the current cycle differ from previous ones. Applied Materials and Lam Research saw their absolute inventory rise by 8% over the past two quarters, but their days of inventory on hand decreased by 19 days, as both companies depleted inventory to fulfill customers' urgent orders. Days of inventory on hand have fallen to the lowest level since the December 2021 quarter.
Morgan Stanley believes there is a valuation misalignment between AEIS and MKS. Subsystem suppliers are expected to outperform compared to previous cycles, as OEMs have no accumulated subsystem inventory to draw down. When the cycle slows, Applied Materials and Lam Research will not face the same negative impact from subsystem inventory as they did in the prior cycle.
However, AEIS and MKS are currently valued as if they were "early-cycle" stocks, while OEM is priced based on 2028 earnings. Morgan Stanley believes this misalignment represents a mispricing, as the inventory digestion factors that hampered subsystem suppliers in the previous cycle are unlikely to recur to the same extent in this cycle.
Disclaimer
This article is a compilation and interpretation by Chaoxiang Research of a third-party brokerage research report (Morgan Stanley, September 7, 2026), combined with publicly available market information. The ratings, price targets, earnings forecasts, and related judgments cited herein reflect the views of the brokerage's analysts and represent only the position of their respective institution; they do not reflect the views of Chaoxiang Research nor constitute any investment advice.
The market carries risks; make decisions independently. This article should not be used as a basis for buying or selling any securities.