European semiconductors Report Interpretation

Overview

Morgan Stanley’s European semiconductor review argues that AI demand and a broadening cyclical recovery remain supportive, but that investors should be more selective as memory momentum matures, risks to data-center investment rise and valuation gaps widen. The report reshapes relative preferences across semiconductor equipment, analog/optics and EDA/IP.

Core views

The report remains constructive on semiconductor fundamentals. Stronger-than-expected 1H26 SIA data point to a broader recovery, firmer pricing, improving demand, inventory replenishment and constructive book-to-bill trends. AI data-center investment and LLM adoption have supported sentiment, with the SOX outperforming the S&P 500 by more than 50% year-to-date. Morgan Stanley’s European coverage is up 88% year-to-date, although the sector has been volatile: its stocks had de-rated from June peaks and SOXX daily volatility widened to a 4.1% standard deviation from 2.5% in January through May. The firm views the sell-off and lower valuations in selected EDA and semi-cap names as creating opportunities, but stresses that the sector remains well owned. The principal cyclical concern is memory. DRAM prices are up more than 600% year-on-year and could reach roughly 700% year-on-year in 3Q26. Because the cycle peak is assessed through the year-on-year rate of price change, Morgan Stanley believes a further acceleration would require roughly a 70% sequential price increase in 4Q. It therefore expects memory to move into a late-cycle phase by 4Q26. This could weigh on sentiment toward memory-exposed equipment names, although structural AI demand, including Agentic AI, may cushion the effect. The firm also sees 2027 WFE strength as broadly understood: it forecasts $163bn in 2026 and $223bn in 2027, up 37% year-on-year, supported by DRAM capacity, logic spending and NAND greenfield investment. Its $254bn 2028 WFE forecast implies only 14% growth, and it does not view WFE exceeding $300bn in 2028 as likely; the market debate is therefore shifting to 2028 capex and supply-demand conditions. Morgan Stanley identifies four sector risks into year-end. First, data-center grid build-out may fall short of elevated expectations because of grid-connection delays and political opposition; it cites waits of five to seven years in some regions and more than 300 US AI-data-center moratoriums since 2023. Second, higher long-end Treasury yields and renewed oil-price concerns could pressure risk appetite and intensify scrutiny of NVIDIA’s ecosystem investments. Third, falling token prices and efficiency gains must prove sufficient to support attractive LLM returns on invested capital as investor focus shifts to monthly ARR and profitability. Fourth, a delay to 800V-native rack architecture could defer demand for components whose content rises with the transition from 400V to 800V, especially GaN. Within semiconductor equipment, ASML remains Morgan Stanley’s Top Pick and Overweight. The firm continues to see DRAM and logic demand, product-led pricing, immersion demand and capacity optionality supporting medium-term margin expansion. However, it lowers the price target to €1,700 from €1,930 despite unchanged estimates, applying 30x FY28 EPS of €56.42 rather than 35x to reflect near-term sentiment overhangs. Those overhangs include China and export-control uncertainty, EUV capacity, China system sales and possible High-NA margin dilution. Morgan Stanley considers EUV-capacity and export-control risks overstated, noting management’s guidance for at least 60 low-NA systems in 2026, 85 in 2027 and 110 in 2028 may represent a floor rather than a ceiling. It expects the Brainport Industries Campus, planned to begin construction in 3Q26 and receive first employees in 1Q28, to add flexibility. The report estimates China at 18% of ASML sales in both FY26 and FY27 and sees no risk to FY26 estimates from a possible MATCH Act implementation, though enactment could affect FY27 and beyond. Besi remains Overweight but its price target falls to €220 from €260. Morgan Stanley concludes that hybrid bonding for HBM is likely to arrive later and at a lower level than previously expected because broader memory architectures can use combinations of SRAM, serialised HBM, peer HBM, LPDDR and CXL instead of relying solely on taller HBM stacks. It cuts expected hybrid-bonding tool shipments for 2026-28 to 65, 115 and 190, from 65, 140 and 230, and forecasts earnings power of €4.33, €7.09 and €7.93, changes of +1%, -4% and -6%. A stronger mainstream bonder cycle for 2.5D applications partly offsets the lower hybrid-bonding outlook. The revised target uses 31x FY27 earnings versus 35x previously, while the firm retains Overweight because it believes much of the risk is priced in and sees optionality from mainstream demand and adoption outside HBM. In analog and optics, Morgan Stanley expects margins to improve but cautions that the pace may disappoint elevated expectations. It highlights IDM-specific start-up and utilisation costs at Infineon and STMicroelectronics, including Dresden and Crolles fab ramp costs, as reasons for a more gradual margin ramp. Vertical power delivery remains strategically important, but competition is rising as ADI and Navitas enter the field. Morgan Stanley still sees Infineon as a major data-center power beneficiary, particularly after its C2i Semiconductor acquisition, but moves it to Equal-weight because the opportunity is already understood and near-term catalysts are limited. It maintains FY27 data-center sales expectations of €2.8bn, below the €4bn-plus level some investors expect, and sits 18% and 24% below consensus for PSS sales in FY27 and FY28. Its €65 target, down from €81, applies a 30x multiple to the server-rack business, reduced from 45x, and 20x to the rest of the group. The report remains positive on optical connectivity and Soitec. Multi-rack AI systems require direct optical links between racks where copper becomes less practical, supporting NPO/CPO configurations enabled by Soitec silicon-photonics wafers. Morgan Stanley estimates a multi-rack topology could triple Soitec revenue per GPU versus a standard 72-GPU rack with no direct cross-rack connections; even a gradual shift to 30% adoption could increase revenue per GPU by around 50%. It is ahead of the Street on Soitec revenue by 1%, 5% and 16% for FY27-29 and on gross profit by 10% and 36% in FY28 and FY29, respectively, due largely to a more optimistic photonics mix. Soitec remains a Top Pick and Overweight with a €200 target. In EDA/IP, Morgan Stanley judges LLM disruption concerns to be overdone. It notes that claims of an LLM using open-source EDA to design a 45nm chiplet differ materially from high-volume commercial production of leading-edge 2/3nm AI chips. The firm expects proof of earnings resilience rather than a rapid sentiment recovery, but sees physical AI as an underappreciated opportunity for Synopsys and Cadence because simulation tools address thermal, mechanical-stress and fluid-dynamics challenges in sim-to-real workflows. Synopsys is upgraded from Equal-weight to Overweight with an unchanged $500 target. Morgan Stanley cites improving confidence in Ansys integration synergies, a Design IP recovery and physical-AI potential. It values Synopsys at 33x FY26 EPS of $15.12, compared with a current 28x multiple and Cadence at 37x. Its September 30 investor day watch points are Factory 2 IP, Multiphysics Fusion and management’s response to LLM-risk concerns. The report also raises its 2030 CPU total-addressable-market estimate to $125bn from Agentic AI, with a $283bn bull case. It sees Arm’s key issue as conversion of demand into supply-backed revenue rather than end demand itself; Arm expects more than $2bn of AGI CPU demand across FY27-28, while secured manufacturing capacity supports at least $1bn of cumulative revenue through FY28. Morgan Stanley retains Equal-weight and a $212 target for Arm because it believes the valuation already discounts much of the longer-term CPU opportunity.

Analysis framework

Morgan Stanley combines semiconductor-cycle indicators, SIA data, DRAM pricing, WFE forecasts, investor positioning and valuation comparisons with company-specific earnings, capacity, technology and competitive analysis. It then adjusts ratings and price targets using earnings estimates, P/E multiples and, where relevant, sum-of-the-parts valuation.

Methodology notes

• Industry AnalysisSupply-demand framework

  Semiconductor-cycle analysis using sales data, inventories, pricing, memory supply-demand and WFE spending.

  The report links pricing, inventory replenishment, DRAM-cycle timing and capital spending to expected demand for semiconductor equipment and components.

• Valuation methodsP/E and PEG Valuation

  Forward P/E valuation.

  Morgan Stanley derives several price targets by applying stated P/E multiples to forecast EPS, including ASML at 30x FY28 EPS and Synopsys at 33x FY26 EPS.

• Valuation methodsSOTP (Sum-of-the-Parts) Valuation

  Sum-of-the-parts valuation.

  The report values Infineon’s server-rack business and remaining operations at different multiples, and similarly separates structural and cyclical businesses for selected names.

• Industry AnalysisUpstream-Midstream-Downstream Transmission

  AI data-center architecture and semiconductor supply-chain transmission.

  The analysis traces how rack scaling, optical links, power architecture, memory configuration and fab capital expenditure affect suppliers across the semiconductor value chain.

Asset mapping & comparison

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

• ASML Holding NV (ASML.AS)

  Top Pick semiconductor-equipment beneficiary of DRAM and logic demand

  Strengths

  EUV leadership, pricing power, capacity optionality and medium-term margin expansion potential.

  Weaknesses

  Near-term sentiment is constrained by China, export controls, capacity questions and High-NA margin concerns.

  Comparison

  Morgan Stanley views ASML as its preferred semi-cap name alongside ASM and Besi.

  Risks

  Export restrictions, China sales pressure, weaker WFE demand, slow EUV order growth and delayed High-NA adoption.

• BE Semiconductor Industries NV (BESI.AS)

  Covered semiconductor-packaging supplier

  Strengths

  Improving core bonder demand and optionality from hybrid bonding outside HBM.

  Weaknesses

  HBM hybrid-bonding demand is expected to materialize later and at lower levels.

  Comparison

  Remains preferred within semi-cap despite lower estimates and target price.

  Risks

  Limited hybrid-bonding adoption, fading core-business strength and FX-related margin pressure.

• Infineon Technologies AG (IFXGn.DE)

  Covered analog and data-center power supplier

  Strengths

  Morgan Stanley sees Infineon as a clear data-center power beneficiary with broad rack-level power-semiconductor exposure.

  Weaknesses

  Limited near-term upside to expectations and lower confidence in a near-term catalyst.

  Comparison

  The report sees stronger structural data-center positioning than European analog peers but a more balanced risk-reward profile.

  Risks

  800V delays, VPD competition, weaker automotive recovery and delayed Module 4 manufacturing efficiencies.

• Synopsys Inc. (SNPS.O)

  Upgraded EDA/IP coverage name

  Strengths

  Ansys synergies, Design IP recovery, simulation capabilities and physical-AI exposure.

  Weaknesses

  LLM-related EDA concerns and the need for proof of earnings resilience remain sentiment overhangs.

  Comparison

  Morgan Stanley views the valuation gap versus Cadence as too wide.

  Risks

  A deeper semiconductor downturn, export controls, new EDA entrants and failure to deliver Ansys cost synergies.

• Soitec SA (SOIT.PA)

  Top Pick optical-connectivity beneficiary

  Strengths

  Leadership in photonics SOI substrates and leverage to NPO/CPO optical connectivity in multi-rack AI systems.

  Weaknesses

  Broader adoption timing for advanced rack topologies remains uncertain.

  Comparison

  Morgan Stanley is above the Street on photonics-led revenue and gross-profit growth.

  Risks

  Slower optical-connectivity adoption and delayed data-center architecture transitions.

Key data

• Coverage performancec.70% YTD; c.60% retreat from June peaksThe report describes strong year-to-date gains followed by a sharp summer pullback.
• DRAM pricing>600% YoY; c.700% YoY in 3Q26Morgan Stanley expects the memory cycle to become late-cycle by 4Q26.
• WFE forecast$163bn in 2026; $223bn in 2027; $254bn in 2028Implies 37% year-on-year growth in 2027 and 14% in 2028.
• ASML target price€1,700Reduced from €1,930 using 30x FY28 EPS of €56.42.
• Besi hybrid-bonding tool shipments65/115/190 in 2026e/2027e/2028eReduced from 65/140/230 as HBM hybrid-bonding adoption is expected later and lower.
• Infineon data-center sales forecast€2.8bn in FY27Morgan Stanley does not expect guidance to rise to the €4bn-plus level anticipated by some investors.
• Synopsys valuation28x current-year P/E versus Cadence at 37x; $500 targetMorgan Stanley applies 33x to FY26 EPS of $15.12.
• Agentic AI CPU TAM$125bn by 2030; $283bn bull caseThe report sees Agentic AI expanding CPU demand.

Impact & implications

Morgan Stanley’s sector view favors selective exposure to AI-linked semiconductor equipment, optical connectivity and EDA/IP rather than broad cyclical positioning. The firm sees improving fundamentals and lower relative valuations as supportive, but expects 2028 WFE growth, the late-cycle DRAM transition, data-center build-out, LLM profitability and 800V deployment timing to determine whether enthusiasm persists.

Risks

• Data-center grid build-out could slow because of connection delays and political opposition.
• Higher long-end yields and oil prices could reduce semiconductor investor enthusiasm.
• Falling token prices and efficiency gains may not be enough to support attractive LLM returns on invested capital.
• Delays to 800V-native architecture could defer demand for GaN and other affected power-semiconductor content.
• The DRAM cycle may turn late-cycle by 4Q26, weighing on memory-exposed equipment sentiment.

What to watch

• Evidence on DRAM pricing, inventory replenishment and the expected late-cycle transition in 4Q26.
• 2028 WFE outlooks from memory makers, logic suppliers and hyperscalers.
• ASML’s EUV capacity progress, Brainport Industries Campus build-out, China sales and export-control developments.
• The timeline for 800V rack deployment, including commentary at the AI Infra Summit and OCP Global Summit.
• Synopsys’ September 30 investor day, especially Factory 2 IP, Multiphysics Fusion and Ansys synergy updates.
• Soitec photonics demand visibility and NPO/CPO adoption timelines.
• Besi hybrid-bonding orders and adoption in HBM, China and wearables.