ASML roadshow signals long-term growth, while the memory cycle enters a more cautious phase

Overview

This is a Morgan Stanley webcast and management-roadshow document covering ASML’s management roadshow, the memory cycle and innovation, Apple AI-related catalysts, and topics including Chinese AI models and the CXMT IPO. It is not a single-company in-depth valuation report, but rather connects AI infrastructure spending, memory supply and demand cycles, long-term semiconductor-equipment growth, and regional competitive dynamics through an industry framework and meeting takeaways.

Core views

The core view is that ASML’s long-term growth logic remains clear, supported by EUV shipments, value pricing, end-market expansion, and mid-cycle margin leverage. In memory, AI demand may extend the cycle but does not eliminate cyclicality; DRAM year-over-year price growth, inventory, and earnings-revision momentum suggest that the cycle may be approaching a high point. The development of Chinese AI and CXMT enhances domestic technological competitiveness, but HBM remains a structural bottleneck.

Analysis framework

The report uses a thematic webcast framework: it first defines the debate around AI capital expenditure, long-term agreements, and the memory-cycle position from the perspective of investors’ key concerns; it then assesses the cycle stage using DRAM/NAND inventory, pricing, earnings revisions, and valuation metrics; finally, it supplements the industry-chain implications with the CXMT IPO, DDR5 supply, AI model performance, and ASML management-roadshow takeaways.

Methodology notes

• Industry Cycle AnalysisThree-Debate Framework for the Memory Cycle

  AI spending, LTA re-rating, cycle inflection point

  The report assesses the memory industry along three themes: whether AI capital expenditure will continue to support demand, whether long-term agreements will change the market’s approach to valuing earnings, and whether inventory, pricing, and earnings revisions indicate that the cycle has entered a high-stage phase.

• Company Roadshow TakeawaysASML NDR Takeaways

  Long-term growth drivers

  The ASML section distills long-term growth drivers from management-roadshow information, including value pricing, order and demand visibility, the EUV shipment outlook, margin leverage, and end-market expansion.

• Valuation and Earnings RevisionsP/B and EPS Revision Monitor

  Price peak and valuation re-rating

  The report uses year-over-year changes in DRAM prices, NTM P/B, EPS changes, and the net earnings-revision ratio to assess whether memory stocks have already priced in cyclical improvement and whether revision momentum is beginning to slow.

Asset mapping & comparison

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

• ASML HOLDING NV

  Core covered company and subject of the management roadshow

  Strengths

  High EUV technology barriers, clear long-term shipment expectations, and value pricing and demand expansion supporting mid-cycle margins.

  Weaknesses

  In the near term, it remains exposed to the pace of semiconductor capital expenditure, customer-order fluctuations, and valuation expectations.

  Comparison

  Compared with ordinary semiconductor-equipment companies, ASML has greater scarcity and supply-chain pricing power in EUV.

  Risks

  Delays in advanced-process investment, export controls, customer concentration, and a pullback in elevated valuations.

• Samsung Electronics

  Subject of memory-cycle monitoring

  Strengths

  Leading DRAM/NAND scale and exposure to AI-server and HBM-related demand.

  Weaknesses

  P/B remains above its long-term average, while a decline from elevated year-over-year pricing could pressure valuation.

  Comparison

  Compared with SK Hynix, its valuation multiple is lower, but the market’s focus on HBM execution differs.

  Risks

  A peak in memory prices, rising inventories, and slowing earnings-revision momentum.

• SK Hynix

  Representative asset for the memory cycle and AI/HBM

  Strengths

  Strong AI-memory demand and HBM exposure, with notable earnings leverage during the cyclical upturn.

  Weaknesses

  One-year forward EPS has recently declined, while P/B of approximately 2.5x remains above its long-term average.

  Comparison

  Compared with Samsung, the market assigns it a higher P/B, reflecting greater AI-memory leverage but also increasing expectation risk.

  Risks

  HBM demand below expectations, a price peak, and weakening earnings revisions.

• CXMT

  Chinese DRAM supply and IPO-event asset

  Strengths

  China’s only large-scale DRAM manufacturer, with substantial planned fundraising that could fill part of the DDR5 supply gap.

  Weaknesses

  Its process node is D1z, it has no EUV, and revenue is highly concentrated in DDR/LPDDR.

  Comparison

  Compared with the Big Three, CXMT is expanding in commodity DDR5, but it still has a structural gap in HBM capabilities.

  Risks

  Technology-generation gaps, the HBM bottleneck, capital-expenditure efficiency, and post-listing valuation volatility.

• Apple Inc.

  Potential catalyst within the Apple AI webcast theme

  Strengths

  Apple AI’s entry into China is listed as one of the key catalysts for AAPL.

  Weaknesses

  The material does not provide specific financial quantification.

  Comparison

  Compared with pure-play AI infrastructure companies, Apple represents a more end-device-oriented AI ecosystem catalyst.

  Risks

  Uncertainty around progress in China, regulatory approval, user adoption, and model experience.

Key data

• ASML EUV shipment expectationsApproximately 92 units in FY27 and 104 units in FY28The report title lists ASML shipment expectations to support its assessment of long-term growth drivers.
• DRAM price-cycle viewMay peak around 4Q26DRAM contract-price year-over-year growth has declined from its cyclical high, while valuations have not yet undergone a significant re-rating.
• Earnings-revision momentumNet upward-revision ratio declined from 92% to 77%The ratio remains positive but indicates slowing momentum in the upward-revision cycle.
• Samsung and SK Hynix valuationsSamsung at approximately 1.7x P/B and SK Hynix at approximately 2.5x P/BThe multiples have declined from recent highs but remain above long-term averages.
• Kimi K3 model scale2.8T total parameters, with 16/896 experts activated per MoE invocationThe report says the model approaches frontier models in capability and has a 1M-token context window.
• Kimi K3 pricingApproximately $3/$15 per million tokens (input/output)The report emphasizes its clear pricing advantage relative to Fable 5 and GPT-5.6 Sol.
• Kimi K3 risk alertApproximately 51% independent hallucination rateThe report explicitly notes that a validation process remains necessary.
• CXMT IPO proceedsApproximately ¥57.9bn, or approximately $8.5bn; up to approximately $9.8bnIt is described as the largest IPO in the history of Shanghai’s STAR Market and Asia’s largest IPO in 2026.
• CXMT implied valuationApproximately ¥579bn, or approximately $85bnThe report describes CXMT as China’s only large-scale DRAM manufacturer.
• ASML disclosed rating and priceOverweight; €1,520.80The disclosure table shows an O rating for ASML Holding NV and a price as of 07/20/2026.

Impact & implications

For investors, the implication is that ASML remains a long-term core beneficiary of the AI and advanced-process capital-expenditure chain, but in the near term it is important to distinguish the long-term certainty of semiconductor equipment from the risk of a peak in the memory-price cycle. Memory companies’ earnings are still being revised upward, but weakening momentum and rising inventories mean that further valuation expansion requires stronger evidence of demand or pricing. Chinese AI models and the CXMT IPO demonstrate progress in domestic substitution and the open-source AI ecosystem, while constraints related to HBM, EUV, and advanced processes remain key limitations.

Risks

• Memory prices may peak around 4Q26; if year-over-year pricing continues to decline, related earnings and valuations could come under pressure.
• DRAM and NAND inventories rose in 2Q26, driven mainly by module makers, potentially indicating that the quality of channel restocking requires monitoring.
• The net earnings-revision ratio declined from 92% to 77%. Although it remains positive, this indicates slowing momentum in the upward-revision cycle.
• Although open models such as Kimi K3 enhance competitiveness, the report flags an approximately 51% hallucination rate, meaning that actual commercialization still requires validation.
• ASML’s long-term thesis depends on advanced-process capital expenditure, EUV demand, and the export environment; any customer delays or policy restrictions could affect order timing.
• CXMT’s capacity expansion and IPO strengthen China’s DRAM supply, but the HBM gap remains a structural constraint.
• Morgan Stanley discloses investment-banking or other business relationships with several covered companies, and investors should be mindful of potential conflicts of interest.

What to watch

• Whether ASML’s subsequent EUV orders, FY27/FY28 shipment visibility, and management commentary on value pricing remain consistent.
• 2Q26 and subsequent hyperscaler capital expenditure, to verify whether AI-infrastructure monetization continues to support semiconductor demand.
• Year-over-year changes in DRAM contract prices, NAND/DRAM inventories, and the pace of inventory digestion by module makers.
• Whether memory companies’ EPS revision rates continue to decline from elevated levels and whether P/B valuations contract further.
• The implementation of CXMT’s STAR Market financing, capacity expansion, DDR5 share, and HBM technology progress after listing.
• Progress in Apple AI’s rollout in China and its actual contribution to AAPL and the on-device AI ecosystem.
• Developer adoption, inference pricing, performance reproducibility, and hallucination-rate improvement following the release of Kimi K3’s open weights.