JPMorgan maintains an overweight view on semiconductors and semiconductor equipment, seeing demand momentum continue to strengthen after Q1

Overview

This report summarizes management feedback from 13 semiconductor and semiconductor capital equipment companies at J.P. Morgan's 54th Global Technology, Media, and Communications Conference. The core conclusion is that industry demand momentum did not weaken after Q1 results; instead, it continued to strengthen in terms of orders, backlog, customer expedites, and long-term visibility. The report attributes its constructive view on semiconductors and equipment to four themes: an AI inference inflection expanding from training into full-stack compute and connectivity demand, persistent tightness in memory and storage supply/demand, a WFE cycle extending beyond 2027, and a broad recovery beginning in cyclical end markets such as industrial, automotive, and aerospace.

Core views

The report argues that the sector's upside momentum is broad-based and durable. On the equipment side, KLAC said visibility into the next year is the best in more than a decade, with CY27 WFE expected to exceed CY26; AMAT said the CY26 upside is not being driven by CY27 pull-ins, but by higher AI intensity and creative reuse of capacity; MKSI is preparing for a $165 billion to $180 billion WFE environment after 2027. On the semiconductor side, MU said its financial outlook strengthened further versus the last earnings call; MCHP said April orders were the highest monthly level in nearly four years; and NXPI, VSH, SYNA, SNDK, and GFS also cited evidence of improving demand or structural transition. On AI, inference and agentic workloads are increasing demand for CPUs, custom ASICs, advanced packaging, CXL, PCIe, UALink, optical interconnects, and storage. On memory, MU and SNDK are locking in demand and pricing frameworks through long-term customer agreements, supporting a longer favorable pricing cycle.

Analysis framework

The report mainly uses a conference-minutes and management-comment cross-checking approach, comparing orders, backlog, capacity plans, customer agreements, product roadmaps, and long-term financial models across companies to assess whether the semiconductor cycle has broadened from a single AI theme into a wider industry recovery. The emphasis is not on a single company's valuation model, but on demand durability, supply-demand constraints, WFE structural changes, AI workload migration, and the restocking phase in end markets.

Methodology notes

• Industry conference trackingJ.P. Morgan TMC conference minutes

  Cross-checking management feedback

  By comparing management commentary from 13 semiconductor and semiconductor capital equipment companies at the same conference, the report reviews orders, backlog, customer expedites, capacity, and product roadmaps to test whether demand momentum is broad-based across the industry.

• Cycle assessmentPost-earnings demand tracking

  Continued improvement in orders and backlog

  The report treats post-Q1 changes in orders, backlog, and customer expedites as key evidence of cycle durability, with particular focus on whether these indicators continue to strengthen after earnings.

• Structural themeAI inference and full-stack opportunity

  Demand diffusion from inference, agents, and physical AI

  The report argues that AI demand is spreading from training accelerators into CPUs, custom chips, memory, connectivity, optical interconnects, advanced packaging, and edge/physical AI, expanding semiconductor content per system.

Asset mapping & comparison

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

• Semiconductor equipment sector

  Direct beneficiary of an extended WFE cycle and AI-driven investment in advanced process nodes, DRAM, and advanced packaging.

  Strengths

  KLAC, AMAT, and MKSI all provided strong WFE visibility; CY27 demand may be higher than CY26; and about 80% of CY26 incremental WFE is concentrated in areas where AMAT has leadership.

  Weaknesses

  WFE still depends on the capital spending cadence of large customers, and the pace of NAND greenfield investment is not yet fully clear.

  Comparison

  Compared with the traditional one-third logic, one-third memory, one-third ICAPS framework, incremental spending is now more concentrated in AI-related advanced logic, DRAM, and advanced packaging.

  Risks

  If AI capex slows, customers delay expansion, or CY27 demand is pulled forward, the extended WFE cycle could be weaker than expected.

• Memory and storage

  Benefiting from supply tightness in HBM, DRAM, NAND, and AI context-window expansion.

  Strengths

  MU says supply tightness will persist beyond CY26, and SNDK extended its shortage outlook to CY27; long-term customer agreements help lock in value.

  Weaknesses

  The industry has historically been highly cyclical, and supply discipline plus execution of technology transitions remain key variables.

  Comparison

  Compared with prior short-cycle memory upcycles, this report emphasizes structural supply factors and long-term customer commitments that improve cycle visibility.

  Risks

  If new supply comes online faster than expected or AI memory demand comes in below expectations, the favorable pricing environment could end earlier than anticipated.

• AI connectivity and custom chip ecosystem

  Benefiting from higher complexity in inference, agentic workloads, custom ASICs, and multi-GPU/multi-XPU communication.

  Strengths

  ALAB's PCIe Gen 6, Scorpio, Leo CXL, UALink, and optical interconnect roadmap shows continued growth in AI system connectivity content.

  Weaknesses

  Engineering resources are becoming a constraint, and new markets such as UALink, NPO, and CPO may not scale meaningfully until 2027 and beyond.

  Comparison

  Compared with pure GPU training demand, inference and agentic workloads create more dispersed, full-stack connectivity and memory bottlenecks.

  Risks

  Standards evolution, customer in-house solutions, intensifying competition, or delays in platform transitions could affect revenue realization.

• Foundry and advanced packaging

  Benefiting from custom chips, silicon photonics, CPO, supply-chain diversification, and specialty process demand.

  Strengths

  GFS highlights three major trends—data center, physical AI, and supply-chain diversification—and shows structural transition through margin improvement and design wins.

  Weaknesses

  Some opportunities, such as CPO contribution, are expected to become more visible only after 2028/2029, so near-term revenue realization will take time.

  Comparison

  GFS is positioned more as a structural transition story than a pure cyclical recovery; Intel, by contrast, is emphasizing 18A, 14A, and a longer-term node roadmap to rebuild foundry trust.

  Risks

  Yield, delivery, PDK maturity, customer trust, and advanced-node competition could all affect foundry monetization.

• Industrial, auto, and aerospace semiconductors

  Benefiting from just-beginning inventory replenishment and demand recovery after several years of weak orders.

  Strengths

  MCHP, VSH, NXPI, and others all cited improving orders, book-to-bill, customer expedites, and restocking momentum.

  Weaknesses

  Broad-based pricing increases have not yet emerged, and companies are focusing more on design wins and customer relationships, leaving limited near-term ASP leverage.

  Comparison

  Compared with the AI theme, this area reflects a more traditional but now broadening cyclical recovery.

  Risks

  If end-market demand recovers less than expected or customer restocking weakens, the recovery may not be sustained.

Key data

• Number of participating companies13 companiesSemiconductor and semiconductor capital equipment companies at J.P. Morgan's 54th Global Technology, Media, and Communications Conference.
• KLAC's view on WFE visibilityBest in more than a decadeManagement said early visibility into the next year is the best in more than a decade, and CY27 WFE is expected to exceed CY26.
• CY26 incremental WFE mixAbout 80% concentrated in advanced foundry logic, DRAM, and advanced packagingAMAT said these AI-driven areas are markets where it has share leadership.
• WFE environment MKSI is preparing forAbout $165 billion to $180 billion after 2027The company is preparing for the post-2027 demand step-up through initiatives such as its Malaysia superhub.
• MCHP order conditionsApril orders were the highest monthly level in nearly four yearsOrder momentum continued into May, showing that industrial and related cyclical recovery is underway.
• ALAB Q1 revenue$308 million, up 93% year over year and 14% quarter over quarterQ2 guidance is about $360 million, up 17% quarter over quarter.
• ALAB PCIe Gen 6 mixNow exceeds one-third of revenueThis has already become a sizable revenue contributor with NVIDIA Blackwell currently serving as the main Gen 6 GPU.
• SNDK supply shortage outlookExtended to CY27The outlook was extended from earlier guidance through the end of CY26 to CY27, with demand locked in through long-term customer agreements.
• GFS long-term revenue growth model10% to 12% long-term revenue CAGR at the corporate levelSupported by assumptions for data center, automotive, IoT, and smart mobile end markets.
• INTC 18A yield improvementImproving by about 7% per month18A is supporting Panther Lake ramp, while 14A is progressing toward the 0.9 PDK target for external customers.

Impact & implications

The report's investment implication is constructive: semiconductors are no longer just a single AI training-accelerator trade, but have broadened into a multi-dimensional upcycle driven by inference, agents, custom chips, advanced packaging, memory, WFE, and traditional cyclical end markets. Equipment makers benefit from concentrated capex in advanced logic, DRAM, and advanced packaging; memory makers benefit from supply tightness and long-term customer agreements; connectivity, CXL, PCIe, UALink, and optical interconnect suppliers benefit from AI cluster architecture evolution; and industrial, auto, and aerospace supply chains benefit from just-starting inventory replenishment.

Risks

• A slowdown in AI infrastructure capex could weaken demand for custom chips, advanced packaging, connectivity, memory, and WFE.
• If memory and storage supply comes on faster than expected, pricing upside and the benefit from long-term customer agreements could be lower than expected.
• The view that the WFE cycle is lengthening depends on 2027 demand remaining above 2026; if customers delay expansion or pull demand forward, equipment stocks could re-rate lower.
• The recovery in industrial, auto, and aerospace is still in the early restocking phase, and if end demand does not stay strong, order improvement may not last.
• Foundry-related opportunities at Intel and GFS remain subject to yield, PDK maturity, customer trust, delivery cycles, and competitive dynamics.
• New technology paths such as UALink, NPO, CPO, physical AI, and quantum computing are still developing, and the timing and scale of commercialization remain uncertain.

What to watch

• Whether CY26 second-half and CY27 WFE guidance continues to move higher, especially the statements from KLAC, AMAT, and MKSI on demand for advanced logic, DRAM, and advanced packaging.
• Follow-up disclosures from MU and SNDK on HBM, DRAM, NAND supply tightness, long-term customer agreements, and pricing frameworks.
• Whether AI inference and agentic workloads continue to drive higher content per system in CPUs, CXL, PCIe, UALink, optical interconnects, and custom chips.
• Whether orders, book-to-bill, inventory days, and customer expedites continue to improve across industrial/auto/A&D companies such as MCHP, NXPI, and VSH.
• The execution progress of GFS in silicon photonics, CPO, the Renesas partnership, MIPS integration, and its long-term gross margin roadmap.
• Whether Intel's 18A production ramp, 14A PDK, customer design wins, and organizational transformation continue to improve external customer trust.