JPM Semiconductors Semi

J P M O R G A N North America Equity Research 12 October 2026

Semiconductors/Semi Cap Equipment 3Q26 Sector Preview: Constructive Setup Intact as Supply Constraints, AI Momentum, and Cyclical Recovery Support Further Revisions / Continued Equity Outperformance

We are entering the 3Q26 earnings season with a constructive, yet selective, stance Semiconductors & Semiconductor on the semis / semicap equipment sector. Fundamentals in our sector have Capital Equipment / IT Hardware continued to improve over the last 3 months, supported by the same forces that have Harlan Sur AC driven the group's YTD outperformance (SMH up 68%): sustained AI ( infrastructure investment, persistent supply tightness across the broader semis

supply chain, and a broadening cyclical recovery in non-AI end markets. As a Mayur Ramdhani result, we expect a majority of our covered companies to deliver solid 3Q results ( and constructive 4Q/CY27 commentary, extending the beat-and-raise cadence that has defined the group over the past several quarters. However, we think the focus Apoorva Kumar ( this quarter will be less on demand strength, which is already well understood, and more on supply, including where constraints remain most acute, how long they Akanksh Chauhan could persist, and when incremental capacity can realistically begin to alleviate ( tightness across different parts of the market (e.g., leading-edge foundry, memory, etc). That said, although the overall sector setup remains favorable, the dynamics J.P. Morgan Securities LLC vary meaningfully across the group, with AI compute, custom silicon, networking, memory, semicap equipment, and analog each facing different supply/demand dynamics, pricing environments, and revision paths. Accordingly, we remain selective across the group, favoring names where AI exposure, supply-constrained growth, pricing power and/or leverage to the broadening cyclical recovery should translate most directly into positive revisions. Our top picks continue to be AVGO and MRVL, though we would also highlight OW-rated NVDA and AMD in compute, MU and SNDK in memory, TXN and ADI in analog, and AMAT, LRCX, and KLAC in semicap equipment. On the smid-cap side, preferred names include the likes of ALAB, AMKR, and MKSI.

• GPUs / XPUs: Sustained supply/demand imbalance supports further upside to earnings revisions; custom XPUs and networking the fastest- growing layers of the AI stack; AI monetization increasingly supportive of capex durability... AI/accelerated compute demand continues to accelerate, with Street-consensus cloud capex for CY26/CY27/CY28 now standing at +97%/+47%/+9% Y/Y as hyperscalers and neoclouds remain structurally supply-constrained and demand visibility continues to extend further into the decade. We expect commentary from NVDA, AMD, AVGO, and MRVL to continue to point to upside to backlog/order visibility heading into CY27, with supply (leading-edge foundry, HBM, advanced packaging, and increasingly power) remaining the gating factor on near-term growth rather than demand. On the AI compute mix, we continue to expect custom XPUs/ASICs to take share of the accelerator TAM (ASIC/XPU rising from ~32% of units in CY25 to ~43% in CY26E and ~56% in CY27E) as hyperscalers scale TCO- advantaged internal programs, with AVGO and MRVL remaining the best- positioned beneficiaries alongside continued leadership from NVDA (merchant GPU) and AMD (MI4xx/MI5xx ramps). Networking has also emerged as one of the fastest-growing layers of the AI stack. Our hardware team’s latest industry model points to Server, Ethernet DC Switch, DCI, and

See page 26 for analyst certification and important disclosures.

Harlan Sur AC North America Equity Research ( October 2026 JPMORGAN

AI DC Switch markets growing at 31%/36%/43%/42% CAGRs through CY30, with the 800G to 1.6T transition and continued Ethernet share gains in AI DC fabrics the key drivers. On the demand-durability side, AI monetization indicators continue to strengthen - improving backlog metrics, lengthening long-term commitments, and expanding operating profitability in hyperscaler cloud/AI businesses all support the view that AI capex can sustain at these elevated levels through…