Global Semiconductors: AI cycle intact, but positioning and volatility pressure are rising

Overview

This report is J.P. Morgan's global semiconductor industry note, covering market color, buy-side expectations, positioning and flows across the US, Europe, and APAC semiconductor and related TMT segments. The central theme is that after SOX's strong Q2 gains, it has corrected and investors are reevaluating AI trade crowding, option pricing, hyperscale capex, and the timing of future earnings catalysts. The authors believe the semiconductor fundamental cycle has not been broken, but that the short term has entered a harder-to-trade phase.

Core views

Key views in the report include: first, most investors still endorse AI-driven semiconductor demand and capex logic, and the recent volatility is seen more as a positioning and risk-management issue rather than a structural start of decline. Second, long-dated options and valuations of long-duration AI beneficiaries are already expensive, so some traders are shifting from buying long-dated upside volatility to selling farther-dated upside, while still seeing value in short-dated gamma. Third, storage, WFE, and advanced packaging remain the most-watched themes, but these trades are highly crowded, and any change in demand, pricing, or capex expectations is being amplified by the market. Fourth, discussion in Europe and APAC focuses on earnings upgrades in FY27-FY28, including ASML's EUV and DUV outlook, Infineon and STMicro's AI and power semiconductor opportunities, Nokia's AI network infrastructure, Aixtron's InP/GaN/SiC multi-throat growth, and storage-capex sensitivity in Japanese equipment names like Tokyo Electron and SCREEN.

Analysis framework

The report combines desk feedback, buy-side surveys, regional industry expert views, derivatives and positioning data, flow observations, and company-level earnings expectation discussion. It is not a single-company deep valuation report; instead, through market participants' views on AI, storage, equipment, logic chips, and broad technology components, it assesses near-term risk-reward and medium-term earnings revision direction for the semiconductor segment.

Methodology notes

• Market MicrostructurePositioning and Flow Analysis

  Assesses short-term trading risk by evaluating hedge fund net flows, crowded positioning, long-short basket performance, ETF inflows, and deleveraging signals.

  The report repeatedly stresses that current risk comes primarily not from semiconductor fundamentals, but from crowding, elevated gross leverage, VAR constraints, and a lack of summer catalysts.

• Derivatives AnalysisVolatility and Term Structure Observation

  Uses preferences for long-dated vega, short-dated gamma, downside hedges, and upside shorting in options to gauge sentiment toward AI trading.

  The report notes that options for long-duration AI beneficiaries are relatively expensive, and some traders have started to consider buying 2-month downside hedges or selling 6-month upside volatility.

• Earnings Expectation AnalysisBuy-Side Consensus and Earnings Upgrade Path

  Compares buy-side expectations for FY27-FY28 EPS, equipment shipments, capex, and revenue with sell-side or company guidance.

  ASML, Infineon, STMicro, Nokia, and parts of Aixtron center on whether the buy-side has already sufficiently upgraded long-duration earnings enough to continue supporting valuation repricing.

Asset mapping & comparison

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

• ASML HOLDING NV (ASML.US)

  European semiconductor equipment leader, with OW framing maintained and buy-side expectations discussed.

  Strengths

  Long-term EUV and DUV demand remains positive, and FY27-FY28 equipment shipments, High-NA, platform upgrades, and DUV/immersion intensity could support EPS upgrades.

  Weaknesses

  Many buy-side participants have already modeled FY27 EPS above EUR 50 and revised FY28 assumptions upward; further upside in the near term will require stronger confirmation.

  Comparison

  ASML trades at a valuation discount versus US semiconductor equipment peers; if the market accepts higher EUV and DUV assumptions, that discount may be difficult to sustain.

  Risks

  If ASML cannot provide specific tool counts or if buy-side assumptions are too aggressive, the stock could repeat the past six months' pattern where expectations got ahead of the shares and price continuation stalled.

• ADVANCED MICRO DEVICES INC (AMD.US)

  Core AI compute and CPU/GPU-related discussion name.

  Strengths

  Investors are focused on the July 23 Advancing AI event, expecting progress on MI450 and potential new partnerships; the CPU segment has its own separate long narrative.

  Weaknesses

  Like ARM and INTC on the CPU side, it is already crowded; GPU compute is also viewed by some investors as a source of additional liquidity.

  Comparison

  Among AMD, ARM, and INTC, investors are debating whether they are holding the correct CPU name; AVGO short-covering also affects AI compute positioning dynamics.

  Risks

  If the AI demand narrative, LLM ARR, or model-efficiency debate weakens expectations for future compute demand, high-beta AI beneficiaries like AMD may be hit.

• Micron / Storage Chain

  Storage is one of the most controversial and most crowded sub-segments in the report.

  Strengths

  Demand for DRAM, NAND, SSD, and HDD remains constructive, and capex revisions in storage support the equipment chain; the NAND narrative of lowering inference cost has attracted attention.

  Weaknesses

  Storage trades including MU are fatigued, with some investors taking profit; Kioxia and Samsung have very high implied volatility, and VAR constraints by risk managers are forcing lower exposure.

  Comparison

  HDD remains a favored long direction for some long-duration investors, with WDC particularly mentioned; SNDK was relatively weak during the month.

  Risks

  Price, supply-driven moves, LTA, capex, and crowded positioning collectively amplify volatility, and the market may shift from dip buying to selling into strength.

• Tokyo Electron / SCREEN / Lasertec

  Japanese semiconductor capital equipment chain, driven by storage capex and advanced-process investment.

  Strengths

  Tokyo Electron is listed as a favored name, benefiting from DRAM etch share recovery, pricing expansion, and rising storage capex; SCREEN benefits from DRAM cleaning-equipment orders; Lasertec has a catalyst from TSMC A14 qualification.

  Weaknesses

  Some names are already heavily monitored by global investors, and ongoing revisions to price and share gains require continued validation.

  Comparison

  Japanese equipment suppliers are yen-denominated and have relatively lower cost structures, which may give them greater pricing power versus US peers.

  Risks

  If storage capex or TSMC advanced-node progress falls short of expectations, related equipment orders and valuations may decline.

• Nokia

  European AI network infrastructure beneficiary, discussed in an OW context.

  Strengths

  The report believes the recent roughly 20% pullback came more from cooling in the optical network segment than from a fundamental change; AI and cloud demand, plus an IP Networks and Optical Networks inflection in profitability, could support a multi-year beat-and-raise cycle.

  Weaknesses

  European long-duration investors remain underweight; although the pullback has provided an entry point, commitments to add meaningful positions are still insufficient.

  Comparison

  Compared with US optical and AI network peers, Nokia is seen as a cheaper, less crowded AI network buildout exposure.

  Risks

  Further execution proof is needed to demonstrate the scale, durability, and visibility of AI network demand, otherwise repricing may be limited.

Key data

• SOX Q2 Performancearound 88%The report says SOX rose nearly 88% in Q2, the largest quarterly gain since inception; MU, AMD, GOOGL, and INTC contributed materially.
• Hedge Fund June Performance+4.2%, YTD +11.1%Even though MSCI AC World fell 0.8%, hedge funds still performed strongly in June, with APAC, North America, and EMEA all contributing positive alpha or spread.
• US Crowded Long Basket+12.4% MTDThe US crowded-long basket's month-to-date increase was at least the highest of the past four years; the top 10 long-tech names outperformed top 10 shorts by about 40% MTD.
• US Overall Positioningaround 40th percentileUS TPM fell from around the 60th percentile at the end of May to around the 40th percentile at the end of June, down about 0.9 z-score over roughly four weeks.
• Gross Leverage Level92nd percentile over 12 months, 98th percentile over 5 yearsThe report believes gross leverage peaked after June option expiry, but the absolute level remains elevated, with seasonality pointing to continued deleveraging into late July.
• ASML Forward Assumptions90 Low-NA EUV in 2027, 105 EUV in 2028Sandeep Deshpande is more constructive on DUV and gives estimates of EPS of EUR 54.4 in 2027 and EUR 64.4 in 2028.
• Infineon Buy-Side AI Revenue ExpectationAbout EUR 2.7–3.0bn over the next 12 months, EUR 3.5–4.0bn in 2028This is above the company’s own next-12-month guidance of EUR 2.5bn, and some investors are discussing higher scenarios.
• FY26-28 Storage Capex Upliftraised from $300bn to $450bnThe report thus continues to favor front-end semiconductor capital equipment, especially companies benefiting from the storage cycle.

Impact & implications

For portfolios, the report suggests the semiconductor segment still has medium-term support from earnings upgrades and AI infrastructure buildout, but short term requires greater attention to crowded positioning, option pricing, flow roll-off, and the post-earnings-catalyst vacuum. A more suitable approach may be to participate in fundamental confirmation through earnings windows, control risk after strong moves, and wait until the September conference season and new catalysts to rebuild exposure. For stock selection, the report emphasizes names with clear capex, capacity, pricing, advanced packaging, or AI network revenue verification rather than merely chasing already-crowded high-beta AI trades.

Risks

• The AI demand narrative may come under pressure if hyperscaler capex slows, AI monetization is weaker than expected, or LLM ARR underperforms.
• Discussion around "token minimization" from model-efficiency improvements could weaken the market's extrapolation of future compute demand.
• Semiconductors, storage, and AI beneficiaries are crowded, and combined with high gross leverage and VAR constraints, they are prone to deleveraging.
• Long-dated options and valuations already reflect a substantial amount of positive AI expectation, so strong earnings alone may not be enough to drive the next leg higher.
• Summer conference and news flow is thinner, and the lack of post-earnings catalysts could lead to profit-taking and heightened volatility.
• Buy-side assumptions for FY27-FY28 EPS, EUV tools, WFE TAM, and AI revenue may be too aggressive.

What to watch

• The July semiconductor earnings window and company guidance on AI demand, order books, capex, and gross margins.
• ASML July 15 earnings and sell-side/buy-side updates, especially 2027-2028 EUV/DUV capacity and shipment assumptions.
• AMD July 23 Advancing AI activity, with focus on MI450 progress and new partnerships.
• TSMC FY26 Q2 preview and earnings, tracking FY26/FY27 capex, gross margin, N3/N2 pricing, and mature-node pricing commentary.
• Storage implied volatility, Korean supplier capex posture, NAND/SSD inference-cost narratives, and HDD earnings upgrades.
• Hedge-fund deleveraging, US Momentum drawdowns, crowded long-basket performance, and flows into/out of high-beta AI long/short names.
• Whether the September conference season provides new validation on orders, capacity, pricing, and AI infrastructure demand.