U.S. semiconductors and semiconductor capital equipment: Bernstein sees improving semiconductor fundamentals despite a valuation-led pullback, with AI compute and equipment remaining preferred.
The report argues that strong AI demand, hyperscaler spending and improving core-semiconductor activity support continued growth into 2027–28. It remains selective because consumer demand, elevated inventories and some end-market exposures remain weak.
Summary
The report argues that strong AI demand, hyperscaler spending and improving core-semiconductor activity support continued growth into 2027–28. It remains selective because consumer demand, elevated inventories and some end-market exposures remain weak.
- SOX is up about 76% YTD despite falling 15% from its June peak; NTM earnings are up about 113% YTD while multiples are down about 18%.
- Bernstein expects WFE to exceed $250 billion in 2028 and is comfortable with $300 billion-plus by the end of the decade.
- AI demand, tight component capacity and hyperscaler capex remain the central supports.
- PC and smartphone demand are weakening amid high memory prices, while inventories remain elevated.
- Preferred exposures include NVDA, AVGO, AMD, semiconductor equipment and ADI; Bernstein is cautious on QCOM's near-term smartphone exposure.
Report Interpretation
Overview
This semiconductor cycle tearsheet assesses industry demand, earnings revisions, inventories, valuation and positioning. Bernstein’s central view is that AI-led compute and data-center investment continue to strengthen sector fundamentals, though elevated inventories and weakening consumer end markets warrant selectivity.
Core views
Bernstein characterizes semiconductor equities as simultaneously strong and volatile. The SOX was up about 76% year to date as of September 21, despite being down 15% from its June peak. The institution emphasizes that the recent decline was valuation-led rather than a deterioration in earnings: SOX next-twelve-month earnings had risen about 113% YTD and were still about 25% above their June level, while multiples had fallen about 18%. The SOX traded at an approximately 17% premium to the S&P 500, versus roughly 62% three months earlier, and sector crowding had declined into its historical range. Bernstein therefore sees improving fundamentals beneath sentiment-driven day-to-day volatility. AI remains the report’s principal growth driver. Bernstein sees no sign of a slowdown in AI demand, citing multi-year visibility, tight supply and capacity across needed components, and very strong hyperscaler capital expenditure in Q2. It is particularly constructive on GPUs, CPUs and semiconductor capital equipment. For equipment, it expects data-center buildouts to lift wafer-fab-equipment spending above $250 billion in 2028 and supports $300 billion-plus by the end of the decade. The report favors NVDA and AVGO as AI beneficiaries with strong growth visibility in a constrained environment; it cites NVDA’s indicated 70%-plus growth into 2027 and AVGO’s potential to double revenue in 2027 and again in 2028 or beyond. It also favors AMD on CPU demand tied to the agentic-AI push and on AI opportunities including deals with OAI, Meta and Anthropic, while viewing Intel’s CPU-market and foundry narrative as improving even though its fundamentals remain challenging. Industry demand data support the positive AI and broader-cycle thesis. Total semiconductor sales rose about 131% year on year in July, following 144% growth in June; memory sales were up about 452% year on year, while non-memory sales rose about 35%. Ex-memory sales growth was 34%, 38% and 35% year on year in May, June and July, respectively. Bernstein notes that Q2 industry revenue grew about 130% year on year and that normal seasonality would imply approximately $1.6 trillion of 2026 industry sales, up about 104% year on year. It also observes generally positive revenue revisions, no current-quarter revenue misses among 24 companies in its sample, and only two misses to next-quarter guidance. Q2 results, Q3 guidance and Q4 expectations were above seasonal, while negative revisions remained near trough levels. The report distinguishes this strength from weaker consumer markets. PC shipments declined year on year in Q2, with Taiwan ODM notebook shipments down about 8% year on year in Q2 and down 24% year on year in July; July shipments were down about 31% sequentially. CPU shipments exceeded PC shipments by about 11% in Q2 versus about 2% in Q1, which Bernstein views as evidence of rising overshipment. Global smartphone shipments fell 7% year on year in Q2 and about 6% sequentially, below normal seasonality. Bernstein remains concerned about PCs, smartphones and broader consumer electronics in the second half and beyond because soaring memory prices are impairing end demand. Automotive and analog present a more mixed recovery. European and U.S. auto sales were up about 5% year on year in June and about 2% in August, respectively, but China auto retail sales fell about 21% year on year in July. Semiconductor dollar content per vehicle remained above trend, although inventories stayed elevated: auto-semiconductor and OEM inventory days fell, but Tier 1 inventory days increased. Bernstein says analog recovery appears underway, with much of the group growing year on year, and prefers ADI because of its more secular industrial exposure, a modest but attractive data-center opportunity and a lower valuation. It is less favorable on TXN because of more cyclical industrial and private-equity exposure and a valuation roughly five turns higher, while it sees NXPI as inexpensive but too exposed to automotive at this stage. Inventory remains the key counterweight to the constructive demand narrative. Channel inventory days ticked up slightly and remained well above average, although below the historical peak; semiconductor-company inventory days declined slightly but remained well above the normal range. Inventory dollars increased both sequentially and year on year, and about 79% of the sample reported higher absolute inventory dollars in Q2. Bernstein therefore treats inventories as a negative or neutral-to-negative indicator even as revenue expectations, valuation and crowding have improved. At the stock level, Bernstein keeps Outperform ratings on NVDA, AVGO, AMD, ADI, AMAT, KLAC, LRCX and PDFS. It prefers AMAT within equipment for DRAM exposure and valuation, and sees PDFS as a beneficiary of rising complexity and investment through its analytics transition and eProbe opportunity. It is Market-Perform on INTC, NXPI, QCOM and TXN. QCOM illustrates the report’s selective stance: Bernstein likes its longer-term data-center opportunity but expects near-term smartphone demand destruction, margin headwinds and Apple QTL renewal uncertainty to weigh. For INTC, it sees improving server demand and narrative support but still-challenging underlying fundamentals.
Analysis framework
Bernstein combines sector performance and valuation analysis with sales trackers, earnings and guidance comparisons, estimate-revision trends, inventory measures, end-market shipment data, hyperscaler capital-spending data, and company financial models. It then uses demand and supply conditions by end market to determine relative stock preferences.
Methodology notes
Semiconductor cycle tracking through AI demand, capacity constraints, end-market demand and inventories.
The report links AI and data-center demand, component supply tightness, consumer demand weakness and inventory levels to the direction of the semiconductor cycle.
Sales growth analysis separating memory from non-memory semiconductor demand.
Bernstein distinguishes memory-led growth from ex-memory growth to assess whether the industry expansion is broadening beyond one segment.
Forward P/E and valuation premium comparisons versus the S&P 500.
The report compares semiconductor multiples and the SOX premium to the S&P to argue that valuation has reset while earnings expectations have risen.
Sector crowding and stock reactions around earnings.
Bernstein uses crowding and post-earnings performance to assess positioning risk and the gap between operating results and market reactions.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- NVIDIA (NVDA)Key AI compute beneficiary
- Strengths
- Enormous data-center opportunity; indicated 70%+ growth into 2027; attractive valuation in Bernstein’s view.
- Comparison
- Preferred alongside AVGO.
- Risks
- A reversal in AI sentiment could pressure shares.
- Broadcom (AVGO)Key AI beneficiary
- Strengths
- Strong AI trajectory with potential revenue doubling in 2027 and again in 2028 or beyond.
- Comparison
- Preferred alongside NVDA.
- Risks
- AI-spending sustainability remains relevant.
- Advanced Micro Devices (AMD)Covered AI and CPU beneficiary
- Strengths
- CPU opportunity from agentic AI and AI deals with OAI, Meta and Anthropic.
- Weaknesses
- High expectations.
- Comparison
- Preferred for CPU exposure; Intel is improving but remains more challenged fundamentally.
- Risks
- Execution of Helios rack ramp timing.
- Analog Devices (ADI)Preferred analog exposure
- Strengths
- Secular industrial contribution, data-center exposure and improved valuation.
- Comparison
- Preferred over TXN and NXPI.
- Risks
- Industrial recovery remains cyclical.
- Applied Materials (AMAT)Preferred semiconductor-capital-equipment exposure
- Strengths
- DRAM exposure, attractive valuation versus peers and exposure to key inflections.
- Comparison
- Bernstein would lean toward AMAT among equipment names.
- Risks
- Dependent on data-center buildouts and WFE spending.
- KLA (KLAC)Covered semiconductor-equipment exposure
- Strengths
- Structural growth, durable competitive position, lower China replacement risk and disciplined capital allocation.
- Comparison
- Supported by positive WFE trends.
- Risks
- Dependent on equipment-cycle conditions.
- Lam Research (LRCX)Covered semiconductor-equipment exposure
- Strengths
- Benefits from GAA, packaging, HBM and NAND upgrades.
- Comparison
- Supported by positive CY26/CY27 commentary.
- Risks
- Dependent on WFE and technology-transition spending.
- PDF Solutions (PDFS)Covered structural complexity beneficiary
- Strengths
- Analytics transition, margin-expansion potential and eProbe as a meaningful new opportunity.
- Comparison
- A non-pure-play semicap way to participate in rising semiconductor complexity and investment.
- Qualcomm (QCOM)Covered mixed-exposure company
- Strengths
- Longer-term data-center diversification story.
- Weaknesses
- Smartphone demand destruction and margin headwinds.
- Comparison
- Less favored than AI-compute leaders and ADI.
- Risks
- Apple QTL renewal uncertainty and near-term smartphone weakness.
- Intel (INTC)Covered CPU and foundry recovery exposure
- Strengths
- Server strength and a more supportive CPU and foundry narrative.
- Weaknesses
- Fundamentals remain challenging.
- Comparison
- Bernstein is more positive than it has been for some time, but keeps Market-Perform.
- Risks
- Operational fundamentals may not validate the improved narrative.
- NXP Semiconductors (NXPI)Covered automotive semiconductor company
- Strengths
- Solid execution and increasingly inexpensive valuation.
- Weaknesses
- High automotive exposure.
- Comparison
- Cheaper than peers but less preferred than ADI.
- Risks
- Recovery pace and composition remain uncertain.
- Texas Instruments (TXN)Covered analog company
- Strengths
- Potential gross-margin upside.
- Weaknesses
- More cyclical industrial presence, private-equity exposure and higher valuation.
- Comparison
- Approximately five turns more expensive than ADI according to the report.
- Risks
- Cyclical industrial exposure.
Key data
- SOX year-to-date return~76%As of September 21, 2026; approximately 15% below the June peak.
- SOX NTM earnings change~113% YTDStill about 25% above the June peak period.
- SOX valuation change~18% decline in multiples YTDSOX traded at an approximately 17% premium to the S&P 500, versus approximately 62% three months earlier.
- July semiconductor sales growth~131% YoYMemory sales rose about 452% YoY and non-memory sales about 35% YoY.
- 2026 semiconductor industry revenue implication~$1.6T and ~104% YoY growthBased on Q2 growth and typical seasonality through 2026.
- WFE outlookAbove $250B in 2028; $300B+ by end of decadeBernstein’s view assuming continuing data-center buildouts.
- Consumer demandPC shipments -4% to -5% YoY in Q2; smartphones -7% YoY in Q2The report attributes pressure in part to high memory prices.
- Inventory increase~79% of sampled companiesReported higher absolute inventory dollars in Q2.
Impact & implications
Bernstein views the pullback in semiconductor equities as occurring alongside stronger earnings rather than weaker fundamentals. It favors AI compute and semiconductor-equipment exposures, while arguing that consumer, smartphone and inventory-sensitive exposures require a more selective approach.
Risks
- Elevated channel and semiconductor-company inventories could create a future demand or correction risk.
- High memory prices are weakening PC, smartphone and broader consumer-electronics demand.
- AI-spending funding, sustainability and sentiment remain sources of volatility.
- QCOM faces smartphone demand, margin and Apple QTL renewal uncertainty.
- China automotive demand showed weakness in July, while automotive supply-chain inventories remain elevated.
What to watch
- Hyperscaler capital expenditure and evidence that AI demand remains durable.
- Data-center buildouts and the resulting trajectory for WFE spending.
- PC and smartphone shipment trends as memory pricing affects end demand.
- Channel, manufacturer, automotive OEM and Tier 1 inventory levels.
- Revenue guidance, estimate revisions and the frequency of guidance misses.
- Whether analog and automotive recovery broadens while China auto demand stabilizes.