Industry
Industry - iShares Semiconductor ETF (SOXX)
1. Industry in One Page
SOXX sits at the intersection of two industries. The product industry is the small US-listed pure-play semiconductor ETF arena — roughly seven funds competing for the same investor wallet, where the issuer (BlackRock) earns a fixed expense-ratio fee on assets under management and the products differ on index methodology, concentration, and price. The asset industry is the global semiconductor sector itself: a $791.7B (2025) industry growing toward roughly $1T in 2026, dominated at the leading edge by a handful of firms (TSMC in foundry, ASML in EUV lithography, NVIDIA in AI accelerators), and historically one of the most cyclical corners of the public equity market.
What a newcomer usually misses: SOXX is not "the chip industry" — it is a US-listing-only, modified-cap-weighted slice of it that excludes most of the foundry profit pool (TSMC is a 2.9% ADR, not the 11%+ weight it has in peer SMH) and tilts toward fabless designers, IDMs, and US-listed equipment makers. The fund's economic returns reflect the chip cycle filtered through a specific weighting choice and a 0.34% fee.
Takeaway: SOXX's cycle, returns, and risk come from the asset layer; its fee, fund flows, and competitive position come from the product layer.
2. How This Industry Makes Money
Two revenue engines need to be separated.
The ETF issuer's revenue engine. BlackRock earns the 0.34% expense ratio applied to SOXX's daily net assets. FY2025 produced $47.4M of net investment advisory fees — the entire P&L of running the fund. The fee accrues on average daily AUM, so the $47.4M implies an ~$13.9B average AUM during fiscal 2025 even though period-end net assets sat at $10.82B after the −16.21% NAV drawdown. Operating costs are minimal (sub-advisory, custody, indexing fees, distribution); the gross-to-net spread is high, which is why issuers compete fee tooth-and-nail for assets at scale. A 0.34% fee on $20–30B of AUM is a high-margin, scalable royalty on chip-sector beta.
The asset's economic engine. Underlying chip companies make money in four very different ways, with very different margins, capital intensity, and cyclicality. Fabless designers (NVDA, AMD, QCOM, AVGO) sell IP-rich silicon at high gross margins and outsource manufacturing — capital-light, IP-heavy. Integrated device manufacturers ("IDMs": INTC, TXN, MU, ADI) own their fabs and earn a wider spread but carry the capex cycle on balance sheet. Pure-play foundries (TSM) sell wafer capacity at advanced nodes; pricing power scales with node leadership (TSMC commands ~67% of pure-play foundry revenue and roughly 90% of sub-5nm output). Semiconductor equipment makers (ASML, AMAT, LRCX, KLAC, TER) sell the tools that build the fabs — order books are visible 12–24 months out, but bookings turn first when the cycle rolls over.
The economic implication: profit pools at the leading edge are highly concentrated. ASML is the only company that sells EUV lithography systems; TSMC is the only foundry currently shipping commercial 3nm/2nm wafers at scale; NVIDIA captures the bulk of merchant AI accelerator dollars. SOXX's US-only mandate gives partial exposure to the foundry profit pool (TSM ADR capped, no SMIC) but full exposure to the US fabless and equipment pools — a structural choice that diverges materially from a cap-weighted view of the entire chip industry.
3. Demand, Supply, and the Cycle
The semiconductor industry is one of the most cyclical in public equities, and the SOX index has historically led the broader tech sector into and out of recessions. Demand drivers and supply constraints sit on different clocks, which is what produces the cycle.
Demand drivers — multiple, partially uncorrelated. AI compute (datacenter accelerators, HBM memory) is the dominant 2024–26 driver. Below it sit smartphones, PCs, automotive, industrial, and traditional datacenter — markets that bottomed in 2023 and have been recovering at different speeds. The AI cycle has been so dominant that traditional end-markets matter less to fund returns than they would in a normal up-cycle.
Supply constraints — long lead times. Adding a leading-edge fab takes 2–3 years and $15–25B; adding HBM memory capacity is similarly slow. EUV lithography tools have multi-year lead times, gating how fast capacity can grow. When demand surprises to the upside, prices and margins spike before capacity arrives; when demand rolls over, the new capacity arrives into a glut and ASPs collapse.
Where the cycle hits first — in this order. Order book and book-to-bill at equipment makers, then memory ASPs (memory turns first because pricing is most exposed), then foundry utilization, then fabless inventory builds at customers, then logic ASPs. Investor sentiment usually turns 6–9 months ahead of trough earnings.
The point of the chart is the dispersion: a buy-and-hold investor in SOXX has compounded above 12% per year since inception, but path-dependent drawdowns of 30–50% on the underlying index are routine. A reader who treats SOXX as a steady-growth allocation will be surprised at the wrong moment.
4. Competitive Structure
The industry is two-layered, and so is the competitive structure.
Product layer — the ETF arena. Highly concentrated, fee-sensitive, and methodology-differentiated. SMH (VanEck, $63.4B AUM) and SOXX ($32.8B AUM at the May 2026 NAV) together hold roughly 90% of pure-play US semiconductor sector ETF assets. Below them sit four sub-scale peers and a 2x-leveraged tactical product. Fee competition is real: Invesco's SOXQ at 0.19% (15 bps below SOXX) has been growing, though it remains <5% of SOXX's AUM. There is no proprietary moat at the product level — every fund tracks a third-party index. Differentiation comes from index methodology (cap-weighted vs equal-weighted vs smart-beta), liquidity (bid/ask spreads, options market depth), and brand (BlackRock/iShares is the largest ETF brand globally).
Asset layer — the chip industry itself. Highly consolidated at the leading edge, fragmented in legacy nodes and analog. Foundry is effectively a duopoly leaning toward monopoly (TSMC ~67% of pure-play foundry revenue Q4'24, ~90% of sub-5nm production; Samsung distant second). EUV lithography is a true monopoly (ASML). AI accelerators are dominated by NVIDIA, with AMD a meaningful #2 and custom silicon (Broadcom, Marvell) selling to hyperscalers. Memory is an oligopoly (Micron, Samsung, SK Hynix). Analog/embedded chips (TXN, ADI, NXPI, MCHP, ON) are a more fragmented, relationship-driven, mature business with longer product cycles and less cyclicality.
A reader should note: SOXX's modified-cap construction systematically dampens the foundry concentration — TSMC at ~$1.7T market cap holds only a 2.9% weight in SOXX vs ~11% in SMH. Investors who want full exposure to TSMC's foundry monopoly should know that the largest profit pool in the industry is structurally underweighted in this fund.
5. Regulation, Technology, and Rules of the Game
Three external forces dominate semiconductor economics today: US export controls on advanced chips and tools to China, industrial policy (CHIPS Act and equivalents), and the technology cadence at the leading edge. None of these are features of the ETF wrapper itself; they all flow through to NAV via constituent earnings.
The most important thing to internalize: every major US tightening of export controls is a direct cap on the addressable market of SOXX's largest holdings (NVIDIA, AMD, AMAT, LRCX, KLAC). Every CHIPS Act disbursement is a subsidy to the same names' capex partners. The two forces partially offset, which is why the fund can rally through tightening regulatory headlines as long as AI capex absorbs more demand than China loses.
6. The Metrics Professionals Watch
Two metric stacks matter: one for the chip cycle (drives NAV), one for the ETF wrapper (drives whether SOXX is the right vehicle to express the chip view).
A working investor mental model: chip-cycle metrics decide whether you should own semiconductor exposure at all; wrapper metrics decide whether SOXX is the right vehicle versus its peers.
7. Where iShares Semiconductor ETF Fits
SOXX is a scale incumbent in the product layer — second by AUM behind SMH, comfortably ahead of the smaller smart-beta and equal-weighted peers — and a modified-cap-weighted, US-listed slice of the global chip industry in the asset layer. Its differentiation is the index it tracks: the NYSE Semiconductor Index applies a top-5 cap of 8%, an other-position cap of 4%, and a 10% aggregate ADR limit, producing a more even distribution of weight across roughly 30 names than SMH does across its top 25.
The relevance for the rest of the report: SOXX is not an operating company. The fund's "earnings" are advisory fees on AUM, and AUM moves with the chip cycle. When the rest of this report discusses risk, valuation, and catalysts, it is mostly discussing the chip cycle filtered through this specific weighting choice — not company-specific fundamentals at the issuer level.
8. What to Watch First
A short investor checklist for whether the industry backdrop is improving or deteriorating for SOXX:
Monthly WSTS/SIA global chip sales — three-month moving average. Sustained YoY growth above 10% has historically supported semi-ETF up-cycles; sub-zero YoY growth has preceded SOXX drawdowns of 20%+ in every cycle since 2008. Source: semiconductors.org press releases.
Equipment book-to-bill (SEMI) and ASML / AMAT / LRCX / KLAC bookings. First mover in the cycle. A turn from above-1.0x to below-1.0x has preceded the last three SOX corrections by 4–8 months.
TSMC monthly revenue release — released around the 10th of every month. The cleanest real-time read on advanced-node demand and indirectly on NVIDIA/AMD/AVGO/MRVL pull-through.
Hyperscaler capex guides at Microsoft, Alphabet, Amazon, Meta, and Oracle quarterly earnings. AI capex is currently the dominant marginal demand driver; a coordinated guide-down would be a leading bear signal across the fund's top-10 holdings.
US export-control announcements (BIS, Federal Register). Tightenings on H-series/B-series GPUs to China or on DUV/EUV tool sales are the single largest non-cycle headline risk to NVDA, AMD, AMAT, LRCX, and KLAC — each a top-15 SOXX holding. A meaningful loosening (Trump-era H20/MI308 license approvals) is a tactical positive.
Memory ASPs (HBM, DDR5, NAND). Memory ASPs turn first in corrections. Micron is a ~9% top-of-fund weight; sequential ASP weakness in any monthly Trendforce read is an early-warning signal disproportionate to the segment's industry weight.
SOXX vs SMH spread (rolling 90-day return delta). When SOXX leads, breadth is widening across mid-cap chip names — historically a constructive signal that the cycle has legs beyond NVIDIA/TSMC. When SMH leads materially, the rally is mega-cap-narrow, which has historically been late-cycle.