iShares Semiconductor ETF
Rating
Accumulate
Adding on Dips — Active Accumulation
Combined average of Moat (AI Resilience), Growth, and Valuation scores.
Moat Score
A diversified wrapper around the AI capex cycle's most defensible companies — TSMC's foundry monopoly, ASML's EUV monopoly, NVIDIA's CUDA network, and Broadcom's custom silicon franchise.
SOXX's moat is derived, not direct — the wrapper itself is commoditised, but the basket aggregates moats:
- Concentrated Quality: The ICE Semiconductor Index caps individual weights at ~8%, but the top 10 holdings — NVDA, AVGO, TSM, AMD, QCOM, TXN, AMAT, LRCX, KLAC, MU — represent ~60% of the fund and include several of the strongest-moat businesses in global tech.
- Vertical Coverage: Exposure spans the entire semi value chain: design (NVDA, AMD, QCOM), foundry (TSM, INTC), EDA (in adjacent SMH but absent here), equipment (AMAT, LRCX, KLAC), and memory (MU). A bet on AI compute without single-stock blowup risk.
- Liquidity & Cost: Mid-tier 0.35% expense ratio with daily liquidity above $1B and tight bid/ask spreads. Tax-efficient ETF wrapper avoids the K-1 and constituent rebalancing friction of holding individual semis directly.
Ten Moats Verdict
SOXX inherits the AI-resilient moats of its constituents — TSMC's foundry monopoly, ASML's EUV monopoly, NVIDIA's CUDA network — while diluting single-name risk. The wrapper itself has no moat, but the underlying basket is structurally the AI build-out's most defensible exposure.
N/A — SOXX is a passive index ETF wrapper with no user-facing software interface.
N/A — the fund mechanically tracks the ICE Semiconductor Index; there is no proprietary business logic.
N/A — index methodology and constituents are fully public; the wrapper has no data moat of its own.
Underlying basket depends on the world's scarcest specialist talent — TSMC process engineers, ASML EUV physicists, NVIDIA CUDA architects — that takes decades to replicate.
The ETF is itself a curated bundle of ~30 critical semiconductor names — the bundling IS the product, and replicating it via single-stock purchases incurs commission and tax friction.
Constituents control deeply proprietary IP — TSMC's process recipes, ASML's EUV optics, NVIDIA's CUDA libraries, Synopsys/Cadence-adjacent EDA — that compound with each node generation.
CHIPS Act subsidies, export controls (EAR, BIS), and EU/Japan/Korea industrial policy entrench the incumbents in the basket and raise barriers for any new entrants.
Several constituents enjoy strong network effects — NVIDIA's CUDA developer base (4M+), TSMC's foundry ecosystem of 500+ customers, and standardised x86/Arm/RISC-V ISAs.
Semiconductors are embedded in every modern device and workload, but the ETF wrapper itself is not embedded in any customer workflow.
N/A — an index fund is not a system of record for any customer or business function.
Combined average of Moat (AI Resilience), Growth, and Valuation scores.
Moat Score
A diversified wrapper around the AI capex cycle's most defensible companies — TSMC's foundry monopoly, ASML's EUV monopoly, NVIDIA's CUDA network, and Broadcom's custom silicon franchise.
Growth Score
The semiconductor industry is in a multi-year AI-driven super-cycle: hyperscaler capex ($300B+ in 2026), sovereign AI infrastructure builds, and the migration from 5nm → 3nm → 2nm processes are all simultaneously expanding TAM. SOXX captures this without single-name risk — TSMC/ASML benefit from the fab build-out regardless of which AI chip designer wins, and AMAT/LRCX/KLAC sell tools to all of them. The trade-off is dilution: NVDA's hypergrowth gets averaged down by the cyclical memory and analog names. WSTS forecasts 12–14% industry revenue growth in 2026 with high-performance compute and HBM leading.
Valuation Score
At ~$215 — recovered from the April 2026 tariff-driven correction low of ~$185 — SOXX trades roughly halfway between the bear ($170) and base ($260) cases. The fund's blended forward P/E sits around 24× on consensus 2026 EPS, a modest premium to its 5-year average of 22× but well below the late-2024 peak of 32×. The diversified basket trades at a discount to its highest-quality constituents (NVDA, AVGO, TSM) while still capturing the bulk of their upside — an attractive risk-adjusted entry mid-cycle.
The Diversified Semiconductor Moat
SOXX's moat is derived, not direct — the wrapper itself is commoditised, but the basket aggregates moats:
- Concentrated Quality: The ICE Semiconductor Index caps individual weights at ~8%, but the top 10 holdings — NVDA, AVGO, TSM, AMD, QCOM, TXN, AMAT, LRCX, KLAC, MU — represent ~60% of the fund and include several of the strongest-moat businesses in global tech.
- Vertical Coverage: Exposure spans the entire semi value chain: design (NVDA, AMD, QCOM), foundry (TSM, INTC), EDA (in adjacent SMH but absent here), equipment (AMAT, LRCX, KLAC), and memory (MU). A bet on AI compute without single-stock blowup risk.
- Liquidity & Cost: Mid-tier 0.35% expense ratio with daily liquidity above $1B and tight bid/ask spreads. Tax-efficient ETF wrapper avoids the K-1 and constituent rebalancing friction of holding individual semis directly.
Ten Moats Verdict
SOXX inherits the AI-resilient moats of its constituents — TSMC's foundry monopoly, ASML's EUV monopoly, NVIDIA's CUDA network — while diluting single-name risk. The wrapper itself has no moat, but the underlying basket is structurally the AI build-out's most defensible exposure.
N/A — SOXX is a passive index ETF wrapper with no user-facing software interface.
N/A — the fund mechanically tracks the ICE Semiconductor Index; there is no proprietary business logic.
N/A — index methodology and constituents are fully public; the wrapper has no data moat of its own.
Underlying basket depends on the world's scarcest specialist talent — TSMC process engineers, ASML EUV physicists, NVIDIA CUDA architects — that takes decades to replicate.
The ETF is itself a curated bundle of ~30 critical semiconductor names — the bundling IS the product, and replicating it via single-stock purchases incurs commission and tax friction.
Constituents control deeply proprietary IP — TSMC's process recipes, ASML's EUV optics, NVIDIA's CUDA libraries, Synopsys/Cadence-adjacent EDA — that compound with each node generation.
CHIPS Act subsidies, export controls (EAR, BIS), and EU/Japan/Korea industrial policy entrench the incumbents in the basket and raise barriers for any new entrants.
Several constituents enjoy strong network effects — NVIDIA's CUDA developer base (4M+), TSMC's foundry ecosystem of 500+ customers, and standardised x86/Arm/RISC-V ISAs.
Semiconductors are embedded in every modern device and workload, but the ETF wrapper itself is not embedded in any customer workflow.
N/A — an index fund is not a system of record for any customer or business function.
Growth Analysis
Growth Drivers
Key Risk
If AI capex digestion arrives in 2H26 — hyperscalers deferring orders by 1–2 quarters as inference workloads consolidate on existing fleets — the basket drops 25–35% as NVDA, AVGO, AMAT all derate together. Semiconductor cycles are notoriously synchronous, and ETF diversification provides limited protection inside a single sector
Score Derivation
Base 80 (15% midpoint — semi industry CAGR through 2028 driven by AI capex, leading-edge nodes, and HBM demand) +1 trajectory (NVDA, AVGO, TSM accelerating offsetting INTC/MCHP decel) +3 TAM expansion (sovereign AI, edge inference, advanced packaging) −5 cycle risk (semi peak typically lasts 18–24 months from trough; current cycle entering year 2) = 79
Price Scenarios (12–24 Months)
Where We Are vs Targets
Loading live price…
AI capex pause and semi cycle peak combine with renewed US-China export controls, dragging the entire basket through a synchronous derating.
- Hyperscaler capex guidance for 2027 cut by 15%+ as inference workloads consolidate on existing GPU fleets and Microsoft/Meta defer 2H26 orders
- US-China escalation expands semi export controls to 14nm equipment; AMAT/LRCX/KLAC China revenue (20–25%) drops to zero by Q4
- HBM/DRAM oversupply emerges by Q3 26 as Samsung qualifies HBM3E into NVDA's supply chain, breaking the SK Hynix/Micron pricing duopoly
AI capex sustains through 2027 as sovereign and enterprise demand offsets any hyperscaler digestion, while TSMC's 2nm ramp drives a full equipment WFE up-cycle.
- 2026 industry revenue grows 13% (WSTS midpoint) with HBM and high-performance compute up 30%+, dragging the SOXX basket to a 15% earnings beat for the year
- TSMC 2nm volume production ramps in 2H26 with N2P bookings already filling 2027 capacity — ASML, AMAT, and LRCX all guide CY27 WFE above $130B
- Sovereign AI infrastructure (UAE G42, Saudi HUMAIN, India Mission, EU AI Factories) materialises as $50B+ of incremental 2026–27 demand, broadening the customer base beyond US hyperscalers
AI capex super-cycle extends through 2028 as inference scales 10× and edge AI emerges, lifting all boats — design, foundry, equipment, and memory — simultaneously.
- Inference workloads scale 10× as agentic AI deployment goes mainstream in 2027, requiring a multi-year second wave of capex on top of the training build-out
- On-device AI silicon (Apple, Qualcomm, AMD) drives a refresh cycle in PC, smartphone, and auto end-markets — broadening growth beyond the data centre
- TSMC and Samsung sustain 50%+ gross margins on 2nm/A14, while equipment makers ride the most expensive node transition in industry history (multi-pattern EUV → high-NA EUV → A2 generations)