Why the iShares Semiconductor ETF Rallied 12% in January

The iShares Semiconductor ETF (SOXX) rallied 12% in January following an almost 40% advance in 2025, led by a 45.6% jump in Micron after TrendForce forecasts showed DRAM prices up ~90–95% quarter-over-quarter and NAND up ~55–60%. TSMC’s blowout results and a $52–56 billion 2026 capex guidance (up ~40% year-over-year) boosted equipment and memory-related names, while Nvidia (+2.5%) and Broadcom (-4.3%) lagged. SOXX has pulled back ~4.6% in February amid profit-taking after AMD’s earnings and softer-than-expected guidance, but elevated cloud capex projections support a continued AI-driven semiconductor demand cycle into 2026.

Analysis

Market structure: The AI transition from training to inference is broadening demand from Nvidia GPUs to DRAM, NAND and enterprise CPUs, handing short-term pricing power to memory suppliers (Trendforce: DRAM +90–95% QoQ, NAND +55–60%) and to TSMC/ASML-driven equipment vendors. Winners: MU, TSMC/ASML, semicap names and select CPU suppliers; laggards: high-valuation infrastructure aggregators (Broadcom down 4.3% Jan) and any fab-lite players with limited memory exposure. This shift centralizes value in fabrication and memory ASPs, tightening supply-demand for 2–12 months as hyperscalers ramp.

Risk assessment: Tail risks include export-controls/regulatory shocks (US/China), a hyperscaler capex pullback after 2026, or a fast inventory destocking that reverses DRAM/NAND pricing — each could cut consensus memory gains by >50% within 6–12 months. Immediate risk (days): profit-taking/earnings noise (SOXX -4.6% in Feb); short-term (weeks–months): guidance volatility from AMD/Cloud; long-term (2026+) depends on sustained AI deployment vs cyclical oversupply. Hidden dependency: memory ASPs depend on datacenter inventory cycles and wafer starts; watch Trendforce revisions and TSMC actual fab starts.

Trade implications: Direct: overweight MU (memory) and ASML/TSM (capex/equipment) for 6–18 months; underweight/hedge Broadcom and short-term beaten-up high-multiple GPU beneficiaries if guidance disappoints. Options: use 6–12 month call spreads on MU to cap premium, and sell short-dated calls against NVDA exposures to monetize elevated IV. Entry criteria: initiate if Trendforce next update sustains DRAM >+60% QoQ or cloud capex guides remain above consensus; exit/stop-loss if QoQ revision <+30% or company-level guidance misses by >10%.

Contrarian angles: Consensus understates the speed of a mean-reversion in memory — the 45.6% January move in MU may have front-loaded 6–12 months of gains, creating downside if wafer starts accelerate in 2027. Historical parallel: 2016–19 NAND/DRAM boom-bust shows 12–18 month lags between price spikes and oversupply. Unintended consequence: sharply higher memory ASPs could pressure hyperscaler margins and slow future deployment, reversing demand by late-2026 to 2027.

Key Decisions for Investors

• Establish a 2–3% long position in Micron (MU) shares now, add up to +1% if the next Trendforce update confirms DRAM QoQ >+60%; set stop-loss at -15% and target +40–60% by Dec 31, 2026.
• Initiate a 1–2% structural long in ASML (ASML) or TSM (if accessible) to play TSMC capex ($52–56B in 2026) with a 12–24 month horizon; use a -20% stop and trim 50% at +30% gain.
• Put on a relative-value pair: long MU 1% vs short Broadcom (AVGO) 1% notional for 3–6 months — thesis: memory pricing upside vs Broadcom valuation/guide risk; unwind if MU underperforms by 15% or AVGO outperforms by 10% within 90 days.
• Buy a 9–12 month MU call spread (buy OTM, sell higher strike) sized 0.5–1% of portfolio to capture continued memory upside while limiting premium; alternatively, write short-dated NVDA calls (covered) against existing NVDA exposure to harvest elevated IV, rolling monthly.
• Monitor (within 30–60 days) two metrics as position triggers/kill-switches: Trendforce DRAM/NAND QoQ revisions (kill if DRAM <+30% QoQ) and combined cloud hyperscaler capex guidance (kill/trim if guidance falls >15% below consensus).