SOXX: Thesis Intact, Rally Still A Go, Maintain Buy

iShares Semiconductor ETF (SOXX) remains a BUY, supported by robust AI-driven demand, hyperscaler capex outperformance, and strong memory pricing. The fund has returned +32% YTD and over 133% over the trailing 12 months, with AUM rising to $26.8 billion. Hyperscaler capex is projected to exceed $750 billion in 2026, with 75% allocated to AI infrastructure, reinforcing a durable multi-year demand cycle.

Analysis

The real story is not just that semis are strong, but that AI capex is becoming a self-reinforcing funding loop for the entire compute stack. Hyperscalers are now effectively underwriting a multi-year demand floor for leading-edge logic, memory, networking, and advanced packaging, which means the beneficiaries are shifting from pure beta names to the bottleneck suppliers with pricing power and limited substitutability. That argues for continued outperformance in the highest-content names, while more commoditized or lagging-node exposure risks being crowded out even if the broader ETF stays elevated.

The second-order effect is margin asymmetry: when demand is this concentrated, incremental supply tends to get bid up first by memory and packaging vendors, then by substrate, test, and equipment firms, creating a ripple that can lag the headline move by quarters. The market may still be underestimating how much of the AI buildout is a capacity-constrained infrastructure trade rather than a simple end-demand story; that favors firms with tight utilization, long lead times, and pricing discipline. Conversely, any supplier with weak gross margin leverage or high customer concentration could lag despite sector strength because buyers will keep reallocating wallet share toward true bottlenecks.

The main risk is not a near-term demand collapse but a capex digestion phase: if hyperscalers pull forward too much spending, the market could get a “good news, less upside” setup within 6-12 months as growth rates normalize. Another risk is that consensus becomes too anchored to a single AI capex runway, making the sector vulnerable to a multiple reset if enterprise AI monetization lags infrastructure spend. That said, the trend is hard to break without either a sharp memory downcycle, a capex pause from the largest cloud vendors, or export/control friction that constrains shipments to the highest-value end markets.

The contrarian view is that SOXX may still be under-owned by longer-duration capital despite the strong run, because many investors remain structurally skeptical of semis after prior boom-bust cycles. That skepticism can actually extend the rally if AI capex continues to surprise to the upside, but it also means the index is more vulnerable to crowded positioning unwinds than fundamentals alone would suggest. In other words, the secular thesis is intact, but the cleaner risk/reward may now sit in selectively owning the bottlenecks and hedging the broad basket against multiple compression.

Key Decisions for Investors

• Maintain a core long SOXX position over the next 3-6 months, but trim size on strength and avoid chasing if the ETF gaps further above trend; upside remains intact, but the marginal reward is lower after a 133% trailing-year move.
• Rotate from broad ETF exposure into bottleneck beneficiaries such as MU, AVGO, and AMAT on pullbacks over the next 1-2 months; these names should capture more of the pricing/margin expansion than the index, with better risk/reward if AI capex stays elevated.
• Pair trade: long MU / short a weaker semiconductor laggard or lower-quality diversified tech hardware proxy for 3-6 months; the thesis is that memory pricing and supply discipline should out-earn generic semiconductor beta as the cycle matures.
• Use call spreads rather than outright longs for new money in SOXX over 6-12 months, funded by selling upside above where sentiment becomes euphoric; this preserves participation while limiting drawdown if capex growth normalizes.
• Set a catalyst watch on hyperscaler earnings and 2026 capex guidance over the next 1-2 reporting cycles; a downward revision would be the fastest path to a sector de-rate, while another upside surprise likely extends the trade.