This ETF Is Down 24% From Its Peak. Here's Why I'm Buying More.

The $72.7 billion VanEck Semiconductor ETF (SMH) is down 19.4% over the past month and 24.4% below its 52-week high, despite being up 41% year to date. The article argues the sell-off may offer an opportunity, noting that Taiwan Semiconductor Manufacturing has reported solid earnings and that slowing hyperscaler capex growth does not necessarily mean AI-chip demand is declining. It also cites expectations that hyperscaler free-cash-flow pressure from AI spending in 2026–27 could ease in 2028, while cautioning that any rebound may not be linear.

Analysis

The setup is not simply “earnings holding up, therefore the drawdown is a bargain.” The key transmission mechanism is the gap between hyperscaler capex levels and the growth rate of those budgets: even if spending stays high, slower incremental growth can reset the revenue and valuation multiples investors assign to chip suppliers. A further risk is that aggregate AI spend does not map one-for-one to merchant GPU demand; custom silicon, utilization, and the returns customers earn on deployed capacity will shape the mix and pace of orders. The article’s projected 2028 free-cash-flow recovery is a forecast, not a verified catalyst, and should not anchor valuation without checking capex, depreciation, and cash-flow guidance.

SMH offers diversified exposure relative to a single-name position, but its concentration means it remains highly exposed to NVIDIA and the same AI-capex factor. TSM’s reported resilience is a useful check on near-term demand, not proof that orders or margins will remain insulated if customer budgets reset. Over days, positioning and technical flows can drive a sharp rebound or another leg lower. Over 1–3 months, hyperscaler spending updates and semiconductor-company order commentary matter more than the drawdown alone. Over 6–18 months, the decisive question is whether deployed AI capacity generates returns that sustain another investment cycle. The contrarian opportunity is that current selling may overprice a slowdown in growth as a collapse in demand; the counterpoint is that high spending can coexist with weaker incremental chip orders and lower valuation multiples.

Key Decisions for Investors

• Treat SMH as a staged-risk entry, not a confirmed bottom: consider a small initial long only if willing to tolerate further volatility, and add only after chip-company order commentary or hyperscaler capex updates support continued demand. Avoid sizing it as a low-volatility proxy for NVIDIA.
• For a defined-risk expression, consider a call spread on SMH rather than unhedged calls; use a limited premium and avoid paying up before stabilization. The reward case is a rebound in concentrated AI exposure, while the main risk is continued multiple compression despite resilient current earnings.
• Track NVIDIA and TSM guidance, customer order trends, advanced-packaging capacity/utilization, and hyperscaler capex and free-cash-flow revisions. These are the missing data needed to distinguish a growth-rate reset from a genuine demand downturn; do not treat the article’s 2028 cash-flow recovery claim as established.
• Falsify the long thesis if subsequent guidance or order commentary indicates material AI demand deferral, or if hyperscalers cut planned spending rather than merely report slower growth. In that case, reduce SMH exposure; if spending remains high but semiconductor orders weaken, reassess whether spend is shifting toward custom silicon or other parts of the supply chain.