Here's Why the VanEck Semiconductor ETF Soared in the First Half of 2026 (Hint: Diversification Works in Investing)

The VanEck Semiconductor ETF (SMH) surged 82.1% in 1H 2026, outperforming chip designers like Nvidia as semiconductor equipment makers gained the most from AI-driven capex. The article attributes strength to AI spending exceeding expectations, forcing equipment firms to ramp capital spending, plus a shift from GPU-heavy training toward more CPU/inference spending over time. It argues diversification across the sector (including equipment) can help investors capture these trends, given Nvidia’s underperformance versus the ETF.

Analysis

The real winner here is not simply “semis” but the parts of the stack with the shortest path from capex to revenue: equipment, test, and memory-linked suppliers. If AI spend keeps broadening, the next leg likely favors names tied to buildouts and utilization rather than the highest-multiple GPU designers; that is a relative headwind for NVDA if the market starts treating training as a mature phase and inference as a lower-margin, more incremental one.

Over the next 1-3 months, watch hyperscaler capex commentary and order cadence. Equipment strength can persist even if end-demand slows because backlog converts with a lag, but the thesis weakens quickly if cloud customers signal slower expansion or if lead times normalize. In a 6-18 month frame, the bigger risk is margin compression in the “picks and shovels” trade if AI infrastructure becomes more efficient and less capital intensive than the market assumes.

The contrarian view is that the rotation may be more about positioning than fundamentals: investors may be crowding into the broad semiconductor basket while underappreciating how much of the upside has already been pulled forward. Inference and CPU spending should help AMD and INTC relative to NVDA, but if the market is overcalling a durable mix shift, the semicap equipment outperformance could mean revert once the capex surge normalizes.

Key Decisions for Investors

• Long SMH / short NVDA for 1-3 months: express the internal rotation from training GPUs to broader semicap capex; stop if NVDA reclaims relative strength after earnings or Blackwell ramp commentary.
• Accumulate AMD versus NVDA on pullbacks over the next quarter: better exposure to inference/CPU mix shift with less dependence on a single platform cycle; invalidated if hyperscaler demand remains training-led.
• Treat INTC as a tactical beneficiary only, not a core long: the inference/CPU narrative can help the stock over 1-2 quarters, but execution risk is high; use small sizing or pair against NVDA.
• Add equipment beta on weakness via SMH rather than single names: best risk/reward if AI capex stays elevated, but reduce if hyperscaler capex guide decelerates or order growth rolls over.