Up 102% in 2026, This Chip ETF Mysteriously Avoids Taiwan Semiconductor

Invesco’s PSI ETF has surged 102.37% YTD (Dec 31, 2025 to July 6, 2026), even though it holds zero Taiwan Semiconductor (TSMC) shares due to an index screen that favors US-domiciled operating companies (TSMC excluded via ADR structure). The fund’s top weights include MaxLinear (7.98%), AMD (6.26%), Texas Instruments (4.97%), Broadcom (4.84%), and Micron (4.67%), with equipment makers like KLA (4.39%), Lam Research (3.99%), and Applied Materials (3.94%). Despite the strong YTD run, PSI is down 10.34% over the trailing week as the AI/semiconductor complex cooled, implying elevated near-term volatility driven by concentrated US design-and-equipment exposure rather than direct foundry leverage.

Analysis

PSI is not a pure AI bottleneck trade; it is a levered basket of the customers and toolmakers around the bottleneck. That matters because the highest-quality economic rent in semis usually sits with the scarce manufacturing layer, while designers and equipment vendors are more exposed to capex cycles, inventory digestion, and valuation compression once the growth impulse cools. In other words, the ETF’s upside is real, but its earnings sensitivity is less durable than the asset it omits.

The next 1-3 months should be driven by whether foundry/packaging capex stays elevated enough to keep AMAT/KLAC/LRCX and memory names compounding. If those budgets hold, PSI can keep working even without direct foundry exposure; if they flatten, the basket can give back quickly because it lacks the one name with explicit pricing power. The recent pullback is a warning that positioning is now crowded, so the ETF’s risk/reward is worse than at the start of the year.

Contrarian view: the market may be underestimating the benefit of a US-only basket in a geopolitically noisy tape. Excluding Taiwan exposure removes a real discount and may make PSI a better risk-adjusted semiconductor beta in drawdowns. But over 6-18 months, if the AI buildout remains constrained by manufacturing capacity, the cleaner way to own the trade is through the foundry or the toolmakers, not the broad US basket that sits one step away from the scarcity rent.

Key Decisions for Investors

• Relative value: long TSM / short PSI into the next TSMC earnings and capex commentary cycle (1-3 months). Thesis: if foundry pricing/utilization remains tight, PSI’s omission becomes a structural underexposure. Falsify if PSI outperforms TSM by >5-7% after guidance or if TSMC margin commentary softens.
• Do not chase PSI after a doubled run; use only on a meaningful pullback or after confirmation that semi capex is reaccelerating. Risk/reward is poor for fresh longs because the ETF is more exposed to multiple compression than to scarce-margin expansion.
• Buy AMAT and KLAC on 5-8% dips as a cleaner expression of sustained AI capex intensity over the next 3-6 months. These are the highest-beta beneficiaries if tool spend stays elevated; the trade breaks if bookings or capex guides roll over.
• If you need a hedge against the 'TSMC is missing' thesis, pair long TSM with a partial short PSI rather than a broad SOX hedge. This isolates foundry scarcity versus design/equipment beta and should work best if the market starts rewarding direct supply-chain control.
• Watch the next round of semiconductor capex and booking data as the key falsifier. If equipment orders decelerate while PSI remains bid, the market is telling us the basket is being bought as a liquidity trade, not a fundamentals trade.