This Tech ETF Has More than Doubled in 2026. Is It a Good Buy for the Back Half of the Year?

iShares Semiconductor ETF (NASDAQ: SOXX) is up 108% year-to-date through June 22 and has returned 58.7% annually over the past three years, reflecting strong momentum in AI-linked chip stocks. The fund offers broad semiconductor supply-chain exposure, but its 0.34% expense ratio is materially higher than many index ETFs and could weigh on long-term returns. The article remains constructive on SOXX as a semiconductor allocation, while cautioning investors not to expect this year's pace to continue.

Analysis

The key second-order signal is not that semis are strong, but that the AI capex stack is still being re-rated as a quasi-infrastructure trade. Within this basket, the market is likely paying up for the enablers of model training and memory bandwidth first, which keeps NVIDIA, AMD, Broadcom, TSMC and the equipment names bid even if end-demand visibility is uneven. That tends to compress dispersion inside the group for a while, because investors prefer the ETF wrapper over single-name selection when the narrative is still expanding.

The more interesting divergence is between cyclical exposure and structural exposure. Intel and, to a lesser extent, Qualcomm look like relative laggards in a tape that rewards direct AI monetization and penalizes legacy mix; they can act as a funding source for longs in higher-beta AI beneficiaries. On the supply-chain side, equipment makers should continue to benefit with a lag if foundry and memory capex stays elevated, making AMAT, KLAC, and LRCX the cleaner way to express the thesis if the market starts worrying about valuation risk in the hyper-scalers and GPU names.

The contrarian miss is that this can be both a great industry and a mediocre entry point. A 60% top-10 concentration means SOXX is less a diversified semiconductor basket and more a crowded proxy for a small set of consensus winners, so ETF ownership may be substituting for active risk management rather than reducing it. If AI spending merely normalizes rather than accelerates, the highest-multiple names can de-rate quickly while the lower-quality legacy semis absorb the downside first, leaving the index vulnerable to a sharp factor unwind over the next 3-6 months.

Near term, the catalyst path is earnings guidance and capex commentary over the next 1-2 quarters, not macro. Any sign that AI server orders, memory pricing, or foundry utilization are plateauing would hit the group hard because current positioning assumes a long runway of above-trend demand. Conversely, if we get another round of upward capex revisions, the tape likely continues to reward a momentum-over-valuation regime, with semis remaining the highest-beta expression of AI optimism.

Key Decisions for Investors

• Stay long NVDA/AMD/AVGO into the next earnings window, but express via call spreads rather than outright stock; the trade works if AI capex stays hot, while upside is already crowded and needs defined risk.
• Rotate part of any SOXX exposure into AMAT/KLAC/LRCX on a 1-3 month horizon; equipment names lag the headline GPU trade but can outperform if capex holds, with better relative valuation and lower narrative saturation.
• Use INTC as a relative short against a basket of NVDA+TSM or against SOXX; the pair benefits from continued AI leadership and should work over 2-6 months if legacy exposure continues to underperform.
• If SOXX rips another 10-15% without a corresponding upward revision in capex guidance, trim and hedge with puts 3-4 months out; the risk/reward deteriorates quickly once momentum becomes the only thesis.
• For a lower-beta expression, buy TSM on pullbacks and pair against QCOM; TSM captures the manufacturing bottleneck in AI supply chains, while QCOM has weaker immediate AI monetization and lower sensitivity to the current cycle.