This website uses cookies

Read our Privacy policy and Terms of use for more information.

KEY POINTS

- The iShares Semiconductor ETF (SOXX) rose 40.4% in April, its largest monthly return in the fund's 25-year history, and pulled in roughly $2 billion in net new flows during the month.

- The VanEck Semiconductor ETF (SMH) added 32.2% in April, also a record monthly move, and has gathered roughly $7.3 billion in net inflows over the trailing twelve months.

- Traders will watch Computex Taipei next week, Nvidia's May 27 earnings, and the volatility regime around the AI capex narrative for signs the chip ETF rally is mature.

The iShares Semiconductor ETF (SOXX) rose 40.4% in April, its largest monthly return in the fund's 25-year history, and the VanEck Semiconductor ETF (SMH) added 32.2% in the same period — also a record monthly move, ETF.com reported. Both funds extended gains in the first two weeks of May as AI infrastructure spending commitments from the largest cloud platforms continued to lift the entire semiconductor complex.

The flow data confirms the move is institutional. SOXX pulled in roughly $2 billion in net new flows during April, the largest single-month inflow in the product's history, and SMH has gathered approximately $7.3 billion in net inflows over the trailing twelve months. SOXX is up about 60% year-to-date through May 14, SMH has climbed roughly 45%, and the First Trust Nasdaq Semiconductor ETF (FTXL) has surged close to 74%, 24/7 Wall St. noted.

Why the Three Funds Behave Differently

SMH and SOXX look similar on the label and behave differently on the tape. SMH is a market-cap-weighted product that concentrates roughly 20% of its assets in Nvidia, with Taiwan Semiconductor at roughly 10% and Broadcom near 8%. The fund is effectively a leveraged bet on the AI training and inference supply chain — when Nvidia, TSMC, and Broadcom move together, SMH moves harder. SOXX uses a modified-cap-weighted methodology that caps single-name exposure at 8%, which spreads risk across roughly 30 holdings and reduces the volatility drag from any single name. The result is that SMH outperformed SOXX during the Nvidia-led rally phases of 2024 and early 2025, but SOXX has caught up in 2026 because the chip rotation has broadened into Intel, AMD, Micron, Marvell, and Applied Materials — names where SOXX's relative weight is higher.

FTXL is the dark-horse outperformer. The First Trust product follows the Nasdaq US Smart Semiconductor Index, which screens for value, growth, and quality factors and rebalances quarterly. That methodology rotates more aggressively into the names where momentum has been strongest, which explains the 74% YTD gain — but it also makes the fund more vulnerable when the rotation reverses, because positioning is concentrated in whichever cohort has been winning over the prior 90 days.

The Capex Number Behind the Rally

The flow story rests on a single number: Amazon, Alphabet, Microsoft, and Meta plan to spend close to $700 billion combined on AI infrastructure in 2026. That figure, drawn from updated guides on Q1 earnings calls, is roughly double the 2024 hyperscale capex print and represents the largest concentrated industrial buildout since the buildout of the 19th-century US railroad system on a real-dollar-spend basis. Roughly 35-40% of that spend lands in semiconductor revenue across the AI accelerator, networking silicon, HBM memory, and foundry capacity stack — which translates to roughly $250 billion in addressable revenue flowing to the names that dominate SOXX, SMH, and FTXL.

Cisco's earnings print Wednesday (covered in Article 1) added an unexpected $4 billion of additional 2026 AI order visibility on top of that capex frame. Cisco is not held in SOXX or SMH but is a meaningful position in the technology-sector ETFs — XLK, VGT, and IYW — which means the AI infrastructure trade is now showing up across multiple ETF wrappers rather than just the pure-play chip funds.

Where the Risk Lives

The risk is concentration. SMH's top three holdings represent roughly 38% of fund assets, which means a single-name disappointment from Nvidia, TSMC, or Broadcom can pull the fund down 5% to 8% in a session. The May 27 Nvidia earnings print is the most likely vehicle for that kind of move. Implied volatility on NVDA into the print sits near 6% for a single-day move, which translates to roughly 2.5% for SMH and 1.8% for SOXX given the beta.

The second risk is positioning. Retail sentiment surveys captured this week show Reddit-active traders are unusually cautious despite the record YTD performance, with users on the WallStreetBets and ETFs subreddits flagging the parabolic chart structure and the historical mean-reversion pattern after 40%-plus monthly returns. The disconnect between flows (institutional bullish) and retail sentiment (cautious) is often constructive for a continuation, but it also concentrates the unwind risk into the institutional cohort if a macro shock — Fed surprise, geopolitical event, or AI capex guidance cut — triggers a position-management exit.

The decision point arrives next week. Computex Taipei opens May 19 with back-to-back keynotes from Jensen Huang, Lisa Su, and Lip-Bu Tan, followed by Nvidia's May 27 print. A clean set of AI roadmap milestones and an inline-or-better Nvidia guide extends the chip ETF run into the summer. A miss on either resets the entire complex.

Keep Reading