Three years into the artificial intelligence buildout, hyperscaler spending has carried semiconductors from a cyclical play to the highest-beta vehicle in the equity market. Microsoft, Google, Meta, and Amazon are guiding combined 2026 capital expenditures above $400 billion, up from roughly $230 billion in 2024, and most of that flows through chip designers, foundries, memory makers, and lithography vendors. For investors who want the basket rather than picking the next NVIDIA, three funds dominate the conversation: VanEck Semiconductor ETF (NASDAQ:SMH), iShares Semiconductor ETF (NASDAQ:SOXX), and Invesco PHLX Semiconductor ETF (NASDAQ:SOXQ).
Each of the three tracks represents a different semiconductor index, and after three years of capex acceleration, the gap between them has widened into a measurable one. SMH delivered about 75% year-to-date. SOXX returned about 103% over the same stretch. SOXQ landed at roughly 93%. The ranking surprises anyone assuming the most NVIDIA-heavy fund automatically wins, and it explains why fund selection matters as much as the decision to own the sector at all.
VanEck’s concentrated bet on the megacaps
The most direct way to express the AI capex thesis in ETF form comes from a portfolio built around roughly 26 holdings, with the top names carrying most of the weight and defining the character of the exposure. As of the late‑May fact sheet, the heaviest positions were AMD at about 10%, Broadcom near 10%, Micron around 9%, TSMC close to 9%, and NVIDIA near 8%, a lineup that effectively recreates the AI infrastructure stack inside a single product. Accelerators, custom silicon, HBM memory, the dominant foundry, and the lithography vendor that enables leading‑edge nodes all sit in one place, giving the fund its role as the clearest expression of AI hardware demand and semiconductor supply‑chain leverage.
The transmission from capex to chip revenue is straightforward. When hyperscalers raise capex, the dollars flow first to NVIDIA accelerators and TSMC wafer starts, then to Broadcom’s custom ASIC business, Micron’s HBM stacks, and ASML’s EUV scanners. SMH owns each of those revenue chains at meaningful weight. The fund’s $65 billion in assets makes it the deepest pool of liquidity in the category, and the 0.35% expense ratio is competitive with the iShares product.
The trade-off lies in the same concentration that drives returns. A trailing price-to-earnings ratio of roughly 52 indicates how much of the AI premium is already reflected in the top names. A single bad quarter from AMD, Broadcom, or TSMC pulls the entire fund harder than it would a more evenly weighted alternative, and the fund’s 127% one-year return cuts both ways during sentiment reversals.
iShares’ broader semiconductor basket
The appeal of this portfolio lies in how it tracks the NYSE Semiconductor Index and spreads capital across 34 holdings, using a per‑name cap structure that prevents any single position from dominating the basket. That structural difference is the entire reason it sits alongside SMH, and the 2026 tape shows exactly what that design produces in a strong market: a 103% year‑to‑date return and a roughly 148% one‑year gain, both ahead of SMH despite holding the same megacaps at lower weights, a contrast that highlights index construction and weighting discipline.
The counterintuitive result comes from breadth. A longer holdings list captures more of the equipment makers and second-tier designers that benefit from capex expansion-driven capacity additions across the supply chain. With $36.93 billion in assets, the fund has the institutional liquidity of a benchmark product without the megacap tilt. Its 0.34% expense ratio essentially matches SMH, so the choice between them comes down to portfolio construction rather than cost.
Investors paying attention to risk should note the fund’s beta of 1.8, meaning it moves roughly 80% more than the broader market in either direction. A P/E of 59 and a dividend yield of 0.24% confirm this is a growth instrument with minimal income contribution.
Invesco’s cost-advantaged challenger
The appeal of this portfolio comes from the way it tracks the PHLX Semiconductor Index, the sector’s oldest benchmark and the one most commonly quoted on trading desks as the SOX, giving it a lineage that anchors the exposure in the industry’s traditional reference point. The fund launched in 2021 and has grown to 2.61 billion dollars in assets, modest compared with its rivals but easily large enough to deliver retail‑scale liquidity. The fee structure is the differentiator, because at 0.19% it undercuts both SMH and SOXX by roughly 16 basis points, and for a 100,000‑dollar position held for ten years at a 12% growth rate, that gap compounds to roughly 3,500 dollars in savings compared with the VanEck product, a reminder of how expense discipline quietly shapes long‑run returns.
The fund holds 32 names and runs a slight mid-cap tilt relative to SMH, which softens single-stock concentration without abandoning the megacap exposure that drives the theme. SOXQ’s 93% year-to-date return sits between SMH and SOXX, roughly what the holdings overlap implies. For an investor seeking nearly identical exposure to SMH at a meaningfully lower fee, SOXQ is the overlooked option that rarely appears on a sort-by-AUM screen.
The tradeoff shows up in depth and liquidity. The fund is smaller, bid-ask spreads can widen during volatile sessions, and the options market is thinner if hedging matters.
Matching the fund to the investor
The choice tracks investor priorities more than fund quality. An investor who wants maximum exposure to the NVIDIA, TSMC, and Broadcom trio and accepts the concentration that comes with it gravitates to SMH. An investor worried about single-name risk and willing to let the per-name cap dilute the megacap weights leans to SOXX, which has rewarded that breadth in 2026. An investor focused on fee drag over a decade-plus holding period and comfortable with a smaller fund picks SOXQ.
A blended approach also exists. A 40/35/25 split across SMH, SOXX, and SOXQ captures the megacap tilt of the VanEck product, the breadth of the iShares basket, and the cost advantage of Invesco’s challenger in a single allocation. With hyperscaler capex still climbing into 2026 and Taiwan-manufactured leading-edge chips accounting for roughly 90% of global supply, owning the basket rather than a single chip stock remains the cleanest structural way to play the cycle.
Contact [email protected] for any questions or corrections.
Top Gaining Stocks
Top Losing Stocks