Semiconductor ETFs have been the defining investment story of 2026 — SOXX is up 90% in the first half, FTXL has doubled, and DRAM, the AI memory ETF, has grown from a niche fund to nearly $17 billion in assets. But the category has expanded far beyond "just buy SMH." This guide breaks down every major product and explains who each one is actually for.
The Semiconductor ETF Landscape at a Glance
ETF | Focus | Expense Ratio | AUM | H1 2026 Return | Best For |
|---|---|---|---|---|---|
SMH | Broad semiconductor (market-cap weighted) | 0.35% | $68.7B | +68.78% | Core long-term holding |
SOXX | Broad semiconductor (more balanced) | 0.35% | $38.0B | +90.03% | Core long-term holding |
FTXL | Nasdaq semiconductor index | 0.60% | $3.48B | +100.06% | Growth-tilted core |
PSI | S&P semiconductor index | 0.60% | $3.38B | +94.81% | Diversified core |
XSD | Equal-weight semiconductor | 0.35% | $3.38B | +87.03% | Diversified, lower mega-cap concentration |
DRAM | AI memory chips | 0.75% | $16.79B | +77% | AI infrastructure theme |
EUV | EUV lithography technology | 0.75% | ~$600M | +10% | Chip-making equipment layer |
SOXL | 3× daily leveraged semi bull | 0.89% | ~$26B | +440% | Short-term traders only |
SOXS | 3× daily leveraged semi bear | 0.88% | ~$600M | -89.93% | Short-term hedgers only |
SMH vs. SOXX: The Broad Semiconductor Giants
VanEck Semiconductor ETF (SMH) and iShares Semiconductor ETF (SOXX) are the two most widely held semiconductor ETFs. They track different indexes and have meaningfully different performance profiles despite covering the same general category.
SMH tracks the MVIS US Listed Semiconductor 25 Index — a market-cap-weighted index of 25 of the largest US-listed semiconductor companies. Nvidia's roughly 15% weighting is both SMH's greatest strength (when Nvidia runs) and its most meaningful risk. SMH is the more liquid of the two, with daily trading volume and AUM ($68.7B) that dwarf most other sector ETFs.
SOXX tracks the ICE Semiconductor Index, which holds 30 stocks with a more balanced weighting methodology. For three years running (2023–2025), concentration won decisively. But entering 2026, that changed — as the rally broadened across mid- and large-cap names, SOXX returned +90.03% in H1 2026, outpacing SMH by more than 20 percentage points.
DRAM: The AI Memory Pure-Play
Roundhill Memory ETF (DRAM) is the most interesting new fund in the semiconductor category. It focuses on companies making high-bandwidth memory (HBM) — the specialized chips that AI accelerators like Nvidia's H100 and H200 need to process large models at speed.
Why does this deserve its own ETF? Traditional DRAM manufacturers — Micron, Samsung, and SK Hynix — have spent decades trapped in a brutal commodity cycle: prices rise, oversupply crashes them, repeat. HBM is different. It's technologically complex, expensive, supply-constrained, and AI companies need it badly enough to pay a premium. Nvidia reportedly can't get enough of it.
DRAM holds roughly 20–21 stocks, concentrated heavily in SK Hynix, Micron, Samsung, and supporting memory ecosystem companies. It's the only ETF that treats memory not as a component of the chip stack but as the central investment thesis. That's also why it has the potential to outperform during an AI infrastructure buildout while lagging during a memory oversupply cycle.
Best for: Investors who want to express a specific view on AI memory demand rather than hold broad semiconductor exposure. Think of DRAM as a satellite position within a semiconductor allocation — not a replacement for SMH or SOXX.
EUV: The Chip-Making Technology ETF
Corgi Lithography & Semiconductor Photonics ETF (EUV) focuses on companies involved in extreme ultraviolet lithography — the chip-making technology without which the most advanced semiconductors simply cannot be manufactured. ASML, the Dutch company that holds an effective monopoly on EUV machines, is the fund's dominant holding. EUV also captures the surrounding ecosystem, including laser light source companies and specialty materials firms whose products only work with EUV processes.
The investment thesis: without EUV lithography, there are no 3nm or 2nm chips — and without 3nm or 2nm chips, there is no frontier AI hardware. This is a picks-and-shovels play on AI at the most fundamental layer of the supply chain.
SMH and SOXX hold Nvidia, TSMC, and Broadcom, but they do not specifically overweight the photonics bottleneck. A dollar in SMH is spread across the entire semiconductor complex. A dollar in EUV is concentrated on the segment where the AI infrastructure buildout is most supply-constrained right now: light sources, optical integration, and the equipment that makes both possible.
EUV is more concentrated, with heavier ASML weighting — a higher-conviction bet on the equipment layer rather than chip designers.
Best for: Investors who want equipment and tooling exposure rather than chip designer exposure. Complements rather than duplicates SMH or SOXX.
SOXL and SOXS: For Daily Traders Only
Direxion Daily Semiconductors Bull 3x Shares (SOXL) and its inverse counterpart SOXS are not investments — they are instruments. This distinction matters enormously, and these products should only be used by experienced investors.
SOXL's performance in a strong uptrend is extraordinary — it gained over 440% in H1 2026. But leveraged ETFs rebalance daily, which means they are subject to volatility decay: even when the index is flat, value erodes. In a choppy market, this decay compounds, and SOXL can lose value even when the underlying index goes up.
The two most important words here are: daily leverage. Here is a simple example:
Day 1 (index +10%) | Day 2 (index -10%) | Net change | |
|---|---|---|---|
Index | 100 → 110 | 110 → 99 | -1% |
SOXL (3× daily) | 100 → 130 | 130 → 91 | -9% |
After two days, the index is down just 1%. SOXL is down 9%. 3× daily leverage does not equal 3× the long-term cumulative return. When markets swing sharply without a clear direction, the daily reset and compounding effect cause real returns to diverge significantly from the intuitive "index × 3."
SOXL's flow story in 2026 was dramatic: after a rapid chip stock correction in late May, $3.8 billion flowed out in a single week as billions in gains evaporated — then $6.8 billion poured back in as traders re-entered for the next leg. This in-and-out pattern is exactly how SOXL is designed to be used: short-term directional positioning on semiconductor moves.
Direxion Daily Semiconductors Bear 3x Shares returned -89.93% in H1 2026. A strong semiconductor market means devastating losses for short-sellers.
The rule is simple: if your holding period is more than a few days, SOXL and SOXS are not appropriate. If you cannot monitor your position daily and exit rapidly when needed — do not enter.
The South Korea Angle: EWY
One of the most surprising stories of H1 2026 came from a fund that is not, strictly speaking, a semiconductor ETF — the iShares MSCI South Korea ETF (EWY), which returned +87.64% year-to-date. Samsung and SK Hynix together represent a substantial share of the Korean equity index, and both are dominant HBM suppliers to AI chip manufacturers.
EWY is not a semiconductor ETF — it also holds financials, consumer companies, and industrials. But for investors who want AI memory exposure with broader sector diversification and lower fees, EWY is worth understanding. The heavy weighting in Samsung and SK Hynix gives it meaningful semiconductor exposure without carrying the explicit semiconductor ETF label.
Which Semiconductor ETF Is Right for You?
If you want… | Best choice |
|---|---|
A core, liquid long-term semiconductor position | SMH or SOXX |
Broader coverage, lower mega-cap concentration | SOXX or XSD |
Pure AI memory / HBM exposure | DRAM |
Chip-making equipment / lithography exposure | EUV |
Short-term leveraged upside | SOXL (traders only, bullish) |
Short-term semiconductor hedge | SOXS (traders only, bearish) |
Indirect AI memory exposure with built-in diversification | EWY |
Bottom Line
Semiconductor ETFs are no longer a one-size-fits-all category. SMH and SOXX remain the workhorses for long-term investors — liquid, diversified enough, and battle-tested. But the emergence of DRAM (AI memory) and EUV (lithography), and their extraordinary performance this year, reflects a real structural shift in where semiconductor value is being created.
Use SMH or SOXX as your core. Consider DRAM or EUV as focused satellites if you have a specific thesis. Leave SOXL and SOXS to traders who know exactly what they're doing.
Data as of H1 2026. Not investment advice.
Comments