Best AI & Semiconductor ETFs for 2026: Why the Funds With "AI" in the Name Keep Finishing Last
Here's the strangest stat on our Rankings page right now: in the biggest year the AI trade has ever had, the fund literally named "Robotics & Artificial Intelligence" is losing money, while a plain semiconductor index fund is up 69%. Using ETFValuer's live scoring model, we grade every mainstream way to own the AI trade — and unpack why the AI-branded products keep landing at the bottom of the table.
Educational content — not financial advice. Data reflects ETFValuer's daily-updated rankings as of July 17, 2026. ~11 minute read.
1. Quick Picks
Every fund below is scored live by ETFValuer's ranking model — return, Sharpe ratio, expense ratio, max drawdown, AUM, and volatility, each percentile-ranked against our full tracked universe of ~120 ETFs. Full methodology is on the Rankings page.
| Best for | Ticker | YTD Return | Expense Ratio | Grade |
|---|---|---|---|---|
| Best overall (score) | SMH | +52.4% | 0.35% | B+ |
| Best 12-month performance | SOXX | +69.3% | 0.34% | B+ |
| Lowest cost | XLK | +23.3% | 0.08% | B+ |
| Best current valuation | VGT | +21.2% | 0.09% | B |
| Lowest volatility | QQQ | +15.4% | 0.18% | B |
| Bottom of the table | ROBO / BOTZ | +12.3% / -3.2% | 0.95% / 0.68% | D / F |
2. Three Ways to Own the AI Trade
"AI ETF" gets used for three very different kinds of fund, and the difference explains almost everything about the grades below:
- Semiconductor pure plays (SMH, SOXX). Concentrated baskets of 25-30 chip designers, memory makers, foundries and equipment suppliers. This is the "picks and shovels" layer where AI capital spending actually lands — data-center GPUs, high-bandwidth memory, lithography. Highest returns this cycle, and by far the highest volatility.
- Broad tech sector funds (XLK, VGT). Several hundred technology stocks weighted by market cap, so Nvidia, Microsoft, Broadcom and Apple dominate. You still get heavy AI exposure — roughly 40% of these portfolios sits in AI-linked mega caps — but with software and hardware businesses cushioning the semiconductor cycle, at about a quarter of the fee.
- Thematic "AI" funds (BOTZ, ROBO and ~400 others). Rules-based baskets of companies whose products use AI — industrial robotics, factory automation, healthcare robotics, autonomous vehicles. Marketing-friendly, fee-heavy, and structurally underweight the mega caps actually monetizing AI today. US-listed thematic ETFs collectively hold over $250B, so this is not a niche mistake.
QQQ sits just outside these categories — a Nasdaq-100 fund rather than an AI fund — but it's the vehicle many investors actually use for tech exposure, so we grade it alongside. See QQQ vs VGT for that direct comparison.
3. July's Trillion-Dollar Reality Check
Timing matters for this article: over the past month the semiconductor group has had its sharpest pullback of 2026. The Philadelphia Semiconductor Index dropped roughly 10% in early July — over a trillion dollars of combined market value — and the drawdown shows up clearly in our one-month numbers: SMH -12.1%, SOXX -15.6%, against -5.0% for QQQ and -7.3% for XLK.
The commonly cited triggers were a cluster of "custom silicon" headlines — Meta advancing its own AI-chip program, reports of DeepSeek developing an in-house accelerator, SK Hynix signalling a slower pace of high-bandwidth-memory expansion, and fresh doubts about Intel's foundry yields. Most analyst commentary framed the move as a valuation reset rather than a change in AI demand — which is exactly what our model's "Rich" valuation flags on SMH and SOXX had been pointing at (more in section 7).
4. The Semiconductor Pure Plays: SMH vs SOXX
SMH — VanEck Semiconductor ETF (Grade B+, Score 76.5, Rank #15)
The highest-scoring fund in this article and the industry's flagship, with $77.2B AUM. SMH is cap-weighted across roughly 25 names, which currently puts about 18% of the fund in Nvidia and roughly 44% in its top five (Nvidia, TSMC, Broadcom, Micron, AMD). 0.35% expense ratio, +52.4% YTD, +96.3% 1-year return, +275.0% 3-year return — the best three-year number of any fund we track in this group — Sharpe 2.47, volatility 37.0%, max drawdown -35.7%. It is effectively a concentrated bet that the biggest AI-compute names stay the biggest winners. Full SMH breakdown →
SOXX — iShares Semiconductor ETF (Grade B+, Score 75.7, Rank #19)
The same industry with a different weighting philosophy: SOXX uses modified market-cap weighting that caps single-stock concentration, so as of mid-2026 its biggest positions were Micron (~12%), AMD (~9%) and Broadcom (~7%), with Nvidia held to roughly 6%. That broader spread is exactly why it outran SMH this year — +69.3% YTD and +116.4% over 12 months — as the rally rotated from Nvidia into memory and the rest of the supply chain. The trade-off: the highest volatility (42.4%) and deepest max drawdown (-41.4%) in this article, with a 2.24 beta. 0.34% expense ratio, $47.8B AUM, Sharpe 2.63. Full SOXX breakdown →
Which one? SMH if you want the trade expressed through the mega-cap leaders; SOXX if you want the whole supply chain with concentration caps. Their long-run returns come from the same industry engine — owning both mostly duplicates exposure. Check any pair of these funds in the Stock Overlap tool before doubling up.
5. The Quiet Winners: XLK, VGT and QQQ
If the pure plays are too sharp an instrument, the broad tech sector funds deliver most of the same theme at a fraction of the fee and volatility — and they hold three of the top six grades in this group.
XLK — Technology Select Sector SPDR (Grade B+, Score 75.6, Rank #20)
S&P 500 technology stocks, cap-weighted. 0.08% expense ratio, $123.9B AUM, +23.3% YTD, +38.4% 1-year, +105.6% 3-year, Sharpe 1.36, max drawdown -25.7%. Scores within a point of SMH and SOXX on the model while carrying "High" rather than "Very High" risk and a "Fair" rather than "Rich" valuation label. Full XLK breakdown →
VGT — Vanguard Information Technology ETF (Grade B, Score 73.5, Rank #26)
Vanguard's broader take on the same sector — more mid- and small-cap tech names beneath the same mega-cap top end. 0.09% expense ratio, $169.2B AUM, +21.2% YTD, +35.8% 1-year, +107.5% 3-year (slightly ahead of XLK), Sharpe 1.31. Notably, VGT is the only fund in this article our valuation model currently labels "Cheap" relative to its own history. Full VGT breakdown →
QQQ — Invesco QQQ Trust (Grade B, Score 70.6, Rank #38)
Not a tech fund by mandate, but the Nasdaq-100's weighting makes it the most popular AI-adjacent vehicle on the market at $490.1B AUM. +15.4% YTD, +27.4% 1-year, +91.2% 3-year, 0.18% expense ratio, Sharpe 1.20 — and the only fund in this article rated "Moderate" risk, with 18.7% volatility and a -22.8% max drawdown. The sensible default for investors who want growth-tech exposure without a sector-level bet. Full QQQ breakdown →
6. The "AI" Fund Paradox: BOTZ and ROBO
Now the strange part. The two best-known funds with artificial intelligence in their actual mandate are the two worst-graded funds on our entire site:
| Ticker | Fund | YTD | 1Y | 3Y | Sharpe | Fee | Grade |
|---|---|---|---|---|---|---|---|
| BOTZ | Global X Robotics & AI | -3.2% | +10.4% | +23.9% | 0.21 | 0.68% | F |
| ROBO | Robo Global Robotics & Automation | +12.3% | +32.5% | +36.9% | 1.06 | 0.95% | D |
| SOXX | (reference) iShares Semiconductor | +69.3% | +116.4% | +217.9% | 2.63 | 0.34% | B+ |
BOTZ currently ranks dead last — #120 of 120 tracked funds — with ROBO at #114. In the biggest AI year on record, BOTZ is negative. Three structural reasons, none of them bad luck:
- They own AI's customers, not its suppliers. These indexes hold "applied AI" businesses — industrial robotics, Japanese factory automation, surgical robots, autonomous-vehicle parts. Those companies buy compute; the profits this cycle have flowed to the companies that sell it. The mega-cap compute stack is either absent or held at 1-2% weights.
- The fee hurdle is brutal. At 0.68-0.95%, these funds charge eight to twelve times XLK's 0.08%. Over a decade, that difference alone consumes a meaningful slice of returns — run the numbers in the Fee Drag Calculator.
- Dilution by design. Spreading 68-80 holdings across a theme means the rare genuine winner is a rounding error in the portfolio. Thematic indexes systematically add stocks after a theme gets hot — buying high is built into the methodology.
7. Valuation, Risk and Position Sizing
What the model flags right now, post-selloff: SMH and SOXX both carry "Rich" valuation labels and "Very High" risk ratings, with trailing P/E ratios near 37 against QQQ's 31. XLK screens "Fair," and VGT — helped by its mid-cap tail — screens "Cheap." None of that is a prediction; it's a description of how much optimism is already in each price.
A reasonable framework used by many long-term investors:
- Core (70-90%): a broad index fund — see which ETF should I buy first? if you're starting out. This already contains substantial AI exposure through the mega caps.
- Satellite (5-15%): one concentrated position — SMH or SOXX for the pure semiconductor bet, or XLK/VGT for the softer version. One, not several: these funds overlap heavily.
- Skip: high-fee thematic wrappers whose top holdings you can't name. The grades above are the argument.
And if the July drawdown made you check your portfolio twice, that's information too — the Portfolio Manager can paper-trade a semiconductor position alongside your real holdings before you commit capital.
8. Frequently Asked Questions
What is the best AI ETF for 2026?
On ETFValuer's live scoring model, the best-graded ways to own the AI trade are plain semiconductor index funds — SMH (Grade B+, Score 76.5) and SOXX (Grade B+, Score 75.7) — followed by the broad tech sector funds XLK and VGT. The funds with "AI" in the name or mandate, like BOTZ and ROBO, currently sit at the bottom of our entire tracked universe.
Is SMH or SOXX better?
They track the same industry with different concentration rules. SMH is cap-weighted with roughly 18% in Nvidia and about 44% in its top five names, so it behaves like a leveraged bet on the biggest AI-compute stocks. SOXX caps single-name weights, spreading exposure across memory, analog and equipment makers — which helped it outperform in 2026 (+69.3% YTD vs +52.4%) as the rally broadened, at the cost of higher volatility (42.4% vs 37.0%).
Why do AI-branded ETFs like BOTZ underperform?
Three structural reasons: they hold "applied AI" names like industrial robotics and factory-automation firms rather than the mega-cap compute stack actually monetizing AI; their fees (0.68%-0.95%) are up to twelve times higher than a plain sector fund; and their diluted weighting means the handful of true AI winners they do hold are too small a position to move the needle. BOTZ is down 3.2% in 2026 while SOXX is up 69.3%.
Is it too late to buy semiconductor ETFs in 2026?
Nobody can time it, but the data is worth knowing: even after July's roughly 10% industry pullback, SMH and SOXX are still up 52-69% year-to-date, both carry a "Rich" valuation label on our model, and both have max drawdowns of -36% to -41% over the past three years. That argues for position sizing — owning them as a satellite around a core index fund — rather than all-or-nothing timing.
Do I need an AI ETF if I already own QQQ or an S&P 500 fund?
You already have meaningful AI exposure. Nvidia, Microsoft, Broadcom, Alphabet, Amazon and Meta dominate the S&P 500 and the Nasdaq-100, so a dedicated semiconductor fund on top mostly increases concentration in stocks you already own. Check the Stock Overlap tool before adding — a 5-15% satellite position is the typical approach, not a second core holding.
Compare these funds yourself
Every metric in this article updates daily. See current grades, or check any of these tickers side-by-side.