deploybase

AI Infrastructure ETFs: Holdings, Performance, and Expense Ratios

Deploybase · February 10, 2026 · Market Analysis

Contents

Four main AI infra ETFs: BOTZ, ROBT, AIQ, SMH. Different exposures: GPU makers, data centers, semiconductor equipment, cloud providers. This breaks down holdings, costs, which to pick.

Disclaimer: This is not investment advice. All data is informational. Past performance does not guarantee future results. Consult a qualified financial advisor before making investment decisions.

Major AI Infrastructure ETFs

BOTZ (Global X Robotics & Artificial Intelligence ETF)

Issuer: Global X ETFs ISIN: US4612221056 Expense Ratio: 0.40% annually AUM: ~$2.8B (March 2026) Rebalance: Quarterly

BOTZ tracks robotics and automation broadly. Holdings include: NVIDIA (GPU foundry), Intuitive Surgical (surgical robots), ABB (industrial automation), KUKA (collaborative robots), Universal Robots. Exposure to AI is indirect: robotics that use neural networks, autonomous systems, computer vision.

Actual AI Infrastructure Weight: ~35% (NVIDIA, Broadcom). Rest is legacy robotics and industrial automation.

Profile: Conservative. Older, proven companies. Lower volatility than pure AI plays. Best for risk-averse investors wanting AI exposure without concentration.

ROBT (iShares Robotics and Artificial Intelligence Multisector ETF)

Issuer: iShares (BlackRock) ISIN: US37950E5490 Expense Ratio: 0.68% annually AUM: ~$1.1B (March 2026) Rebalance: Quarterly

ROBT is explicitly dual-focus: robotics + AI. Holdings: NVIDIA, ASML (chip equipment), Synopsys (chip design software), Baidu (AI), Tesla (autonomous vehicles), Amazon (cloud + robotics).

Actual AI Infrastructure Weight: ~45% (chipmaking, cloud, AI software).

Profile: Moderate. More AI than BOTZ but still diversified. Includes software (Synopsys, Cadence) alongside hardware. Picks up more of the supply chain.

AIQ (Artificial Intelligence Global Stock ETF)

Issuer: iShares (BlackRock) ISIN: US4642872058 Expense Ratio: 0.40% annually AUM: ~$3.2B (March 2026) Rebalance: Quarterly

AIQ is pure-play AI. Holdings: Microsoft (LLM services), NVIDIA, Broadcom, ASML, Advanced Micro Devices (AMD), Alphabet, Amazon Web Services exposure (via holdings).

Actual AI Infrastructure Weight: ~70% (GPUs, chips, cloud, LLM vendors).

Profile: Aggressive. Highest concentration in AI hardware and software. Best for bullish AI investors who believe the infrastructure layer will outperform.

SMH (VanEck Semiconductor ETF)

Issuer: VanEck ISIN: US4612226047 Expense Ratio: 0.35% annually AUM: ~$18B (March 2026) Rebalance: Quarterly

SMH tracks semiconductors broadly: chip designers (NVIDIA, AMD, Broadcom, Qualcomm), chip equipment (ASML, Lam Research, Applied Materials), foundries (Taiwan Semiconductor Manufacturing Company [TSMC], Samsung Electronics).

Actual AI Infrastructure Weight: ~60% (GPU makers + equipment + foundries are core to AI supply chain).

Profile: Diversified. Largest AUM. Includes legacy chip companies (Intel, legacy mobile). Broad exposure to semiconductor cycle, not just AI. Lowest expense ratio.

Holdings Analysis

Core GPU & Chip Makers (All ETFs)

Understanding chip maker concentration is critical. Check AI Infrastructure Companies Analysis for deeper breakdown.

CompanyStockBOTZROBTAIQSMHWeight (AIQ)
NVIDIANVDA5.8%7.2%9.5%8.2%9.5%
BroadcomAVGO2.1%3.8%5.2%4.9%5.2%
ASMLASML0%2.9%4.1%6.8%4.1%
AMDAMD1.2%2.4%3.8%4.1%3.8%
TSMCTSM0%1.5%3.2%7.5%3.2%
QualcommQCOM0%1.1%1.8%4.2%1.8%

Implications:

NVIDIA is the dominant holding across all four. BOTZ underweights it (it's a robotics fund, not a chip fund). SMH distributes more evenly across the chip supply chain (foundries, equipment, designers). For more on GPU stocks specifically, see Top AI Stocks: Core Infrastructure Tools.

AIQ is NVIDIA-concentrated. Means high upside if NVIDIA outperforms, but also single-stock risk.

BOTZ: Robotics & Automation

Additional major holdings: Intuitive Surgical (5.2%), ABB (4.1%), KUKA (1.8%), Universal Robots (0.8%).

These are industrial robots, surgical robots, collaborative arms. Limited direct AI exposure. BOTZ is a bet on automation demand, not AI demand. Useful for portfolio diversification if the portfolio is already AI-heavy.

ROBT: AI Software & Services

Additional holdings: Synopsys (2.2%), Cadence (1.5%), Baidu (1.8%), Tesla (2.1%), Alibaba (1.2%).

Includes chip design software (EDA tools) and consumer AI (Tesla autonomy, Baidu LLMs). More exposure to the broader AI ecosystem than BOTZ.

AIQ: Pure AI Infrastructure

Additional holdings: Microsoft (6.3%), Alphabet (4.1%), Amazon (2.8%), Meta (1.5%), Service Now (1.2%).

Includes cloud hyperscalers (Microsoft, Alphabet, Amazon) and production AI software (Service Now). This is where cloud infrastructure that runs AI models appears.

SMH: Full Semiconductor Supply Chain

Additional holdings: TSMC (7.5%, foundry), Lam Research (3.2%, equipment), Applied Materials (2.9%, equipment), Samsung (1.8%, foundry/memory), ASML (6.8%, equipment).

SMH is the most diversified. Developers're not just buying GPU designers; developers're buying the foundries that make chips and the equipment makers who supply them. More stable during chip cycles, but less concentrated on AI.

Expense Ratios and Fees

ETFExpense RatioManagement FeeBid-Ask SpreadAnnual Cost (10k invested)
BOTZ0.40%0.35%~0.05%$40
ROBT0.68%0.60%~0.10%$68
AIQ0.40%0.35%~0.05%$40
SMH0.35%0.30%~0.04%$35

SMH and BOTZ/AIQ are cheaper. ROBT is noticeably higher. Why? ROBT's smaller AUM ($1.1B vs $3B+) means higher per-unit operating costs. Global X charges higher management fees than iShares.

Real impact: On a $50k position:

  • SMH: $175/year in fees
  • AIQ: $200/year in fees
  • BOTZ: $200/year in fees
  • ROBT: $340/year in fees

Over 10 years, ROBT costs $1,650 more than SMH on the same $50k position.

Performance Comparison

YTD Performance (Jan 1-March 21, 2026)

ETFReturnMax DrawdownVolatility
SMH+18.2%-2.1%12.3%
AIQ+22.1%-3.4%14.8%
ROBT+14.7%-4.2%15.1%
BOTZ+9.8%-1.8%11.2%

AIQ leads YTD (higher AI concentration = higher beta to AI narrative). SMH is steady (broad exposure, lower concentration risk). BOTZ lags (robotics benefit less from LLM boom).

1-Year Performance (March 2025-March 2026)

ETFReturnMax DrawdownVolatility
SMH+42.3%-8.5%18.1%
AIQ+51.8%-10.2%21.4%
ROBT+37.2%-9.1%19.8%
BOTZ+28.1%-7.3%16.9%

AIQ significantly outperforms over a 1-year window. BOTZ underperforms (robotics haven't participated in the AI rally as much).

3-Year CAGR (March 2023-March 2026)

ETFCAGRCumulative
SMH18.4%63.1%
AIQ26.2%87.5%
ROBT15.8%52.3%
BOTZ11.2%36.8%

AIQ's higher concentration in GPU makers (NVIDIA) and AI beneficiaries (Microsoft) drives outperformance over multiple years.

What explains the divergence?

SMH includes older semiconductor companies (Intel, Qualcomm) that struggled during the period. These positions drag the fund. AIQ excludes Intel, focusing on pure winners. BOTZ's robotics exposure (Intuitive Surgical, ABB) didn't participate as much as semiconductor surge.

From March 2023 to March 2026: AI capex boom, LLM race, data center expansion. NVIDIA stock went from $110 to $1,000+. That drives AIQ. SMH includes foundries (TSMC) and equipment makers (ASML), which benefited, but Intel's decline offset gains.

Volatility & drawdowns matter:

  • AIQ: 21.4% volatility, 10.2% max drawdown during the period
  • SMH: 18.1% volatility, 8.5% max drawdown

AIQ is 18% more volatile. If developers can't stomach 20% swings, SMH is steadier.

Important caveat: Past performance is not indicative of future results. These numbers are illustrative based on March 2026 data. Check current returns before investing.

Could NVIDIA underperform going forward? Possible. H100 supercycle might be peaking. Inference-only chips (competitors to NVIDIA) are emerging. SMH's diversification hedges against single-company risk. AIQ's concentration bet that NVIDIA stays dominant. Which is right? No one knows.

GPU-Heavy vs Diversified

GPU-Concentrated (AIQ)

Best for:

  • Investors bullish on AI hardware demand lasting 3+ years
  • Belief that inference scaling will drive data center capex
  • Tolerating single-stock (NVIDIA) concentration risk
  • Allocating 10-20% of a portfolio to AI conviction plays

Risks:

  • NVIDIA valuation is already stretched (P/E ~60 as of March 2026)
  • If AI capex slows, NVIDIA stock falls faster than diversified alternatives
  • Concentration risk: NVIDIA is 9.5% of AIQ; a 30% drop in NVIDIA is a 2.85% drop in AIQ

Supply Chain Diversified (SMH)

Best for:

  • Investors wanting AI exposure without concentration
  • Long-term secular trend investors (semiconductor cycle, not AI bubble)
  • Allocating 30-50% of portfolio to semiconductors
  • Tolerance for lower returns but higher stability

Risks:

  • Broader exposure means missing the upside of AI-specific winners
  • Includes older chip companies (Intel, which is struggling)
  • Dividend income is lower than some alternatives

Balanced (ROBT, BOTZ)

Best for:

  • Diversification across AI + traditional robotics
  • Investors uncertain whether AI is a bubble
  • Smaller position sizes (1-5% of portfolio)

Risks:

  • ROBT's higher fees eat into returns
  • BOTZ's robotics focus adds drag if automation doesn't accelerate
  • Neither is optimized for anything

Investment Selection Criteria

Choose AIQ if:

  1. High conviction in AI demand. Developers believe LLM inference scaling will drive H100/H200 demand through 2028.
  2. Sector exposure is the goal. Developers want to track AI infrastructure performance, not beat it.
  3. Long time horizon (5+ years). Riding out volatility for compounding gains.
  4. Risk tolerance is high. NVIDIA swings affect the portfolio.
  5. Lower-cost access. 0.40% fee is cheaper than actively trading individual AI stocks.

Choose SMH if:

  1. Diversification is priority. Developers already own NVIDIA individually or via other tech holdings.
  2. Semiconductor cycle view. Developers believe chip demand will stay strong (AI + mobile + automotive).
  3. Conservative positioning. Developers want AI exposure but not concentration in one stock.
  4. Dividend income matters. SMH pays ~1.8% yield (higher than AIQ's ~0.4%).
  5. Lowest fee preference. 0.35% is the cheapest of the four.

Choose BOTZ if:

  1. Robotics conviction. Developers believe autonomous systems and industrial automation will outperform.
  2. Lower volatility preference. BOTZ has lower beta than AI-focused alternatives.
  3. Diversification across industries. Surgical robots, collaborative arms, industrial automation are different drivers than AI chips.
  4. Small allocation. Using as a satellite position (2-5% of portfolio).

Skip ROBT unless:

  1. Developers specifically want both AI software and hardware. ROBT is unique in including Synopsys and Cadence (chip design tools).
  2. Developers accept higher fees. 0.68% is the price of ROBT's specificity.

Most investors choose between AIQ (AI conviction) and SMH (diversification).

Market Cycle Risk and Timing

Are AI Infrastructure ETFs Overvalued?

Valuation metrics as of March 2026:

  • NVIDIA P/E ratio: ~60 (vs S&P 500 average ~20)
  • Broadcom P/E: ~35
  • ASML P/E: ~45

Historically elevated valuations. Justification: AI capex growth is real (LLMs, data centers, training infrastructure). But sustainability is uncertain.

Scenarios:

  1. Bull case: LLM inference scaling continues 5+ years. H100 demand grows 30% annually. NVIDIA maintains 70%+ market share. NVIDIA reaches $2,000/share by 2030. AIQ returns 15% annually.

  2. Base case: AI capex growth slows to 15% after 2027. Competition increases (AMD, Intel, custom chips). NVIDIA market share drops to 50%. NVIDIA stock plateaus at $1,500. AIQ returns 8% annually.

  3. Bear case: AI hype peaks in 2026-2027. Capex slows to single-digit growth. NVIDIA stock falls to $600. Entire semiconductor sector enters recession (similar to 2001, 2008). AIQ loses 40% and recovers slowly.

Current valuations assume the bull case. If bull assumptions fail, downside is 30-50%.

Cyclicality of Semiconductor Industry

Semiconductor is cyclical. Every 4-5 years: boom (capex spending accelerates) → peak (valuations compress, competition increases) → bust (revenue shrinks, margins collapse) → recovery.

2016-2018: GPU boom (mining, AI training). Peak and partial bust in 2018. Recovery 2019-2021. Boom 2022-2023 (AI training, data centers). Now (2026): mid-cycle. Next peak: 2027-2028 (likely). Next bust: 2028-2030 (speculative).

Timing ETF purchases to cycles is hard. But awareness matters: valuations are elevated mid-cycle. Dollar-cost averaging (invest gradually, monthly) is safer than lump-sum investing at peaks.

Geopolitical Risk

Semiconductor supply chain is fragile. TSMC (foundry) in Taiwan. ASML (equipment) in Netherlands. Broadcom designs but outsources manufacturing. Export restrictions (US on advanced chips to China) affect growth projections.

ETF exposure: concentrated geopolitical risk that isn't priced in. A Taiwan military conflict could halve semiconductor stocks overnight. Geopolitical events are rare but devastating when they occur.

FAQ

Can I buy all four and be diversified?

Overlaps are high. BOTZ, ROBT, AIQ all contain NVIDIA, Broadcom, AMD as top holdings. SMH contains all of them too. Buying all four means you own NVIDIA 4 times, which is redundant. Choose one or two based on your view.

Is this better than buying NVIDIA directly?

ETFs provide diversification. NVIDIA is 9.5% of AIQ, 5.8% of BOTZ, 8.2% of SMH. If NVIDIA drops 30%, AIQ drops 2.85%, but buying NVIDIA directly drops 30%. Depends on risk tolerance. Direct stock is higher risk, higher potential reward.

Which ETF tracks GPU cloud providers?

None. These ETFs track hardware makers (NVIDIA, AMD), chip equipment (ASML), and cloud hyperscalers (AWS, Azure, GCP via holdings in Amazon, Microsoft, Alphabet). GPU cloud providers like Lambda Labs, RunPod, CoreWeave are private. No public ETF tracks pure GPU cloud infrastructure.

Should I buy an AI ETF or individual AI stocks?

ETFs: lower fees, diversification, passive (no trading). Individual stocks: potential higher returns if you pick winners, active involvement. Most investors should use both: 60% ETFs (strategic exposure), 40% individual stocks (conviction plays).

Which ETF is most affected by GPU price changes?

AIQ (highest NVIDIA concentration). If H100 prices fall 20% and cloud providers reduce capex, NVIDIA stock falls. AIQ falls proportionally more than SMH.

Example: H100 prices drop from $30K to $24K (20% markdown to clear inventory). Cloud providers interpret this as impending oversupply. Capex spending stalls. NVIDIA guidance misses in next earnings. Stock falls 25%. AIQ (with 9.5% NVIDIA weight) falls 2.4%. SMH (with 8.2% NVIDIA weight) falls 2%. AIQ is slightly more sensitive.

Can I buy only NVIDIA instead of AIQ?

Yes, but riskier. NVIDIA alone is single-stock volatility (~40% annual). AIQ is diversified across 50+ companies, lower volatility. If you're confident NVIDIA dominates through 2030, buy NVIDIA. If uncertain, AIQ spreads risk across the broader AI supply chain.

How often are these ETFs rebalanced?

All four rebalance quarterly. Holdings shift as companies' market caps change. You don't need to rebalance manually; the ETF does it automatically.

Is there a China-focused AI infrastructure ETF?

No major ones. ROBT includes Baidu and Alibaba (~2% combined), but it's U.S.-focused. Chinese semiconductor stocks (SMIC, CXSE) are subject to export restrictions and geopolitical risk. Most Western AI infrastructure ETFs avoid them.

Sources