The numbers are staggering. Global data center investment is expected to exceed $1 trillion by 2027. Power consumption from AI workloads is doubling every two years. Nvidia’s data center revenue alone surpassed $47 billion in fiscal 2024. And this is still early innings.
For investors who want exposure to the AI infrastructure supercycle without picking individual stocks, ETFs offer a diversified, cost-efficient way to position a portfolio across the four pillars driving this boom: semiconductors, data center REITs, power infrastructure, and the broad AI buildout. In this guide, we cover the best U.S. and Canadian ETFs to play each theme in 2026.
The Complete 2026 AI Data Center ETF Shortlist
| Category | Best U.S. ETF | Ticker | Best Canadian ETF | Ticker |
| Semiconductors | VanEck Semiconductor | SMH | iShares Semiconductor | XCHP |
| Data Center REITs | Global X Data Center | DTCR | No TSX equivalent | — |
| Power Infrastructure | SPDR Energy Select | XLE | Ninepoint Energy Fund | NNRG |
| Grid Modernization | First Trust Smart Grid | GRID | iShares S&P/TSX Energy | XEG |
| Broad AI Infrastructure | Defiance AI & Power | AIPO | Global X AI Semi | CHPS |
| Single-Ticker Solution | Defiance AI & Power | AIPO | Global X AI Semi | CHPS |
Why the AI Data Center Boom is the Investing Theme of the Decade
Every AI model you interact with — ChatGPT, Claude, Gemini — requires massive amounts of computing power to train and run. That computing power lives in data centers. Those data centers need chips, real estate, cooling systems, and above all, electricity — enormous amounts of it.
The AI infrastructure supercycle is not a software story. It is a physical infrastructure story. It requires steel, concrete, land, power grids, transformers, and the most advanced semiconductors ever built. The companies building, supplying, and powering this infrastructure are among the most strategically positioned businesses in the world today.
Investors who understand this are not simply buying “AI stocks.” They are buying the picks and shovels of the most capital-intensive technological revolution in history.
Pillar 1 — Semiconductors: The Brain of the AI Data Center
No chips, no AI. The semiconductor sector sits at the absolute center of the data center boom. Nvidia GPUs, AMD accelerators, TSMC fabrication, ASML lithography equipment, and Micron memory are all essential components of every AI data center built today.
Side-by-side comparison of all semiconductor ETFs covered in the section.
| ETF | Ticker | Market | AUM | MER | # Holdings | Best For |
| VanEck Semiconductor | SMH | 🇺🇸 | $65.6B USD | 0.35% | 25 | Liquidity + core exposure |
| iShares PHLX SOX | SOXX | 🇺🇸 | $34.9B USD | 0.34% | 30 | Broader diversification |
| Invesco PHLX Semi | SOXQ | 🇺🇸 | $2.1B USD | 0.19% | 30 | Lowest MER |
| iShares Semiconductor | XCHP | 🇨🇦 | $202M CAD | 1.46% | ~30 | TFSA-friendly, CAD |
| Global X AI Semi | CHPS | 🇨🇦 | $311M CAD | 1.45% | ~30 | AI tilt, CAD |
SMH — VanEck Semiconductor ETF
SMH is the benchmark semiconductor ETF for U.S. investors and arguably the single best way to play the AI chip theme in one ticker. With over $65 billion in assets under management and a rock-bottom expense ratio of 0.35%, it holds 25 of the largest global semiconductor companies. Nvidia, TSMC, and Broadcom are among the top positions. The fund returned over 50% year-to-date in 2026, making it one of the strongest performing ETFs across all categories.
Best for: investors who want the broadest, most liquid semiconductor exposure with a low MER.
SOXX — iShares PHLX Semiconductor ETF
SOXX tracks the Philadelphia Semiconductor Index — the industry’s oldest and most established benchmark. With $34 billion in AUM and an expense ratio of 0.34%, it offers slightly different weighting methodology than SMH, with more equal distribution across its 30 holdings. This makes it somewhat less concentrated in Nvidia than SMH, which can be either an advantage or a disadvantage depending on your view of Nvidia’s dominance.
Best for: investors who want semiconductor exposure with slightly more diversification across the sector.
SOXQ — Invesco PHLX Semiconductor ETF
For cost-conscious investors, SOXQ tracks the same Philadelphia Semiconductor Index as SOXX but charges only 0.19% — nearly half the cost. With $2 billion in AUM it is smaller and less liquid than SOXX, but for long-term buy-and-hold investors the cost advantage compounds significantly over time.
Best for: long-term investors who want to minimize MER on semiconductor exposure.
Canadian Option — XCHP (iShares Semiconductor Index ETF)
For Canadian investors who want semiconductor exposure in Canadian dollars without U.S. dollar currency risk, XCHP is the go-to option. It returned +65.7% year-to-date in 2026 with $202 million in AUM. It trades on the TSX and eliminates withholding tax issues when held in a TFSA — a significant advantage over U.S.-listed equivalents held in the same account.
Canadian Option — CHPS (Global X AI Semiconductor Index ETF)
CHPS takes a more focused approach, specifically targeting semiconductors tied to artificial intelligence. Its portfolio is more concentrated on AI chip designers, advanced manufacturers, and equipment makers. At $311 million in AUM with a YTD return of +37.4%, it is a strong complement to XCHP for investors who want a dual-layer semiconductor position with an explicit AI tilt.
Pillar 2 — Data Center REITs: The Real Estate of the Digital Economy
Every AI model needs a home. Data center REITs own, operate, and lease the physical facilities that house the servers, networking equipment, and cooling systems powering the AI revolution. These are infrastructure landlords collecting rent from hyperscalers like Amazon, Microsoft, Google, and Meta — the most creditworthy tenants on earth.
DTCR — Global X Data Center & Digital Infrastructure ETF
DTCR is the most direct U.S.-listed ETF for data center real estate exposure. It holds companies like Equinix, Digital Realty, Iron Mountain, and other digital infrastructure operators. With $1.96 billion in AUM and a 0.50% expense ratio, it delivered +37.3% year-to-date in 2026. It also provides exposure to cell towers, fiber networks, and other digital infrastructure assets — giving investors a broader “digital real estate” basket rather than pure data centers alone.
Best for: investors who want direct exposure to the physical real estate of the AI data center boom, with some additional digital infrastructure diversification.
VNQ — Vanguard Real Estate ETF
While not specifically a data center ETF, VNQ is the largest REIT ETF in the world with over $30 billion in AUM and a 0.13% expense ratio. Data center REITs have grown to represent a meaningful portion of its holdings. For investors who want broad REIT diversification while still capturing data center growth, VNQ offers an ultra-low-cost approach. The tradeoff is dilution — your data center exposure is blended with residential, retail, office, and industrial REITs.
Best for: investors who want REIT diversification with data center exposure embedded at rock-bottom cost.
The Data Center REIT Thesis in Plain Language
The demand side for data center space is essentially unlimited in the near term. Microsoft, Google, Amazon, and Meta have collectively committed over $300 billion in capital expenditure for AI infrastructure through 2026. Every dollar of that capex eventually needs physical space in a data center owned by a REIT. Meanwhile, the supply side is constrained — building a new hyperscale data center takes 18 to 36 months and faces acute shortages of power connections, cooling capacity, and skilled labor. This supply-demand imbalance is why data center REITs have been among the strongest performing real estate subsectors over the past three years.
Pillar 3 — Power Infrastructure: The Hidden Bottleneck
This is the most underappreciated angle of the AI data center boom — and potentially the most important for the next five years.
A single hyperscale AI data center consumes between 100 and 500 megawatts of electricity. The entire U.S. data center sector is expected to consume 8% of the country’s total electricity by 2030, up from roughly 2% today. This is creating an unprecedented demand for power generation, transmission infrastructure, and grid modernization equipment.
The bottleneck is not chips and not real estate. The bottleneck is electricity.
| ETF | Ticker | Market | AUM | MER | YTD 2026 | Angle |
| First Trust Smart Grid | GRID | 🇺🇸 | — | 0.58% | — | Grid modernization & transformers |
| SPDR Energy Select | XLE | 🇺🇸 | $40B+ USD | 0.08% | +32.9% | Broad energy — lowest cost |
| Pacer Energy Independence | AMPS | 🇺🇸 | — | 0.55% | — | North American energy infra |
| iShares S&P/TSX Energy | XEG | 🇨🇦 | $2.33B CAD | 0.61% | +46.3% | Canadian oil & gas — TSX |
| Ninepoint Energy Fund | NNRG | 🇨🇦 | $438M CAD | 1.12% | +50.1% | Active management — highest return |
GRID — First Trust NASDAQ Clean Edge Smart Grid Infrastructure ETF
GRID focuses on companies building and upgrading electrical grid infrastructure — transformers, smart grid technology, transmission systems, and power management equipment. These companies are the unsung heroes of the AI boom, supplying the physical electricity infrastructure that makes data centers possible. With a 0.58% expense ratio, it offers targeted exposure to this often-overlooked layer of the AI stack.
Best for: investors who want to play the power infrastructure angle specifically — the “picks and shovels of the picks and shovels.”
AMPS — Pacer American Energy Independence ETF
AMPS provides exposure to North American energy infrastructure including natural gas pipelines, electricity transmission, and independent power producers. As data centers increasingly turn to natural gas and nuclear power to meet their energy demands — often bypassing the traditional grid entirely through direct power purchase agreements — midstream energy infrastructure becomes a direct beneficiary of AI data center growth.
XLE — SPDR Energy Select Sector ETF
For broader energy exposure tied to the power demand theme, XLE is the largest and most liquid U.S. energy ETF with over $40 billion in AUM and an expense ratio of just 0.08%. While it includes traditional oil and gas exposure, the sector increasingly overlaps with AI power infrastructure as major oil companies invest in power generation and natural gas supply contracts for data centers.
Canadian Options — XEG and NNRG
Canadian energy ETFs benefit from the same power infrastructure thesis through a different lens. XEG (iShares S&P/TSX Capped Energy Index ETF) and NNRG (Ninepoint Energy Fund) both delivered strong YTD returns in 2026 (+46.3% and +50.1% respectively), partly driven by the growing global energy demand tied to AI infrastructure. Natural gas — Canada’s most exportable energy commodity — is becoming the fuel of choice for data center operators seeking reliable, cost-effective baseload power.
Pillar 4 — Broad AI Infrastructure: Capturing the Full Stack
For investors who want exposure across all layers of the AI buildout — chips, software, infrastructure, and services — these broader AI-focused ETFs offer a single-ticket solution.
DTAI — Global X Artificial Intelligence ETF
DTAI holds companies across the full AI value chain: chip designers, cloud providers, enterprise software companies, and AI application developers. It is a broader bet on AI as a technology platform rather than a specific layer of the infrastructure stack.
AIQ — Global X Artificial Intelligence & Technology ETF
AIQ combines AI infrastructure companies with the enterprise software and cloud businesses that monetize AI capabilities. With significant holdings in Microsoft, Nvidia, Alphabet, and Meta alongside less well-known infrastructure players, it provides a balanced exposure to both the buildout and the monetization of AI.
AIPO — Defiance AI & Power Infrastructure ETF
AIPO is one of the most interesting recent additions to the AI ETF landscape. It explicitly combines AI infrastructure companies with power generation and grid infrastructure holdings — effectively building a single ETF around the thesis that AI and power demand are inseparable. With a +43.5% YTD return in 2026 and a 0.69% expense ratio, it is a compelling single-ticker solution for investors who want to capture both the compute and the power dimensions of the AI boom.
How to Build a Portfolio Around the AI Data Center Theme
A well-constructed AI infrastructure portfolio does not need to hold all of the above. Here is how to think about allocation depending on your objective.
For a concentrated high-conviction AI infrastructure position, SMH or SOXX for semiconductors, DTCR for data center real estate, and either GRID or AIPO for power infrastructure gives you coverage across all three physical layers of the buildout. Canadian investors can replace SMH with XCHP to eliminate currency and withholding tax friction in a TFSA.
For a broader thematic position with less concentration risk, a broad AI ETF like AIQ combined with XEG or NNRG for Canadian energy exposure gives you participation in the AI infrastructure boom with more diversification across the technology sector.
The most important principle: this is a multi-year, multi-decade theme. The AI infrastructure supercycle is not a trade — it is a structural shift in how computing power is built, located, and powered. Short-term volatility in any of these ETFs should be evaluated against a 5 to 10 year investment horizon, not a quarterly earnings cycle.
Key Risks to Monitor
The AI infrastructure theme is compelling but not without risk. Valuation compression remains a real concern — many semiconductor and AI-adjacent ETFs trade at elevated multiples that leave limited room for disappointment. Power infrastructure permitting and grid connection timelines are proving longer than expected in many markets, which could delay data center buildout schedules. Geopolitical risk around TSMC and Taiwan remains a structural vulnerability for the semiconductor supply chain. And the concentration of hyperscaler capex — which drives the data center REIT thesis — means that any slowdown in spending from Microsoft, Amazon, or Google would ripple quickly across the entire infrastructure ecosystem.
Conclusion
The AI data center boom is not a software story. It is a physical infrastructure story — and physical infrastructure requires chips, real estate, and electricity. The ETFs covered in this guide give investors exposure to all three layers of this buildout, in both U.S. and Canadian markets.
Whether you are a TFSA investor in Quebec looking for semiconductor exposure without currency friction, or a U.S. investor building a concentrated AI infrastructure position, the ETF ecosystem has never offered better tools to participate in this theme at low cost and with genuine diversification.
The supercycle is underway. The question is not whether to have exposure — it is how to structure it intelligently.