You can't buy NVIDIA shares directly on the ASX. You can't buy TSMC. You can't buy ASML. You can't buy OpenAI at all — it's still private. So how does an Australian investor actually get exposure to the AI boom?
The answer is ETFs — and there are more options than most people realise. 17 ASX-listed ETFs give you direct or indirect exposure to the AI value chain, spanning everything from broad US tech to specific bets on chips, uranium, copper miners, and humanoid robotics.
This guide breaks down each of the six layers of the AI value chain, shows the May 2026 performance data, and gives you three concrete portfolio approaches depending on how much conviction you want to express.
Disclaimer: No fund manager wrote this article. No issuer is paying for placement. This is general information only and does not constitute financial advice. Thematic ETFs are higher-volatility positions and should be considered satellite holdings, not core portfolio building blocks. Always consider your personal circumstances and consult a licensed financial adviser before investing.
The lazy take vs the real take
The lazy take on AI ETFs goes like this: "Buy NDQ, it's the Nasdaq, it's full of AI stocks, done."
That's not wrong. NDQ is the largest AI-adjacent ETF on the ASX at $8.96 billion as at May 2026, charging 0.48% MER, with roughly half its holdings in AI-leveraged megacaps (Microsoft, Alphabet, NVIDIA, Meta, Apple, Amazon). Over 3 years it returned +88.8% — solid, but not the standout.
The real take is this: the AI trade over the past 3 years wasn't "buy more tech." It was "buy what tech needs."
Building AI doesn't just take software. It takes:
Chips to run the models
Robots as the physical embodiment of AI
Power to feed the chips (and lots of it)
Copper to wire the data centres
Uranium and gas to generate the power
Critical minerals for batteries and storage
Cybersecurity and defence to defend the stack
If you only own NDQ, you're betting on the customers of the AI build-out. The companies actually building it — and the commodities they consume — sit in completely different ETFs.
The AI value chain in six layers
Every dollar invested in AI flows through this chain. Here's how each layer maps to ASX-listed ETFs.
You can browse the full sortable list on our AI & Robotics ETF category page (note: typo in slug — kept for legacy SEO), and compare any two side-by-side with the Compare ETFs tool.
Layer 1 — AI software & models
The companies actually building AI products.
Ticker | Name | MER | AUM (May 2026) | 1Y | 3Y |
|---|---|---|---|---|---|
BetaShares Nasdaq 100 | 0.48% | $8.96B | +27.2% | +88.8% | |
BetaShares Nasdaq 100 Hedged | 0.51% | $973M | +39.5% | +100.6% | |
Global X Artificial Intelligence | 0.57% | $266M | +48.1% | n/a (new) | |
Global X AI Infrastructure | 0.57% | $144M | +64.3% | n/a (new) |
The takeaway: NDQ is the broad option. GXAI and AINF are pure-play AI but too new for 3-year track record. HNDQ outperformed NDQ by 12pp over 1 year because of AUD recovery — see our hedged vs unhedged guide for more.
Layer 2 — Chips & semiconductors
The brains of every model. Without chips, there is no AI.
Ticker | Name | MER | AUM | 1Y | 3Y |
|---|---|---|---|---|---|
Global X Semiconductor | 0.45% | $1.01B | +151.7% | +269.7% | |
Global X Ultra Long Nasdaq 100 (2× geared) | 1.00% | $93M | +93.7% | +194.5% |
SEMI is the standout ETF of the past 3 years on the ASX — returning +269.7% over 3 years to May 2026, holding NVIDIA, TSMC, ASML, AMD and Broadcom. At 0.45% MER it's also one of the cheaper thematic options. LNAS is 2× leveraged Nasdaq — amplifies both upside and downside, suitable only for experienced investors who understand daily reset and volatility decay.
Layer 3 — Robotics & automation
AI in physical form.
Ticker | Name | MER | AUM | 1Y | 3Y |
|---|---|---|---|---|---|
Global X Robotics & Automation | 0.69% | $344M | +41.2% | +48.1% | |
BetaShares Global Robotics & AI | 0.57% | $329M | +15.9% | +32.1% | |
Global X Humanoid Robotics | 0.57% | $3M | n/a (new) | n/a (new) |
ROBO and RBTZ track similar themes — global robotics and industrial automation — with ROBO winning over 3 years. HMND is the most concentrated humanoid-robotics fund (Tesla Optimus, Figure, etc.) but listed in March 2026 with only $3M AUM, so it's a very early-stage option.
Layer 4 — Power & energy
This is the layer most retail investors miss. AI data centres consume staggering amounts of electricity. Goldman Sachs estimates US data centre power demand will grow 165% by 2030. That power has to come from somewhere — and the past 3 years show exactly where the money flowed.
Ticker | Name | MER | AUM | 1Y | 3Y |
|---|---|---|---|---|---|
Global X Uranium | 0.69% | $148M | +47.4% | +160.8% | |
Betashares Global Energy (LNG/Oil) | 0.57% | $259M | +42.6% | +56.6% |
The uranium thesis has been the standout AI power play. Microsoft, Amazon, Google and Oracle have all signed nuclear power deals to feed their data centres. ATOM returned +160.8% over 3 years — second only to SEMI among the AI value chain. Nuclear is back on the agenda specifically because data centres need 24/7 baseload power that wind and solar cannot reliably provide.
FUEL gives you exposure to LNG and natural gas — the second-most-favoured energy source for new AI data centres because gas turbines can be built faster than nuclear plants.
For the full energy theme breakdown, read Commodity & Resource ETFs on the ASX.
Layer 5 — Copper, critical minerals & batteries
The grid build-out and storage. Every gigawatt of AI demand needs new transmission, new substations, new cooling, new batteries. Copper is the main material in all of it.
Ticker | Name | MER | AUM | 1Y | 3Y |
|---|---|---|---|---|---|
Global X Copper Miners | 0.65% | $879M | +95.0% | +139.6% | |
Betashares Energy Transition Metals | 0.69% | $141M | +130.9% | +115.4% | |
Global X Battery Tech & Lithium | 0.69% | $872M | +108.1% | +89.0% |
WIRE returned +139.6% over 3 years and +95% in the past year alone, driven entirely by the AI-grid story. XMET (which holds copper, lithium, nickel, rare earths) just delivered the highest 1-year return in this entire guide at +130.9%. ACDC covers battery technology + lithium miners — both essential for grid storage as AI demand grows.
Browse the full set on the critical minerals ETF category page.
Layer 6 — Cybersecurity & defence
AI as national security. Every major government is now treating AI as a strategic asset.
Ticker | Name | MER | AUM | 1Y | 3Y |
|---|---|---|---|---|---|
BetaShares Global Cybersecurity | 0.67% | $1.46B | +5.6% | +73.4% | |
Global X Defence Tech | 0.50% | $131M | +8.0% | n/a (listed 2024) |
HACK is the largest cybersecurity ETF on the ASX but has had a slower year. DTEC includes defence-tech disruptors like Palantir alongside the traditional primes — covered in detail in our defence ETF guide.
The 3-year leaderboard — what the data actually shows
13 AI-investment ETFs with at least 3 years of track record, sorted by cumulative AUD total return to 31 May 2026.
Rank | Ticker | Category | 3Y Return |
|---|---|---|---|
1 | SEMI | Chips · Semiconductors | +269.7% |
2 | CRYP | Crypto + AI compute | +220.7% |
3 | LNAS | Geared Nasdaq 100 (2×) | +194.5% |
4 | ATOM | Uranium | +160.8% |
5 | WIRE | Copper miners | +139.6% |
6 | XMET | Energy transition metals | +115.4% |
7 | HNDQ | Nasdaq 100 hedged | +100.6% |
8 | ACDC | Battery tech + lithium | +89.0% |
9 | NDQ | Nasdaq 100 | +88.8% |
10 | HACK | Cybersecurity | +73.4% |
11 | FUEL | Global energy / LNG | +56.6% |
12 | ROBO | Robotics + automation | +48.1% |
13 | RBTZ | Robotics + AI | +32.1% |
Eight of the top 13 ETFs beat the broad Nasdaq 100 (NDQ +88.8%). Picking the right thematic exposure was decisive over the past three years.
The S&P 500 (IVV) returned +67.9% over 3 years — meaning nine of the 13 AI-investment ETFs beat the broad US market, and most of the outperformance came from non-software layers: chips, uranium, copper.
Note on CRYP — Betashares Crypto Innovators returned +220.7% partly because it holds many of the same GPU/compute infrastructure companies that benefit from AI compute demand (NVIDIA, Coinbase, MicroStrategy, Bitcoin miners). The crypto-AI overlap is meaningful.
How to build AI exposure — three portfolio approaches
Each approach trades off simplicity, conviction, and risk. Pick the one that matches you.
Approach 1 — Simple (one ETF)
100% NDQ
If you don't want to think about it, NDQ gives you broad AI-adjacent exposure in a single trade. About half the fund is in AI-leveraged megacaps. It's the largest tech ETF on the ASX, the most liquid, and the simplest expression of "buy AI."
Who it suits: Beginners, investors who already have a diversified core (VAS + VGS) and just want a tech tilt, anyone who doesn't want to manage multiple positions.
Track record: 1Y +27.2%, 3Y +88.8%, 5Y +126.5%
For more on whether to add Nasdaq exposure on top of a global ETF, see IVV vs VGS vs VTS.
Approach 2 — Balanced (three ETFs)
Allocation | ETF | Layer |
|---|---|---|
40% | NDQ | Software + broad tech |
30% | SEMI | Chips |
30% | ATOM | Power (uranium) |
This is a three-fund expression of the full AI thesis — tech + chips + power. It covers the three layers that have driven returns most strongly over the past 3 years, without going too deep into satellite picks.
Who it suits: Investors with a strong conviction in AI who want diversified exposure across the value chain, willing to manage three positions, and comfortable with thematic volatility.
Sizing guidance: Keep this thematic allocation to 10–15% of your total portfolio. The rest should be in broad diversified core ETFs like VAS, VGS, IVV, or DHHF.
Approach 3 — Full stack (six ETFs)
Allocation | ETF | Layer |
|---|---|---|
25% | NDQ | Broad tech |
20% | SEMI | Chips |
15% | CRYP | AI compute / crypto |
15% | ATOM | Uranium / power |
15% | WIRE | Copper miners |
10% | HACK or DTEC | Cyber / defence |
The maximum conviction expression — full coverage across all six layers of the AI value chain. Higher complexity, higher volatility, but if AI is going to keep driving markets for the next decade, this captures it.
Who it suits: Highest-conviction AI investors comfortable with thematic concentration. Most appropriate as a small slice (under 20%) of total portfolio with the rest in diversified core funds.
Critical caveat: This is a satellite allocation, not a core portfolio. For the full satellite framework, see How to Build Your Satellite Portfolio with ETFs.
What the past 3 years tell us
A few patterns worth noting before you commit capital.
1. The non-tech layers won. Chips (SEMI), uranium (ATOM), and copper (WIRE) all beat the broad Nasdaq over 3 years. If the AI build-out continues, the commodities and infrastructure layers may keep winning.
2. Pure-play AI software ETFs are too new to assess. GXAI launched April 2024, AINF launched April 2025, HMND launched March 2026. Returns in the first 12-24 months of a thematic ETF are very noisy and shouldn't drive your decision.
3. Cybersecurity has been a relative disappointment. HACK at +73.4% over 3 years is solid in absolute terms, but lags every other AI layer — sector consolidation and competitive pricing have crimped underlying company margins. The thesis is still intact long-term, but execution has been mixed.
4. Australian resources lagged global commodity peers. MVR, QRE and OZR (Australian miners) returned 28-33% over 3 years vs WIRE's 140% and XMET's 115%. The global pure-play funds outperformed the broad AU resources baskets meaningfully.
5. Hedging mattered. HNDQ (hedged) returned +100.6% over 3 years vs NDQ (unhedged) at +88.8%. With the AUD recovering, hedged tech has outperformed — covered in detail in our hedged vs unhedged guide.
Risks to understand before investing
Concentration risk. Thematic ETFs hold 20–80 stocks in a single industry. When the theme is in favour, returns can be spectacular. When sentiment turns, drawdowns can be brutal. SEMI fell ~40% in 2022 before its current rally.
Theme timing risk. Most thematic ETFs launch after the theme has already moved meaningfully. SEMI launched in August 2021, ATOM in December 2022, WIRE in November 2022 — all after the early moves in their underlying themes. That's not a reason to avoid them, but it means the easiest gains were before retail could access them through an ASX wrapper.
Higher fees. Thematic ETFs charge 0.45–0.69% MER vs 0.04–0.18% for broad market funds. The fee gap is justified only if the theme outperforms by more than the difference. So far for SEMI, ATOM and WIRE, it has — but that's no guarantee for the future.
Currency risk. All the international thematic ETFs above are unhedged unless specified (HNDQ is hedged). When the AUD rises against the USD, your returns suffer. The AUD has recovered from US$0.62 to US$0.70 over the past 18 months — a headwind for unhedged positions.
Sequence risk. If you hold concentrated thematic positions during retirement, a 40-50% drawdown in your AI sleeve at the wrong time could force you to sell at losses. Thematic ETFs are better suited to accumulators with long time horizons.
The bottom line
You cannot buy NVIDIA on the ASX directly, but you can absolutely access the AI boom through ASX-listed ETFs — across six distinct layers and 17 different funds.
The trade of the past 3 years was not "buy more tech." It was "buy what tech needs." Chips, uranium and copper miners all beat the broad Nasdaq. The same dynamic may or may not continue — but the data is clear: NDQ alone is missing most of the story.
Three concrete approaches:
Simple — 100% NDQ. Done in one trade.
Balanced — 40% NDQ, 30% SEMI, 30% ATOM. Full thesis in three ETFs.
Full stack — 6 ETFs covering all six layers of the value chain.
Whichever approach you take, size it as a satellite — not a core. Thematic exposure should typically sit at 10–20% of your total portfolio, with the rest in diversified core funds like VAS, VGS, IVV or DHHF.
For broader context on building portfolios with ETFs, see Best ETFs for Beginners 2026 and How to Build Your Portfolio from Scratch.
Frequently asked questions
What's the best AI ETF in Australia?
For broad AI/tech exposure, NDQ (BetaShares Nasdaq 100) is the largest at $8.96 billion as at May 2026 and 0.48% MER — returned +88.8% over 3 years. For chips specifically, SEMI at 0.45% returned +269.7% over 3 years — the best-performing ETF in this guide. For pure-play AI software, GXAI at 0.57% — but it's too new for a 3-year track record.
Are there pure-play AI ETFs on the ASX?
Yes — GXAI (Global X Artificial Intelligence), AINF (Global X AI Infrastructure), and the robotics funds ROBO, RBTZ and HMND. All invest in companies whose primary business is AI development, AI infrastructure, or robotics.
Why are uranium ETFs an AI play?
AI data centres consume vast amounts of electricity — far more than wind or solar can reliably provide for 24/7 workloads. Tech giants (Microsoft, Amazon, Google, Oracle) have all signed nuclear power deals to feed their data centres. Uranium fuels nuclear reactors. ATOM returned +160.8% over 3 years as this thesis played out.
What's the best semiconductor ETF on the ASX?
SEMI (Global X Semiconductor) at 0.45% MER and $1.01B AUM is the dedicated chip ETF — returned +269.7% over 3 years as AI chip demand exploded. For broader chip exposure within mega-cap tech, NDQ at 0.48% includes ~25% chip-stock weighting.
What's the best copper ETF on the ASX?
WIRE (Global X Copper Miners) at 0.65% MER and $879M AUM is the dedicated copper play — up +139.6% over 3 years as AI grid demand drove copper prices. For broader Australian mining exposure including copper, MVR, QRE and OZR all hold significant copper positions.
How much of my portfolio should be in AI ETFs?
Thematic ETFs are higher-volatility satellite positions, not core holdings. The right framework: own a broad-market core (VAS, IVV, VGS, or a diversified fund like DHHF) for 70–80% of your portfolio, then layer thematic plays like AI/uranium/cyber on top for 10–30%. The wrong framework: making thematics your whole portfolio.
Is it too late to invest in AI?
The early gains in AI have already happened — SEMI is up 270% in 3 years. But the AI build-out is a multi-decade structural shift, and several layers (power, copper, robotics, defence) are arguably still in early innings. The right question isn't "is it too late?" but "what's an appropriate sizing given how much has already moved?" Most investors should keep total thematic exposure below 20% of portfolio.
What about Australian AI stocks?
There are some Australian-listed AI-adjacent companies (TechnologyOne, NEXTDC for data centres, Pro Medicus for medical AI), but the Australian AI universe is small compared to the US. Most pure-play AI exposure comes through international ETFs. For Australian shares context, see Every Australian Shares ETF on the ASX.
Related reading
Sources: ReviewETF.com.au, CBOE Australia (performance to 31 May 2026), Goldman Sachs research (data centre power demand projections), Bloomberg, issuer factsheets (Global X, BetaShares, VanEck). All MER and AUM figures verified as at 31 May 2026. Returns shown are cumulative AUD total return. Past performance is not indicative of future returns. Thematic ETFs are higher-volatility positions and should be considered satellite holdings, not core portfolio building blocks. General information only — this is not financial advice.
