Investing · Commodities •
AI needs copper and uranium. Australia has both.
The AI build-out ends in a mine. Here’s what data centres actually need from copper and uranium — and why most Australian portfolios already own the supply side.
Book a free consultation
We've already followed the money behind the AI build-out — how much is being spent and who's paying for it. This article asks a different question: what is all that money actually buying, once you get past the buildings and the chips?
The answer, further down the chain than most coverage goes, is metal and fuel. And Australia happens to sit on a great deal of both. It's the other half of AI infrastructure investing in Australia — the supply side, not the demand side.
- Most of the copper in a data centre isn't in the servers. It's in the transformers, switchgear and cable that carry power to them.
- The IEA puts data-centre copper demand at 512,000 tonnes by 2030 — real, but under 2% of a global copper market it forecasts at over 31 million tonnes.
- Australia holds around 28% of the world's identified uranium but produced only about 7.6% of it in 2024 — and is legally barred from using any of it here.
- If you own Australian shares, you already own the supply side of this. The question isn't whether to get exposure. It's whether you already have more than you meant to.
Why does the AI build-out end in a mine?
Follow any supply chain far enough and it stops being about technology. Chips are made of refined materials. Data centres are made of concrete and steel. And the electricity that runs them has to be generated somewhere, then carried to a specific patch of ground through a great deal of metal.
Two commodities sit at the end of that chain in a way that's easy to miss. Copper, because it is what electricity travels through. Uranium, because a growing share of the round-the-clock power these facilities want is expected to come from nuclear generation.
Neither depends on picking which AI company wins. Both get consumed whoever does. That's the appeal — and, as the rest of this article sets out, it is not the same thing as being lower risk.
Where does the copper actually go?
Not where most people picture it. Very little copper sits inside the server racks themselves.
It's in the electrical spine: the transformers that step voltage down, the switchgear that distributes it, the busbars that carry it across the floor, and long runs of distribution cable. Then, outside the fence, the substations and transmission lines built specifically to serve the site.
That inside-the-fence versus outside-the-fence distinction sounds like a technicality. It turns out to be the whole reason the published forecasts don't agree.
How much copper is it, in context?
The International Energy Agency produced the first detailed estimates of what data centres will need from critical minerals. Their figure for 2030 is 512,000 tonnes of copper, alongside 75,000 tonnes of silicon.
That is a real number, and it is smaller than the coverage often implies. The same IEA modelling puts total global copper demand at just over 31 million tonnes in 2030. So data centres land at under 2% of the market.
Source: International Energy Agency, Energy and AI, which provides the first detailed estimates of data-centre demand for critical minerals, and IEA global copper demand projections. Forecasts, not results. Current as at August 2026.
The wider market is growing too, which is the more important backdrop. The IEA has global copper demand rising from around 25.9 million tonnes in 2023 to just over 31 million by 2030 and above 36 million by 2040.
Source: International Energy Agency projections for total global copper demand, reported 2024–2026. The IEA also expects annual copper demand for electricity grids to double by 2040. Forecasts, not results. Current as at August 2026.
One caution about any copper figure you see quoted. Estimates differ substantially between analysts, and the reason is definitional rather than a disagreement about direction. Counting only the copper inside the building gives one number. Counting the substations and transmission built because of that building gives a much larger one. When a figure is quoted to you without saying which it is, it is worth asking.
The proportions matter as well. Copper's largest growth drivers remain electric vehicles, solar and wind, and the replacement of ageing grids. The data-centre build-out adds pressure to a market that was already tight. It did not create the tightness, and any account that presents AI as the main driver of copper demand is overstating it.
Uranium and the nuclear numbers
The uranium link is simpler to state and harder to pin down. Data centres want power that runs at three in the morning regardless of weather, and nuclear is one of the few low-carbon sources that supplies it at scale. That has pushed operators toward long-term nuclear contracts.
Whether the reactors get built is a separate question, and the institutions that model it do not agree either. Global capacity today is around 397 gigawatts across roughly 440 reactors.
Sources: current capacity of about 397 GWe across roughly 440 reactors and the 1,446 GWe figure from the World Nuclear Association, World Nuclear Outlook Report 2025; 992 GWe high case from the IAEA, Energy, Electricity and Nuclear Power Estimates to 2050; 728 GWe from the International Energy Agency's World Energy Outlook 2025 existing-policies scenario. The three differ because of their assumptions, not their underlying data. Current as at August 2026.
The highest figure is roughly double the lowest, and the World Nuclear Association is candid about what its own number would require: building around 65 gigawatts a year between 2046 and 2050, roughly double the peak rate achieved in the 1980s, with about 542 gigawatts of the target not yet backed by identified projects.
Take that as a direction rather than a forecast. The near-term reactor construction underneath it is real. The 2050 destination is an ambition with a steep curve attached.
Australia's oddly shaped position
Australia occupies a strange spot in all of this, and the facts are worth knowing before forming a view.
- We hold a very large share of the resource. Around 28% of the world's identified uranium resources sit in Australia — by some measures closer to a third.
- We produce far less than that share. Australia was the fourth-largest producer in 2024, at roughly 7.6% of global output.
- We use none of it. Commonwealth legislation does not permit a domestic nuclear power industry, so all Australian uranium is exported — in 2024, mainly to the United States, Canada and Europe.
- Mining happens in one state. Active uranium mining is confined to South Australia. Western Australia, Queensland, New South Wales and Victoria each have restrictions or bans of various kinds.
There's a neat detail buried in that. Australia's largest uranium asset, Olympic Dam in South Australia, is principally a copper mine — uranium and gold come out as by-products. Both commodities in this article come out of the same hole in the ground, which says something about how entangled these supply chains are.
Whether Australia should change its domestic nuclear settings is a live political debate and not one for a financial planning article. The point for an investor is narrower: Australia's exposure to a global nuclear build is as a supplier rather than a user, and that supply is shaped by a patchwork of state rules that can change with a state election.
Sources: Australia’s uranium resource share, production ranking, export destinations and state-by-state position — Geoscience Australia, Australian Energy Commodity Resources 2026 and World Nuclear Association, Australia’s Uranium. Resource figures as at end-2024. Current as at August 2026.
You may already own the supply side
This is where the two halves of the story meet, and it's the part most people skip — because for most Australians, this is what AI infrastructure investing in Australia already looks like, whether they chose it or not.
Australian portfolios are unusually concentrated in resources. The ASX is weighted heavily toward mining and energy, so if you hold Australian shares — and virtually every balanced or growth super option does — you already own a slice of copper and uranium production, sitting alongside iron ore, gold and energy.
Most Australians are running a significant commodities tilt they never chose. Adding a specialist copper or uranium holding on top doesn't create exposure you lacked. It concentrates exposure you already had, in the same sector, the same market, the same currency and the same economy that pays your wage and prices your house.
- Look before you add. Check the Australian shares allocation and largest holdings in your super option first.
- Ask what it displaces. Money going into a commodity position comes out of something else. That trade-off is the actual decision.
- Watch the home bias. Australia is roughly 2% of world share markets. Most Australian portfolios hold far more than that, and resources is where the extra tends to sit.
Why commodities behave differently
Commodity investments have a shape of their own, and it isn't the same shape as the companies buying what they produce.
Prices are set by supply as much as demand. A high price invites new mines, new mines eventually add supply, and supply brings the price back down — often just as the demand story is being written up most confidently. Meanwhile the companies themselves carry construction risk: a mine that runs over budget or behind schedule can disappoint badly while the metal it produces does exactly what everyone predicted.
Analysts are openly split on where copper goes next. Some see meaningful supply deficits arriving soon. Others argue recent strength has run ahead of the fundamentals and no real shortage emerges until later in the decade. Both views are held by serious people, which is itself worth knowing.
That volatility is survivable at 40. It's a different proposition at 62, because a sharp fall in the years either side of your retirement date does lasting damage — you start drawing from a smaller balance, and the money withdrawn never recovers. That's sequencing risk, and a concentrated single-sector position is one of the more reliable ways to make it worse. It's something we work through directly in our superannuation and retirement planning work.
Frequently asked questions
Why do AI data centres need so much copper?
Why do different analysts give such different copper numbers?
Does Australia have nuclear power?
How much nuclear capacity will actually be built by 2050?
Why do data centres want nuclear power specifically?
What are small modular reactors, and do they change the picture?
Where is uranium mined in Australia?
Does my super already own copper and uranium miners?
Will the AI build-out push copper and uranium prices up?
What's the difference between investing in a commodity and investing in a miner?
We'll show you what you already own, where it's concentrated, and whether that mix still suits the retirement you're planning.
Book a free consultation
With 17+ years’ experience and over 1,000 retirement plans built for Australian families, Manny works with clients aged 50 to 65 across Bundoora, metropolitan Melbourne, and nationally via video consultation. His focus is helping pre-retirees replace uncertainty with a clear, evidence-based plan.
+61 433 564 003 · manny@planmywealth.com.au · Book a free consultation
Sources & further reading
- Australian uranium resources, production and export destinations — Geoscience Australia, Australian Energy Commodity Resources 2026
- Australia's share of world uranium resources and production ranking — World Nuclear Association, Australia's Uranium
- Nuclear capacity to 2050 and the build rate required — World Nuclear Association, World Nuclear Outlook Report 2025 and IAEA, Energy, Electricity and Nuclear Power Estimates to 2050
- Critical mineral requirements of data centres — including data-centre copper demand of 512 kt in 2030 and total global copper demand projections — International Energy Agency, Energy and AI and the IEA Global Critical Minerals Outlook
- Nuclear's role in data-centre power supply and small modular reactor plans — International Energy Agency, Energy supply for AI
