Shares · Thematic investing
Fusion energy and the AI power crunch: what Australian investors can actually buy
Fusion finally has customers, contracts and a listed share price. It still has no commercial product, and the best companies in the field are private. Here is the honest map of the exposure available from Australia — and how to size it.
A story that sounds a little too neat
The pitch is almost too neat. Artificial intelligence needs enormous quantities of clean, uninterrupted electricity. Fusion promises exactly that, from fuel that is effectively unlimited, with no long-lived radioactive waste. Two of the largest technology companies on earth have already signed contracts to buy it.
The catch is equally neat: no fusion plant anywhere has ever delivered a single watt of electricity to a grid, and the companies closest to doing so are almost all privately held. What retail investors can buy is a much narrower, stranger set of things than the headlines suggest.
This piece maps what is genuinely investable from Australia, what timeframe the thesis realistically runs on, where the money is most likely to be lost, and how a speculative position like this is usually sized.
Four things have moved since this piece first ran. Australia’s capital gains tax change is now law, not pending legislation — it received Royal Assent on 26 June 2026. Almonty (ASX: AII) is leaving the ASX, with its CDIs suspended from 28 August 2026 and de-listing on 1 September. General Fusion has given its first business update as a listed company, and disclosed that its predecessor SPAC’s March 2026 accounts will be restated. And the data centre demand figures have been brought up to the 2026 Berkeley Lab and IEA numbers. Each is covered in the relevant section below.
The demand side is not speculative
Whatever you conclude about fusion, the power problem it is being sold against is real and already measurable. Data centres consumed 4.7% of all electricity generated in the United States in 2024 — 192 terawatt hours — and the International Energy Agency expects global data centre demand to roughly double from 485 TWh in 2025 to about 950 TWh by 2030, or about 3% of world electricity. That is approximately the entire electricity consumption of Japan, added to world demand inside a decade.
In the United States specifically, data centres are projected to draw 649 TWh a year by 2030 — 11.8% of national electricity use, on the Berkeley Lab study published for the US Department of Energy in June 2026, within a range running as high as 843 TWh. Meeting that has already pushed hyperscalers into restarting retired fission plants, contracting geothermal, and signing agreements for generation technologies that do not yet exist.
This is the part of the thesis that does not depend on a scientific breakthrough. Demand for firm, carbon-free, round-the-clock power is growing faster than the grid can supply it, and buyers are willing to pay a premium and sign contracts a decade in advance. The open question is entirely on the supply side.
What actually changed: fusion got customers
Fusion has been perpetually twenty years away for seventy years. The genuine shift over the past three years is not a physics result — it is commercial. Sophisticated counterparties have done due diligence on fusion roadmaps and signed binding contracts against them.
- Microsoft – Helion Energy. The world's first fusion power purchase agreement: 50 MW or more for a data centre in central Washington, with the plant due online in 2028 and full output after a roughly one-year ramp, penalties attached. Helion has raised roughly US$1.5 billion, including US$465 million in June 2026.
- Google – Commonwealth Fusion Systems. 200 MW from the planned ARC plant in Chesterfield County, Virginia, in the early 2030s.
- Eni – Commonwealth Fusion Systems. A power purchase agreement reported at around US$1 billion.
A power purchase agreement with delivery dates and penalty clauses is a stronger signal than a funding round. A venture round can reflect narrative; an offtake contract means a counterparty's procurement and legal teams concluded the timeline was credible enough to underwrite. That said, five signed agreements across an entire global industry is a thin base to build an investment case on.
“The most consequential development in the fusion industry is not a physics breakthrough — it is a commercial one.”
Where the technology actually stands in 2026
The nearest thing to a verdict on fusion's commercial future arrives in 2027, at a facility in Devens, Massachusetts. Commonwealth Fusion Systems — an MIT spin-out that raised a further US$1 billion in July 2026 to reach about US$4 billion, close to 30% of all private capital ever committed to fusion — is assembling SPARC, a compact high-field tokamak designed to produce more fusion energy than is put into the plasma.
As at mid-2026 SPARC is roughly 80% assembled, with its high-temperature superconducting magnets going in and the vacuum vessel being prepared. First plasma and the attempt at net energy gain are targeted for 2027 — already a slip from an original 2025 goal. If SPARC succeeds, it would be the first time a privately built machine has crossed scientific breakeven, and it would revalue the entire sector. If it fails or slips badly, the opposite applies.
The problem: the best companies are not for sale
Here is the structural issue that most fusion investing content skips. Rank the field by capital raised, technical progress and contracted customers, and the leaders are Commonwealth Fusion Systems, Helion, Pacific Fusion, Proxima Fusion and Tokamak Energy. Every one of them is private. None has filed to list. CFS just raised a billion dollars from pension funds, which is precisely what a company does when it does not need retail equity markets.
That creates an adverse selection problem worth stating plainly: in a sector where the strongest players can raise privately at will, the companies that arrive on public markets are disproportionately those that could not. General Fusion completed its listing after heavy redemptions from the SPAC vehicle. TAE, according to reporting in The Wall Street Journal, was facing a funding crunch when it found its merger partner.
This does not make either company a bad investment. It does mean the listed fusion universe is not a representative sample of the fusion industry, and buying it is not the same as buying the theme.
What you can actually buy from Australia
On the ASX
There is no ASX-listed pure-play fusion company. Australia's only home-grown fusion business, Sydney-based HB11 Energy, which is pursuing laser-driven hydrogen-boron fusion, is private and has raised only around A$10 million.
The closest ASX-listed exposure is an ETF and some supply-chain miners:
| Code | What it is | Honest fusion linkage |
|---|---|---|
| ASX: URAN | VanEck Uranium and Energy Innovation ETF — global uranium and nuclear infrastructure | Direct but small. Its index rules explicitly permit companies commercialising nuclear fusion technology. In practice fusion is a minor sleeve inside a mostly-fission, mostly-uranium fund. |
| ASX: EQR | EQ Resources — tungsten producer, Mt Carbine (Qld) and Barruecopardo (Spain) | Indirect and coincidental. Fusion reactors need tungsten — one 500 MWth plant needs over 4,200 tonnes — but EQR's re-rating is driven by Chinese export controls and defence demand, not fusion. Andrew Forrest took a 16.8% stake in July 2026 and the company posted record July revenue of A$51m; shares are up more than 800% in a year. |
| ASX: AII | Almonty Industries — tungsten, multi-jurisdiction | Indirect and coincidental. Same critical-minerals thesis as EQR. Owning it for fusion reasons is telling yourself a story about a defence and supply-security trade. It is also leaving the ASX: CDIs are suspended from 28 August 2026 and the de-listing takes effect 1 September 2026, so holders need to convert 1:1 to the Nasdaq line (ALM) or sell before then. |
If you buy a tungsten miner because of fusion, you have bought a tungsten miner. It will trade on Chinese export policy, defence budgets and the APT price — and it will do that for the entire decade before fusion is a material customer. That may still be a good investment. It is not a fusion investment.
On US markets
Australian investors can access these through CHESS-sponsored international brokers or global platforms. This is where the only genuine pure-play sits — along with the sector's two most obvious traps.
| Code | What it is | What you are taking on |
|---|---|---|
| NASDAQ: GFUZ | General Fusion — the only listed pure-play. Vancouver-based, magnetized target fusion, LM26 demonstration machine, founded 2002 | Pre-revenue, roughly US$24m annual cash burn, small cap; trading around US$7 in late August 2026, some 53% below its debut peak. Listed via SPAC on 13 July 2026 after heavy redemptions. Its first business update as a listed company, on 18 August 2026, reported plasma heating of about 0.72 keV on LM26 and a milestone-based framework agreement with Italy’s Renexia. The same filing disclosed that the SPAC’s March 2026 quarterly accounts should no longer be relied upon and will be restated — non-cash, but it flags a possible material weakness in internal controls. First earnings are expected 11 September 2026. Near-certain future dilution. |
| NASDAQ: DJT | Trump Media & Technology Group — agreed all-stock merger with fusion company TAE Technologies, valued above US$6bn | Deal announced December 2025 and still not closed: no S-4 registration statement has been filed and no shareholder vote scheduled, with completion now targeted for Q4 2026. The US$6bn headline is a December 2025 all-stock figure and DJT has fallen roughly 45% since, so the implied consideration is materially lower. Until it closes you own a loss-making social media and bitcoin business with a fusion option attached. Political and conflict-of-interest risk on top of everything else. |
| NASDAQ: AMSC | American Superconductor — high-temperature superconducting wire | Real revenue from grid, naval and wind applications. Fusion is not yet a material revenue line, so this is a picks-and-shovels position with a fusion call option, not a fusion bet. |
| NYSE: COHR | Coherent — industrial lasers, including the excimer lasers used to deposit HTS tape | A diversified photonics business. Fusion supply chain is a rounding error against its datacom and semiconductor segments. |
| TYO: 5803 TYO: 5801 |
Fujikura and Furukawa Electric — two of the world's leading REBCO superconducting tape makers | Arguably the truest bottleneck exposure — SPARC alone needs around 10,000 km of superconducting wire. Both sit inside large diversified Japanese conglomerates, and Tokyo access is awkward for most Australian retail brokers. |
What the first listed fusion stock did in its first month
General Fusion's listing is the closest thing to a live experiment in how public markets price pre-revenue fusion. It is not reassuring.
That chart is the single most useful thing in this article. The science moves on a decade timescale; the share price moves on sentiment, index flows and lock-up expiries. Anyone buying listed fusion needs to be able to tolerate that gap without selling at the bottom of it.
How to spot the fakes
Every genuine technology theme attracts shell companies that rename themselves to match. Fusion is now firmly in that phase. One example currently trading over the counter in the US: a company that was a dormant biofuels shell until February 2026, when it reverse-merged with a fusion start-up, renamed itself, and now describes a fusion platform. Its most recent quarterly filings showed under US$100,000 in cash against a quarterly net loss of roughly US$670,000, with about 1.3 billion shares on issue after cancelling 1.68 billion more by court order.
Four filters that catch almost all of these:
- Exchange. Nasdaq or NYSE main board, not OTC or OTCQB. Listing standards exist for a reason.
- Cash against burn. Fusion R&D costs tens of millions a year, minimum. A company with six figures in the bank is not building a reactor.
- Peer-reviewed results. Serious programmes publish. Ask what has been through review, and where.
- Share count history. Billions of shares outstanding, recent reverse splits or a name change within the last two years are all warnings.
Investment time frame
Match the horizon to the physics, not to the news cycle. There are three distinct clocks running:
The practical implication: if you cannot leave the money untouched for a decade, this theme is not for you. And even a decade may only get you to the beginning of commercialisation, through several rounds of dilution along the way.
Risk and return, stated plainly
What could go right
Fusion is a genuine power-law opportunity. If a company solves cheap, siteable, dispatchable fusion power, the addressable market is global electricity generation — measured in trillions. A successful first mover could compound for decades. That asymmetry is the entire reason to consider it: a position that can go to zero but might return many multiples of itself.
The problem is that the asymmetry is priced into venture rounds, not necessarily into the listed vehicles you can access. The upside accrues to whoever owns the winning company — which today is a private cap table.
What could go wrong
| Risk | Why it matters here |
|---|---|
| Dilution | Pre-revenue companies fund themselves by issuing shares. The Fusion Industry Association's survey put the median additional capital needed to reach a first pilot plant at around US$700 million per company. Your percentage ownership will shrink, repeatedly. |
| Timeline slippage | Every major fusion milestone in the modern era has moved right. SPARC has already slipped from 2025 to 2027. ITER now runs to the late 2030s, although its assembly has lately run ahead of plan — the sixth of nine tokamak sector modules was lowered into the pit almost six months early in July 2026. |
| Being beaten by boring technology | Fusion does not need to fail to lose. If fission restarts, small modular reactors, geothermal or solar-plus-storage supply AI's demand more cheaply by 2032, fusion arrives to a market that no longer needs it urgently. |
| SPAC mechanics | Redemptions, warrant overhangs and lock-up expiries create predictable selling pressure unrelated to progress. GFUZ's first fortnight is the illustration. |
| Single-experiment risk | Much of the sector's valuation rests on one machine in Massachusetts producing one result in 2027. |
| Currency | US holdings are unhedged AUD/USD exposure. A stronger Australian dollar erodes returns independently of the shares. |
| Liquidity | Small caps with thin volume are easy to buy and hard to exit in a falling market — exactly when you want out. |
How much of a portfolio?
We cannot tell you what to hold — that depends on your circumstances, and it is the kind of question a licensed adviser should answer with your full picture in front of them. What we can set out is how allocations like this are conventionally framed.
The standard structure is core-and-satellite: a diversified core doing the actual compounding, and a small satellite sleeve for high-conviction speculation. Most practitioners cap the entire speculative sleeve — every thematic and single-stock punt combined — at somewhere between 5% and 10% of a growth-oriented portfolio. Fusion is one theme competing for space inside that sleeve, which in practice puts it in the range of 1–3% of the total portfolio, and closer to the bottom of that range for a pre-revenue sector with one listed pure-play.
Three tests are more useful than any percentage:
- The zero test. Write the position down to nothing on paper. If that outcome changes your retirement date, your mortgage, or your sleep, the position is too large. This sector can go to zero.
- The decade test. Could you hold it, without adding, through a 60% drawdown and a decade of no revenue? If not, the size is wrong regardless of the percentage.
- The order test. Speculation comes after the boring work is done — emergency buffer in place, high-interest debt cleared, super and core holdings on track. If those are not sorted, the correct fusion allocation is zero.
For a great many investors, zero is the right answer, and there is nothing timid about it. Choosing not to buy a pre-revenue technology a decade from commercialisation is a legitimate decision, not a failure of imagination. As we have written before in the context of property investing, the returns that actually build wealth usually come from the unglamorous core, not the satellite.
Exit strategy
Speculative positions need exit rules written before you buy, while you are still capable of thinking clearly about them. For a theme with no earnings, no dividends and no valuation anchor, price targets are close to meaningless. Milestone triggers are not.
Sell triggers worth defining in advance
- Thesis broken. SPARC fails to achieve net energy gain, or slips beyond 2029 without a credible technical explanation. Helion misses its contracted 2028 delivery outright.
- Balance sheet broken. Cash runway falls under 12 months, or a raise arrives at a deep discount to the prevailing price. Dilution on bad terms is how these positions die quietly.
- Deal broken. For DJT specifically, the TAE merger failing to close removes the entire reason for a fusion investor to hold it.
- Story changed. A pivot away from fusion, a second name change, or the departure of the founding technical leadership.
Selling into strength
A widely used discipline for speculative positions: if the holding doubles, sell enough to recover your original capital and let the remainder run as a free position. You lock in the fact that you cannot lose money on it, and you keep exposure to the outcome you bought it for. The 2027 SPARC result is the most likely candidate for the kind of spike where this rule earns its keep.
Separately, rebalance the sleeve back to its target weight annually. If fusion runs and becomes 8% of your portfolio, it is no longer the position you sized — it is a concentration you drifted into.
Two Australian tax points that change the maths
Australian residents currently access the 50% CGT discount on assets held longer than 12 months. The 2026–27 Federal Budget announced that from 1 July 2027 this will be replaced by cost base indexation plus a 30% minimum tax on real capital gains. It passed the Senate on 25 June 2026 and received Royal Assent the following day, so this is now law.
For a ten-year fusion position bought today, that matters: gains will be split into a pre-1 July 2027 component under the old rules and a post-2027 component under the new ones. On current Treasury commentary there is no advantage in panic-selling shares before the switchover. This is general information, not tax advice — and it is an area to confirm with a registered tax agent, since the detail is still being finalised. The same Budget reshaped property investing, which we covered in our piece on what actually changed for negative gearing.
Directly held shares in US companies — including many US-domiciled ETFs — are generally treated as US-situs assets for US estate tax purposes. Under US domestic law, a non-US citizen who is not US-domiciled can face estate tax filing requirements once US-situs assets exceed US$60,000, with rates reaching 40%.
Australians, however, may receive substantially more favourable treatment under the Australia–US estate tax treaty, including access to a pro-rata share of the much larger US estate tax credit. In 2026 the US basic exclusion amount is US$15 million. The US$60,000 figure should therefore not be read as a simple tax-free threshold for Australian investors. Anyone with substantial direct US holdings, or a large worldwide estate, should obtain specialist cross-border estate tax advice.
Also worth doing before your first US trade: lodge a W-8BEN with your broker. It reduces US withholding tax on dividends from 30% to 15% under the Australia–US treaty. Pre-revenue fusion companies pay no dividends, so this matters more for the ETF and supply-chain names than for GFUZ.
The bottom line
The demand side of this thesis is the strongest it has ever been. AI's appetite for firm, clean power is real, growing, and already forcing contracts to be signed years in advance. Fusion's commercial validation — real customers, real offtake agreements, record funding — is genuine and new.
The investable side has not caught up. One pure-play on a major exchange, down 59% from its debut within a fortnight. One pending merger wrapped in political risk. A supply chain whose listed participants are driven by entirely different demand. And the sector's leaders sitting comfortably in private hands, funded by pension money, with no reason to sell you a share.
Fusion may well be one of the defining investments of the next thirty years. That does not mean today's listed vehicles are the way to own it. If you take a position, size it as venture capital — small enough to lose entirely, held long enough to matter, with the exit rules written down before the first purchase.
And for the opposite end of the risk spectrum — the boring core a speculative position like this should sit beside — see 7 ETFs that cover the entire market.