The smelters on life support

The potline never stops. That is the first thing anyone learns on the floor of an aluminium smelter, and it is not a slogan about work ethic. It is a physical fact about molten metal. Aluminium is made by passing an enormous electrical current through alumina dissolved in a bath of molten salt held at around 960 degrees Celsius, in long rows of steel-and-carbon cells called pots. The current is the process. Cut it, and the bath begins to cool. Leave it off for more than a couple of hours and the metal freezes solid inside the pots — and a frozen potline is not a plant that has been paused. It is a plant that has been destroyed, because digging cold aluminium out of hundreds of cells and rebuilding their linings costs so much that owners almost never do it.

So a potroom operator on night shift in the Hunter has a relationship with the electricity market that almost no other worker in Australia has. The plant she works in is not merely a large customer. It is the state's single biggest one, and everybody in the control room knows it. She has learned, over about a decade, to read energy news the way other people read the weather. There was a scare in the middle of the last decade, when the wholesale price spiked and the parent companies talked openly about viability. There was another when the coal station down the road announced its closure date. Her mortgage in a town where the smelter is the largest private employer has been paid off, year by year, against a plant whose supply contract has a known expiry date printed on it. What changed her family's ten-year outlook in August 2026 was not the aluminium price and not anything that happened inside the potrooms. It was two paragraphs in a government media release.

That operator is a composite, not a named person, and the vignette is illustrative. The numbers around her are not. The plant is real, the contract expiry is real, the money is real, and the pattern the money belongs to is the subject of this article.

What the governments did

In August 2026 the Commonwealth and New South Wales governments announced a joint support package for Tomago Aluminium, the smelter on the Hunter River near Newcastle, valued at up to $2.5 billion over ten years and split between the two governments (Commonwealth and NSW ministerial announcements, Aug-2026).

The structure matters more than the headline, and it should be described precisely because most commentary has not been. The package is not a cash grant paid on signature. It is a power-price guarantee: an undertaking that the smelter's electricity cost will not exceed a defined level, with governments meeting the difference between that level and the price the plant would otherwise face. The guarantee does not start now. It starts when Tomago's existing electricity supply agreement with AGL — a contract underpinned by coal-fired generation at the Bayswater power station — expires at the end of December 2028. The ten-year window therefore runs from the beginning of 2029 to the end of 2038 (announcement terms, Aug-2026).

What the money is buying is a specific industrial asset. Tomago has a nameplate capacity of up to about 590,000 tonnes of primary aluminium a year, which is roughly 40 per cent of Australia's total primary aluminium production (company and public reporting). It employs more than 1,000 people directly, plus around 200 regular contractors, and is credited with supporting in the order of 5,000 jobs across the Hunter when suppliers, maintenance firms and downstream users are counted (company and public reporting). Its electricity load runs continuously at close to a gigawatt, in the order of 10 per cent of all electricity consumed in New South Wales (approx.; company and public reporting; AEMO demand data). There is no other single customer of that scale in the state, and there is no comparable plant left elsewhere in the country that could absorb its output.

The three that came before it

Tomago is the largest item in a sequence, and the sequence is the reason this topic is worth 5,000 words rather than a news paragraph.

In the same month, the Commonwealth joined the Tasmanian and South Australian governments in a support package of approximately $135 million for the two Nyrstar plants — the zinc smelter at Hobart and the lead smelter at Port Pirie — both operated by a subsidiary of the commodities trader Trafigura, and both of which had been publicly described by their owner as loss-making (government announcements, Aug-2026).

Before that, across 2025 and 2026, the Commonwealth and Queensland governments agreed support for Glencore's copper smelter at Mount Isa and the associated copper refinery at Townsville, again in response to a stated risk of closure. This article does not quote a dollar figure for that package, because the verified record we are working from does not carry one, and inventing a number to complete a pattern would defeat the purpose of documenting it.

Further back, and under governments of both persuasions, the Portland aluminium smelter in western Victoria has been the subject of repeated Commonwealth and Victorian support arrangements, most prominently in 2017 and again at the start of the 2020s, each time framed as a time-limited bridge (Commonwealth and Victorian announcements, various).

Alongside the facility-by-facility rescues sits a sector-wide instrument. In January 2025 the Commonwealth announced a Green Aluminium Production Credit worth approximately $2 billion, a per-tonne production credit intended to reward smelters that shift to renewable electricity, available for up to a decade per participating plant (Commonwealth announcement, Jan-2025; subsequently reflected in budget documentation). That is a different kind of policy from a bilateral rescue — it is published, rules-based and open to any qualifying smelter — and the distinction between the two kinds of support is central to the critique later in this article.

One more piece of context belongs here, because it shows how far the underwriting habit now extends across the electricity system. In May 2024 the New South Wales government entered an underwriting arrangement with Origin Energy to keep the Eraring power station — the largest coal-fired generator in the country — operating beyond its announced closure date. Two years later, the same state government is helping to underwrite the state's largest electricity consumer. The public balance sheet is now standing behind both ends of the same wire.

Governments are now underwriting both the largest generator on the grid and the largest consumer on the grid, and neither arrangement was legislated as a policy.

The evidence: what has been committed, and what has not been written down

Set the packages side by side and two things become visible that are invisible one at a time: the scale has escalated sharply, and the design has become less rules-based as the amounts have grown.

The sequence, drawn

What $2.5 billion actually means

Big numbers are only useful once they are converted into something a reader can weigh, so here are four conversions, each with its own caveat.

Divided evenly, $2.5 billion over ten years is $250 million a year, split between two governments — so in the order of $125 million a year each (author's calculation on the announced envelope). Against roughly 10.9 million Australian households (approx.; ABS household estimates), the Commonwealth's share is a few dollars per household per year. Measured purely as a fiscal impost, this is not a large program. Anyone arguing against it on the grounds that the country cannot afford it is arguing from the wrong number.

Divided by jobs, the arithmetic reads differently. Against 1,200 direct and contractor positions, $250 million a year is in the order of $200,000 per job per year (author's calculation). Against the broader figure of around 5,000 supported jobs, it is in the order of $50,000 per job per year. Both numbers are crude — the package buys an industrial asset and a grid service, not a payroll — but the first of them is the number that will be quoted at every future negotiation, and a government that does not have an answer to it will lose the argument in public.

Divided by electricity, the figure becomes a price. A load of close to a gigawatt running continuously consumes something in the order of 8 terawatt hours a year (author's calculation). Two hundred and fifty million dollars spread across that volume is roughly $30 per megawatt hour (author's calculation, approximate and illustrative). That is a plausible order of magnitude for the gap between a legacy coal-backed contract struck years ago and the cost of firmed renewable supply at the end of this decade, which suggests the envelope was sized against a real commercial estimate rather than plucked from the air.

And divided by nothing at all, the honest answer is that we do not know what it will cost. A price guarantee is a contingent liability, not a cheque. If wholesale prices in the 2030s sit below the guaranteed level, the guarantee is never called and the cost to taxpayers approaches zero. If they sit well above it, the cost could exceed the headline. The $2.5 billion is a ceiling on an estimate, not a payment schedule. This cuts both ways, and both ways should be said: the taxpayer may pay far less than the headline, and the taxpayer has also accepted an exposure whose size depends on a market nobody can forecast a decade out.

What is not in the record

Here is the finding that the whole article turns on, and it is a finding about absence rather than about any figure.

Search the Commonwealth record for the framework that governs these decisions and there is nothing to find. There is no Act of Parliament establishing an energy-intensive industry support scheme. There is no published eligibility test setting out what a firm must demonstrate to qualify — no minimum employment threshold, no trade-exposure test, no emissions-intensity condition, no requirement to open the books. There is no published methodology for sizing a package against the value of what is being preserved. There is no sunset doctrine stating the conditions under which support ends rather than the date on which one particular contract expires. There is no appropriation line that a reader can look up in Budget Paper No. 2 under a scheme name, because there is no scheme.

What exists instead is a set of bilateral commercial negotiations, conducted between governments and individual companies, concluded confidentially and announced by media release. That is a materially different thing from a policy, and the difference is not a technicality. A policy can be evaluated, audited, extended to a firm that meets the test, and refused to a firm that does not. A negotiation can only be judged by its outcome, and only by people who were in the room.

Compare that with the one instrument in this space that was built properly. The Green Aluminium Production Credit has published parameters: it is a per-tonne credit, it is tied to the use of renewable electricity, it runs for a defined period per participating plant, and any qualifying smelter can apply. Whether the design is right is arguable. That it is a policy rather than a deal is not.

A policy can be audited, extended or refused. A negotiation can only be judged by the people who were in the room.

The critique: bargaining from the weakest possible position

The objection to this pattern is not that any individual package was wrong. On the facts available, each one was arguably right. The objection is to what the accumulation of individually-right decisions has produced in the absence of a framework: a standing, discretionary, growing subsidy for energy-intensive processing that no parliament ever voted on as a policy, and that no rule constrains.

The negotiating problem

Consider the position a government occupies in these talks. The counterparty is a company that has announced, publicly, that a plant may close. The plant is irreplaceable — once cold, it is gone. The employment is geographically concentrated in a region that is politically identifiable to the seat. The timeline is set by a contract expiry the government did not write. The company knows the cost of closure to the government, and the government does not know the company's true reservation price. Everything about that structure hands leverage to one side.

A framework fixes this, which is the entire reason frameworks exist. If the terms are published in advance — this is the test, this is the maximum, these are the conditions, these are the obligations that come with the money — then the government negotiates from a rule rather than from a deadline. Firms that meet the test know what they can expect. Firms that do not meet it stop asking. The government's answer to the fourth applicant is the same as its answer to the first, and it can be defended in public without reference to anything confidential. Without a framework, every negotiation restarts from zero, under time pressure, with the government's own political exposure sitting on the table as a bargaining chip for the other side.

The escalation visible in the chart above is the predictable consequence. Once a government has demonstrated that it will pay to prevent a closure, every subsequent firm facing marginal economics knows the option exists. That is not an accusation of bad faith against any company — it is what a competently run business does with information about its counterparty. It is also why the absence of a published test is not a paperwork failure. It is the substantive problem.

The December 2038 question

The sharpest test of any transitional support is to ask what happens on the day it ends, and the Tomago package invites that question with unusual precision because it has a date on it.

The guarantee runs from the beginning of 2029 to the end of 2038. Ask what is true on 1 January 2039. Tomago will still be a very large electricity load in a state that will have retired most or all of its coal-fired generation by then, on AEMO's published closure timetable and the Integrated System Plan's development pathway. Global aluminium supply conditions in 2039 are unknowable, but there is no particular reason to expect the structural oversupply that drove this decision to have resolved. The plant will still be roughly 40 per cent of national capacity, still irreplaceable, still the largest employer in its region, and still — if the economics have not changed — able to make exactly the same argument it made in 2026, with the same leverage, to a government that will by then have thirteen years of precedent behind it.

There are only three ways that day can go. The support is renewed, in which case it was never transitional and the country has acquired a permanent subsidy without ever debating one. The support ends and the plant closes, in which case $2.5 billion of public money bought a ten-year delay. Or the underlying economics genuinely change, through cheap firmed renewable supply, a durable recovery in metal prices, or a carbon border arrangement that prices the emissions intensity of imported metal — in which case the bridge worked as designed.

The third outcome is the one the policy is betting on, and it is not fanciful. But nothing in the announced package makes it more likely. A guarantee that simply covers a price gap for ten years does not, by itself, build the firmed renewable supply that would close the gap, and it does not create the carbon border arrangement that would neutralise the leakage. Those are separate policies, on separate timetables, in separate portfolios. If they arrive, the bridge lands somewhere. If they do not, the bridge ends in mid-air, and the choice in 2038 will be the same choice as in 2026, with more money already spent.

Who decides, and on what

There is a democratic point underneath the economic one. Australia has a large and mature apparatus for scrutinising industrial assistance. The Productivity Commission publishes an annual Trade and Assistance Review precisely so that assistance to industry is counted, categorised and made visible. Assistance delivered through the tax system appears in Treasury's tax expenditures statement. Grant programs are subject to Commonwealth grants rules and to audit by the Australian National Audit Office. This apparatus was built, over decades and by governments of both persuasions, because Australia's historical experience with unexamined industry protection was poor and the country decided it wanted assistance to be transparent.

Support delivered as a confidential bilateral price guarantee sits awkwardly inside all of that. It is real assistance with a real fiscal value, and it should be counted as such and disclosed as a contingent liability in the budget's statement of risks. But because it is not a program, there is no program to review, no guidelines to audit against, and no vote at which a member of parliament could have said yes or no. The public gets the announcement and the total. It does not get the test.

None of this requires attributing bad motives to anyone. Ministers facing an imminent closure announcement in a region that depends on the plant, with an irreversible asset at stake and no framework available to them, will negotiate a deal — because that is the only instrument in the drawer. The failure is not the deal. It is that after four of them, the drawer still contains nothing else.

What they got right, and the honest caveats

The credit here is substantial and specific, and it should be given before anything else in this section is said.

First, the structure of the Tomago package is materially better than the alternatives that were available. It is a power-price guarantee, not an open-ended cash grant. It has a defined ten-year window rather than an indefinite commitment. Its start is tied to a specific, identifiable transition event — the expiry of a coal-backed supply contract at the end of 2028 — rather than to a political cycle or a bad quarter. It leaves the commercial risk of running the plant with the owners, who remain responsible for operating costs, capital and metal price exposure. A government that wanted simply to write a cheque could have written one. This design is more disciplined than that, and the officials who built it deserve to be told so.

Second, the strategic and system-security arguments are not fig leaves. They are the strongest part of the case and they are genuinely load-bearing. A large interruptible load is worth real money to a grid managing high shares of variable renewable generation, and that value is not captured in any jobs-per-dollar calculation. Sovereign capability in refined metals has moved, across allied economies, from a talking point to a planning assumption. The irreversibility of a cold shutdown is a hard engineering fact, not an industry bargaining position, and it justifies a genuine option premium.

Third, the attribution should be stated plainly and in the government's favour. The distress at these plants was overwhelmingly imported. Global refining overcapacity, the compression of treatment charges, the rapid expansion of nickel supply out of Indonesia and the scale of Chinese production in aluminium, zinc and lead are conditions Australian governments operate inside, not conditions they created. To the extent that Australian energy costs contribute at all, they are one input among several, and they are the product of two decades of decisions across multiple governments and both major parties, not of any single administration. The correct verb for what this government did about global metal prices is neither "caused" nor "inherited". It presided over a period in which they moved against Australian producers, and responded.

Fourth, the Green Aluminium Production Credit is evidence that the government is capable of building the rules-based instrument this article says is missing. It was published, it is open to any qualifying smelter, and it ties support to a behavioural condition — the shift to renewable electricity — rather than merely to the threat of closure. That is a template. It exists. It simply has not been extended to cover the situations where the money is actually being spent.

The caveats on the credit

A defined window is only a discipline if it holds, and there is no mechanism in the announced structure that makes it hold. The end date is a term of a commercial agreement, not a statutory sunset with conditions attached, and commercial agreements are renegotiated when the alternative is a closure announcement.

The design also asks nothing in return that would change the position in 2038. On the public record there is no published requirement that the plant procure a stated volume of firmed renewable supply by a stated year, no obligation to maintain demand-response availability at a defined level, no employment floor, no clawback if the owner closes the plant shortly after the guarantee expires, and no equity or profit-share that would return value to taxpayers if metal prices recover strongly. Some or all of these may exist in the confidential agreement. The point is that the public cannot know, and a support arrangement of this size and duration should not be a matter of trust.

And the good instrument and the ad hoc instruments have not been connected. The production credit rewards decarbonised production; the Tomago guarantee bridges a plant from coal-backed supply toward something not yet specified. A coherent policy would make the second conditional on measurable progress toward the first. As announced, they are separate.