Maria Petrovic has run the ovens at her family's bakery in Morwell, in Victoria's Latrobe Valley, since before sunrise most mornings for eleven years. The bakery supplies bread and pastries to cafes across Gippsland, and its walk-in coolers and proving ovens draw power around the clock — a load that cannot simply wait for a calmer hour if the grid gets tight. Morwell sits a few kilometres from Yallourn Power Station, the ageing brown-coal plant that has helped keep Victoria's lights on for more than half a century and that is now scheduled to close well before the decade is out. Maria supports the shift to cleaner energy; she has solar panels on the bakery roof and would install a battery tomorrow if the numbers worked. What worries her is the gap between Yallourn's closing date and the day the wind, solar, batteries and transmission lines meant to replace it are actually finished, connected and reliably delivering power — and what happens to her power bill, already her second-largest cost after flour, while that gap is bridged.
Her question is the same one this article sets out to answer at a system-wide scale: what does it actually cost — not the sticker price of a wind or solar farm alone, but the full bill for a grid that stays reliable through the changeover — to replace coal with a system built mostly on wind and solar? The answer runs through three ideas that headline generation-cost comparisons routinely leave out: how much of the time a wind or solar farm can actually generate, what it costs to store and dispatch power when it cannot, and who ends up paying for the transmission lines, batteries and backup capacity needed to stitch it all together. Those three ideas are why "renewables are the cheapest new generation" and "a firmed, reliable renewable grid is the cheapest system" can both be true statements, and still describe very different bills.
What the government did
The Albanese government's approach to the cost side of the transition rests on three main instruments, each aimed at a different part of the gap between cheap generation and a reliable system. The first is the Capacity Investment Scheme (CIS), a Commonwealth underwriting mechanism expanded in 2023 and 2024 that offers a government-backed floor on revenue to new projects, in exchange for a share of any windfall gains if wholesale prices run unusually high. Of its roughly 32-gigawatt target for new capacity by 2030, a meaningful share — on the order of a quarter to a third — is reserved specifically for dispatchable, "firming" capacity: batteries, pumped hydro and other technologies that can supply power on demand, rather than only when the wind blows or the sun shines, with the remainder allocated to variable wind and solar (DCCEEW, Capacity Investment Scheme design documents, 2023–24; approx.). The scheme is explicitly designed to buy firming capacity alongside generation, rather than assume the market will supply it unprompted.
The second instrument is Rewiring the Nation, the $20 billion concessional financing facility administered through the Clean Energy Finance Corporation and created in the October 2022 Budget to fund the transmission backbone the Australian Energy Market Operator's Integrated System Plan says the grid needs to move renewable power from where it is generated to where it is used (Budget Paper No. 2, Oct-2022 Budget). Transmission is where much of the "firming" cost of the transition actually shows up in practice, because a wind or solar farm in a remote, resource-rich region delivers nothing to a household in a capital city unless a high-voltage line connects the two — and building that line typically takes longer, and has proven harder to cost accurately, than building the generator itself.
The third instrument is Snowy 2.0, the single largest dedicated storage project the Commonwealth has committed to, and the one meant to anchor the firming side of the system at scale. Announced in 2017 as a way of pumping water uphill using cheap or surplus power and releasing it back through turbines at times of peak demand, Snowy 2.0 was originally pitched as a roughly $2 billion project able to add 2,000 megawatts of dispatchable capacity and around 350,000 megawatt-hours of storage to the National Electricity Market (Snowy Hydro Limited, project announcements, 2017; approx.). It is owned by the Commonwealth, run by Snowy Hydro Limited, and treated in AEMO's planning as one of the "actionable" projects the least-cost pathway to 82 per cent renewables depends on.
Alongside these large, centralised projects, the government has also moved on the household side of firming. From 1 July 2025, the Cheaper Home Batteries Program offers a rebate that cuts the upfront cost of installing a home battery, aimed at both easing bills and reducing pressure on the grid at peak times by letting households store their own solar and use it after dark (Australian Government, Cheaper Home Batteries Program, 2025; approx.). It builds on Australia's already high rate of rooftop solar uptake and follows earlier state-based battery rebate schemes in Victoria, South Australia and elsewhere. Together, these instruments — underwritten generation and storage, transmission financing, a flagship pumped-hydro project, and household batteries — are the mechanisms through which the government has tried to buy not just cheap electrons, but a system that keeps delivering them reliably as coal retires.
The strongest case for the policy
Put the government's defence at its best
The government's defence starts from a number most critics of the transition do not actually dispute: nobody is proposing to build a new coal-fired power station in Australia, because no private financier will fund one and no government has offered to build one outright. The live choice is not "cheap coal versus expensive renewables"; it is between a firmed renewable system and whatever mix of ageing coal, gas and, eventually, new nuclear or new coal would have to replace it instead — costed the same honest way, including all of its own firming, backup and replacement costs, not just the price on the label.
On that comparison, CSIRO and AEMO's jointly produced GenCost report — the standard public reference for Australian generation costs, updated annually and open to challenge by anyone who thinks the assumptions are wrong — deliberately builds in the extra cost of storage and additional transmission needed to firm wind and solar as their share of the grid rises. Even modelled at a 90 per cent variable-renewable share of the system, a scenario deep enough to require substantial storage and network augmentation, GenCost's integrated cost estimates for wind and solar plus firming have consistently come in below the equivalent cost of new-build coal, new-build gas with carbon capture, or new nuclear generation of any kind Australia could plausibly build this decade (CSIRO and AEMO, GenCost report, 2023–24 and 2024–25 editions). That is not GenCost comparing renewables to an easy target; it is GenCost comparing a fully firmed renewable system to the honestly costed alternatives, and finding renewables still cheaper.
The government's second point is about the cost of waiting. Ageing coal plants do not become cheaper or more reliable the longer they run past their planned life; they become the opposite. Older coal units already suffer more frequent unplanned breakdowns than newer plant, a maintenance burden that shows up as periodic price spikes and reliability warnings rather than one clean line item, and every year of delay to the replacement build is a year the ageing fleet has to be nursed through instead (AEMO, Electricity Statement of Opportunities, 2023–24 editions; approx.). On the government's account, the choice was never a firmed renewable grid against a costless status quo. It was a firmed renewable grid, built as fast as the country can manage, against an ageing fleet whose maintenance and reliability bill was always going to rise regardless of what replaced it — and the modelling says the former remains the cheaper of the two paths, even accounting for the very real construction costs this article goes on to examine.
The evidence: what the sticker price leaves out
What "capacity factor" means, and why it changes the sum
Start with a concept that does more work in this debate than almost any other, and that headline generation-cost comparisons often leave out entirely: capacity factor. A generator's capacity factor is the share of the time, averaged over a year, that it actually produces power at its full rated output. A coal or gas plant built to run continuously can have a high capacity factor, limited mainly by fuel supply and maintenance. A wind or solar farm's capacity factor is set by the weather: a solar farm generates only during daylight hours, and best around midday, while a wind farm generates only when the wind blows within its useable range. On CSIRO and AEMO's own planning assumptions, utility-scale solar in Australia runs at a capacity factor of roughly a quarter of the time (approx.), and onshore wind at roughly a third to two-fifths of the time (approx.), against an older black or brown coal plant that — even accounting for its own age-related outages — still runs closer to half to two-thirds of the time (CSIRO and AEMO, GenCost report; AEMO, 2024 Integrated System Plan; approx.).
That gap matters because the headline "cost per unit of electricity" figure — the levelised cost of electricity, or LCOE, the average price per unit a generator would need to earn over its lifetime to cover the cost of building and running it — is calculated per unit actually generated, not per unit of capacity installed. A wind or solar farm can have a genuinely low LCOE and still leave a hole in the grid for two-thirds or three-quarters of the year, a hole that a fossil generator running continuously does not leave. Filling that hole — with batteries, pumped hydro, gas peaking plants or extra transmission to import power from somewhere the wind is blowing — is the "firming" cost, and it does not show up in the bare generation LCOE at all. It shows up somewhere else: in a separate storage project, a separate transmission project, or a separate gas contract, each with its own budget line and its own capacity to blow out.
Snowy 2.0: the firming project meant to anchor the system
Snowy 2.0 is the clearest single case study of what happens when a firming project meant to anchor a system runs into the ordinary difficulty of major civil construction. Tunnel-boring machines working through Kosciuszko National Park's granite and complex rock formations struck conditions more difficult than the original feasibility studies assumed, and the project's costed scope grew to include transmission connection works that early public announcements had not fully priced in. Snowy Hydro Limited's own cost disclosures show the project's estimated cost rising from the roughly $2 billion figure used at its 2017 announcement to more than $12 billion by the early 2020s, with some accounts of the fully connected cost, including its associated transmission works, running higher again (Snowy Hydro Limited, project cost and schedule updates, 2017–2024; approx.). The in-service date has moved in step: first generation from the scheme, originally targeted for around 2021, was still not fully delivered several years past that date, with completion pushed out toward the back half of this decade (Snowy Hydro Limited, project updates; approx.). The Australian National Audit Office has separately examined aspects of the project's governance, underscoring that the cost growth has drawn independent scrutiny, not just industry commentary (Australian National Audit Office, review of Snowy Hydro Limited governance; approx.).
None of this means Snowy 2.0 will not eventually work as intended; large pumped-hydro projects elsewhere in the world have also taken longer and cost more than first announced, and still ended up as valuable long-life assets once complete. But it does mean the single project the government chose as the flagship of its firming strategy has, on its own numbers, taken several times longer to plan and several times more to build than first advertised — and every dollar and year of that overrun is a dollar and a year the rest of the system had to do without the dispatchable capacity Snowy 2.0 was meant to supply.
The other ways to firm the system
Snowy 2.0 is not the only technology carrying the firming task, and the mix matters because each option trades off differently against cost, speed of construction and how long it can keep supplying power. Grid-scale batteries can be built in a year or two rather than a decade, and respond within seconds, which makes them well suited to smoothing short dips of a few hours — but most currently deployed at scale store only two to four hours of output, well short of what is needed to cover a multi-day stretch of low wind and overcast skies (AEMO, 2024 Integrated System Plan; approx.). Pumped hydro schemes like Snowy 2.0 can store and dispatch power for much longer stretches, but as this article has shown, they take years of civil construction and carry real cost and schedule risk of their own. Gas peaking plants sit at the other end of the trade-off: faster to build than a pumped-hydro scheme, able to run for as long as fuel supply allows, but reliant on a gas market this site's companion analysis of gas, energy security and the interventions shows has its own price volatility and supply-security questions, and each hour a gas peaker runs is an hour of emissions the 43 per cent target is meant to be cutting. AEMO's own modelling keeps some gas generation in the fleet through the 2030s in every scenario it publishes, including the Step Change pathway the government's target is built on, precisely because no single firming technology yet covers every part of this trade-off on its own (AEMO, 2024 Integrated System Plan). The system cost this article is tracking is, in practice, the cost of assembling several imperfect firming technologies into one reliable system, not the cost of any single "solution."
What network charges show
The firming cost that does not show up in a generation LCOE does not vanish; on current regulatory settings, most of it is recovered through network charges — the part of an electricity bill that pays for the poles, wires and transmission lines used to deliver power, as distinct from the wholesale cost of generating it. Transmission and distribution businesses recover the cost of assets like Snowy 2.0's connecting lines, HumeLink and Marinus Link — the transmission blow-outs examined in this site's companion analysis of the 43% target and the net-zero arithmetic — through revenue determinations set by the Australian Energy Regulator (AER), typically over five-year periods, with the cost of the underlying asset spread and recovered, with a regulated return, over that asset's working life, often several decades (Australian Energy Regulator, transmission and distribution revenue determinations, 2023–24). That is a sensible way to finance long-life infrastructure, spreading the cost over the decades of service it provides rather than requiring an eye-watering bill in the year construction happens. But it also means a network cost overrun does not land as a one-off political embarrassment the way a blown election promise does; it lands as a slightly higher line on every bill, every quarter, for years, largely invisible unless a customer goes looking for the breakdown.
The Australian Energy Market Commission's Residential Electricity Price Trends reporting, which breaks a typical household bill down into its wholesale, network, environmental and retail components, has repeatedly found that network charges make up somewhere between a third and half of a typical household electricity bill — often the single largest component, ahead of the wholesale generation cost itself (Australian Energy Market Commission, Residential Electricity Price Trends; approx.). That is the mechanism through which the "hidden" firming and transmission cost of the transition becomes visible to a household or a business like Maria's bakery: not as a separate transition levy she can identify and debate, but as part of the ordinary network charge on every bill, rising in step with the cost of the assets — Snowy 2.0, HumeLink, Marinus Link and the rest — being built to deliver it.
A kilowatt-hour that is cheap to generate and expensive to deliver reliably is not a cheap kilowatt-hour. It is a bill deferred, not a bill avoided.
The critique: the cost is real, and it is being paid quietly
The government's defence rests on a genuine and carefully documented modelling result: a firmed renewable system, costed honestly and in full, is projected to be cheaper than the fully costed alternatives. The problem is that GenCost's integrated cost estimate is a long-run, whole-of-system model, not a guarantee about any individual project's delivered cost or delivery date. It answers the question "if the system is built roughly as planned, what would it cost?" It cannot answer, and does not claim to answer, "will Snowy 2.0, HumeLink, Marinus Link and the rest actually be built roughly as planned?" The evidence assembled above says, so far, not quite. Snowy 2.0's cost has grown several times over from its original public estimate, and its schedule has slipped by most of a decade. HumeLink and Marinus Link, the transmission lines this site's companion analysis of the 43% target examines in detail, have followed a similar pattern of major cost growth and schedule slippage. Every one of these is a specific input the modelling assumes exists, roughly on budget and roughly on time, in order to reach its "firmed renewables are cheaper" conclusion. The more of them blow out, the further the real, delivered system drifts from the modelled one — and the weaker the guarantee that the cheaper path on paper stays the cheaper path in practice.
The second problem is where the overruns land. A generation project that blows its budget is usually visible: a wind or solar developer that cannot make a project pay simply does not build it, or sells it to someone who thinks they can. A regulated network asset works differently. Once the AER approves a transmission or distribution business's spending as prudent and efficient — the test the regulator is required to apply — the cost of that asset, overruns included where the regulator accepts them as reasonably incurred, is recovered from customers over the asset's working life through the network charge on every bill (Australian Energy Regulator, transmission and distribution revenue determinations, 2023–24). That is not evidence of bad faith by the regulator, which exists precisely to scrutinise these costs on the public's behalf. But it does mean a project like Snowy 2.0's associated transmission works, or a delayed and over-budget interconnector, does not produce a single, dateable, attributable event a voter can respond to at the next election. It produces a slightly larger number on a bill that arrives quarterly, indefinitely, blended in with everything else the network charge covers — power poles replaced after a storm, routine maintenance, the cost of connecting a new suburb. The transition's firming and transmission cost is real, it is being paid, and it is one of the hardest costs in the entire federal ledger for an ordinary household to actually see.
That invisibility helps explain why the government's own 2022 election commitment to cut household power bills by $275 by 2025 — examined in full in this site's analysis of the $275 promise — has proven so difficult to deliver, even as the wholesale cost of new-build renewable generation has, on CSIRO and AEMO's own figures, kept falling. A promise about the bill implicitly promises about the whole bill, network charges included, not just the wholesale generation component that renewables are actually making cheaper. If network charges rise by more than wholesale prices fall — which is roughly the pattern across the period this article examines — the total bill can rise even while the underlying claim, "renewables are getting cheaper to build," remains entirely true. Both things are true at once, and a bill promise pitched against only one of them was always going to be tested against the other.
The network-charge mechanism also does not spread its cost evenly. Areas that host large new transmission projects, or that sit further from existing generation and therefore need more new line built to reach them, tend to carry a higher share of the cost recovery in their local network charges than areas closer to existing infrastructure, an effect the AER's regulatory framework is built to reflect rather than smooth away (Australian Energy Regulator, transmission and distribution revenue determinations, 2023–24; approx.). Regional communities like the Latrobe Valley, already absorbing the disruption of hosting the ageing coal plants the transition is retiring, can end up carrying a disproportionate share of the network cost of replacing them as well — a distributional question the transition's headline "cheaper system" framing does not address, because it is a question about who pays, not only how much is paid in total.
The reliability side of the ledger carries its own, separate risk, and it compounds the cost risk rather than sitting apart from it. AEMO's Electricity Statement of Opportunities has repeatedly flagged reserve-capacity shortfall risk in New South Wales and Victoria in the second half of this decade if new generation, storage and transmission do not arrive close to schedule as coal exits (AEMO, Electricity Statement of Opportunities, 2023 and 2024 editions). This site's companion analysis of the 43% target documents the clearest case already on the record: the 2025 decision by the New South Wales Government and Origin Energy to extend the life of Eraring, the state's largest coal plant, specifically because the storage and transmission meant to replace it were not far enough advanced to let it close on schedule. Victoria faces a comparable question over Yallourn, in Maria's own Latrobe Valley, where AEMO's reliability forecasting flags a similar risk window later this decade as the plant approaches its scheduled retirement (AEMO, Electricity Statement of Opportunities; approx.). Every time a coal plant's closure is delayed to manage this risk, it is itself a cost — extending the operating life, and often the public support, of an asset the transition was meant to be retiring — layered on top of the firming and transmission costs already examined.
Every time a coal plant's closure is delayed to manage reliability risk, it is a sign the sequencing has not been solved — only postponed, at a cost, one plant at a time.
None of this shows up as a single "cost of the transition" line in a budget paper or on a bill, and that is itself part of the problem this article documents. The generation cost sits in wholesale market data. The transmission cost sits in AER revenue determinations. The storage cost sits in Snowy Hydro Limited's own project accounts. The reliability cost sits in a state government's decision to keep one coal plant running a little longer. Each is real, each is separately documented in the primary record cited throughout this article, and each is defensible on its own terms. But no single official document adds them together and asks whether the full, all-in figure is still tracking below the fully costed alternative GenCost compares it to — which means the honest answer to "is the transition still the cheaper path?" currently has to be assembled by a reader piecing together several different regulators' and companies' numbers, rather than read off one page.
What the government got right
Credit is due on several fronts, and a fair audit gives it. The government did not choose to model the transition's cost on optimistic, cherry-picked numbers; the GenCost methodology it relies on is CSIRO and AEMO's own public, annually updated document, and it is one of the more transparent generation-cost exercises undertaken anywhere in the world, deliberately built to be contestable by industry, unions and critics alike rather than presented as a settled political talking point. That the modelling itself explicitly includes firming and integration costs, rather than quoting a bare generation LCOE the way earlier public debates often did, is itself a genuine improvement in how Australia has this argument.
Second, the design of the Capacity Investment Scheme shows a government that learned from an earlier generation of renewable-support schemes. Australia's original Renewable Energy Target, running from 2001, was built almost entirely around supporting variable generation, with no comparable mechanism to guarantee the dispatchable capacity needed to firm it — a gap later reliability forecasting, including AEMO's own, identified as a real contributor to the tightening reserve margins of the 2010s. Reserving a defined share of the CIS target specifically for dispatchable, firming capacity, rather than assuming the market would supply it once enough wind and solar existed, is a direct response to that earlier lesson, and the scheme's early tender rounds have reportedly run ahead of some of their own interim milestones (DCCEEW, Capacity Investment Scheme tender results, 2024–25; approx.).
Third, the household side of the ledger has been a genuine, if quieter, success. Australia already has one of the highest rates of rooftop solar per household in the world, and the Cheaper Home Batteries Program builds directly on that base, aiming to turn millions of individually small, distributed batteries into a form of firming capacity that does not carry the single-project blow-out risk of a Snowy 2.0 or a HumeLink, because it is built one rooftop and one battery at a time, largely with the household's own capital plus a rebate, rather than a single multi-billion-dollar balance-sheet commitment (Clean Energy Regulator, small-scale renewable energy scheme data; Australian Government, Cheaper Home Batteries Program, 2025; approx.). Whatever the shortfalls documented elsewhere in this article, distributed household storage is one part of the firming task where actual delivery has kept pace with, or exceeded, expectations.
Finally, it is worth stating plainly what the alternative would have cost in scrutiny, not just dollars. The cost overruns on Snowy 2.0 and the major transmission projects are on the public record specifically because Australia has regulators — the AER on network revenue, the Australian National Audit Office on Commonwealth project governance — whose job is to make exactly this kind of blow-out visible and contestable. A country that wanted to hide these costs more effectively than Australia has managed would need weaker, not stronger, independent scrutiny of its infrastructure spending. That the numbers in this article can be cited to named public bodies and named reports, rather than assembled from leaks or estimates, is itself part of the case that the system — however costly its delivery has proven — is at least being run in the open.
The risk to the future
Costs postponed into network charges; reliability postponed into a shrinking margin
Two risks compound each other, and neither is speculative; both are already visible in the primary record examined in this article. The first is that the transition's true, all-in cost advantage — the modelled result GenCost finds for a firmed renewable system over the fully costed alternatives — depends on individual projects like Snowy 2.0 and the major interconnectors landing close to their costed estimates and schedules, and so far the largest of them have not. Every further year and every further billion dollars of overrun narrows the margin between "the cheaper system on paper" and the system Australians are actually paying for, and because most of that cost is recovered through network charges amortised over decades rather than a single visible bill, the people ultimately paying for today's delays and blow-outs include households, tenants and small businesses — Maria's bakery among them — who had no vote on the projects' approval, and in some cases are not yet born. The second risk is reliability itself: AEMO's own forecasting has already flagged shortfall risk in the states carrying the largest share of the remaining coal exit, and the Eraring extension shows that risk is not hypothetical — it has already forced one state government to keep an ageing coal plant running past its planned closure date rather than risk the alternative. If similar decisions are needed again, at Yallourn or elsewhere, each one is itself a cost, and a sign the sequencing between what is closing and what is ready to replace it has not been solved, only managed one plant at a time. Neither outcome is certain, and this article does not claim otherwise: the government's underlying wager — that a firmed renewable grid ends up cheaper than the alternative even accounting for these overruns — may yet prove correct on the full, multi-decade horizon GenCost models. But the next generation of Australians will not inherit a modelled outcome. They will inherit whatever grid, and whatever set of network charges, the delivery record documented here actually produces — and on the evidence assembled in this article, that record is currently running well behind the schedule and the budget the cheaper-system case was built on.