Stand at the fence line of a gas peaker on a still winter evening around 6pm and you’ll see the thing that spreadsheets never quite capture: nothing happening, then everything happening. Rooftop solar has fallen off a cliff, the wind’s gone to sleep over the Southern Ocean, and 25,000 air conditioners and heaters across the state have just switched on in the space of forty minutes. Somewhere in a control room, AEMO’s dispatch engine asks for more megawatts, right now, and a turbine that was sitting idle a few minutes earlier starts spinning up. That’s firming. It’s less a technology than a job description, and gas has been doing most of that job for twenty years.
Worth being precise about what the word actually covers, because it gets used loosely enough that half the arguments about it are really arguments about different things.
Firming is a verb, not a generator type #
Firming means making an intermittent or variable resource behave like a reliable one. Wind and solar are variable by nature – brilliant when the resource is there, worth nothing when it isn’t. Firming is whatever fills that gap: a battery discharging for two hours, a gas turbine running for six, a pumped hydro scheme releasing water from an upper reservoir, or in extreme cases, demand response paying a smelter to turn itself off for an afternoon. None of those things generate “firm” power on their own in the way a coal or nuclear plant does by just sitting there burning fuel continuously. They all firm something else.
That distinction matters because people sometimes treat batteries and gas as substitutes doing the identical job, when really they’re doing different shifts. Batteries, as things stand with current economics, are mostly doing the short, sharp firming: an hour or two either side of sunset, catching the evening ramp as rooftop and utility solar drop away together. Gas peakers tend to get called on for the longer, less predictable stretches: a cold snap running three days, a wind drought across the whole southern half of the continent, the kind of event AEMO’s reliability forecasts flag but can’t put a date on.
Why the duration problem keeps bringing gas back into it #
I went back through the AEMO Integrated System Plan modelling again recently – not because I expected the headline number to have moved much since the 2024 draft, but because the duration assumptions are where the real argument lives, and those get revised more often than the top-line capacity figures. The grid doesn’t need firming for fifteen minutes very often. It needs firming for four hours most evenings, and every so often it needs firming for four days. Batteries, at the duration and scale currently being built, mostly two to four hours, are superb at the first problem and close to useless at the second. You’d need an uneconomic amount of battery capacity sitting idle 350 days a year to cover the once-a-decade event with batteries alone.
Gas, by contrast, is mediocre at the short sharp stuff compared to a battery (slower response, higher marginal cost per megawatt-hour when it does run) but very good at showing up for day three of a cold snap, because the fuel supply isn’t a function of how charged something was yesterday. That’s the honest version of why gas keeps coming up in every planning document even as the NEM adds gigawatts of storage: it’s not that batteries have failed, it’s that the two technologies solve different lengths of problem, and nobody has built enough long-duration storage – pumped hydro, mostly – to close that gap yet. We’ve covered the pumped hydro versus batteries trade-off in detail before, and the short version there is the same: duration is the scarce commodity, not megawatts.
What AEMO’s own numbers say about the mix #
AEMO’s 2024 Integrated System Plan, still the working reference document even as the next draft gets finalised, has gas-fired generation shrinking as a share of total energy produced across the NEM over the next two decades, while gas-fired capacity, the nameplate megawatts sitting there able to run, barely shrinks at all, and in some scenarios grows. That’s not a contradiction, it’s the whole point of firming. You don’t need the gas fleet to run often. You need it to exist, fuelled and maintained, for the handful of days a year when wind, solar and batteries between them can’t cover it. A peaker with a five per cent capacity factor is doing its job if it’s there on the one evening it’s needed. Coal plants were never built to run that way, which is part of why the economics of coal closures and the reliability debate keep circling back to this same question about what replaces the always-there assumption.
Readers who want the primary source rather than my summary of it can find AEMO’s ISP documents and the accompanying forecasting reports on AEMO’s own publications page, which is a less exciting read than it sounds but a more reliable one than most of what gets said about it in press releases.
The Capacity Investment Scheme is quietly deciding the mix #
This is the part that doesn’t get enough attention relative to how much it matters. The federal Capacity Investment Scheme runs tenders that effectively decide, region by region, how much new firming capacity gets underwritten and in what form. We’ve written before about what the CIS tender clock shuts out, and the short practical effect is that batteries and long-duration storage are winning most of the contracted capacity in recent rounds, while gas peakers are mostly getting built outside the scheme, financed on their own economics or state-government balance sheets – Kurri Kurri in the Hunter being the obvious recent example of a project that took the long way round, as we covered in detail at the time.
That split is worth sitting with. It suggests the policy settings are, deliberately or not, steering new short-duration firming toward batteries and leaving the long-duration, few-days-a-year firming to whatever gas capacity already exists or gets built commercially. Whether that’s a coherent strategy or an accident of which technology happens to clear each tender round is a fair question, and I don’t think anyone in Canberra has fully answered it yet.
The cost side nobody loves quoting precisely #
Gas peaking is expensive on a per-megawatt-hour basis when it runs, because you’re paying for a power station’s capital cost spread across very few operating hours, on top of a fuel cost that moved a long way in the wrong direction after the east-coast gas market got tight through the 2020s. That’s the core of the argument in gas peakers versus big batteries, which I’d point you to if you want the full maths rather than my abridged version. But the flip side that often gets dropped from the hype-against-gas version of this argument: a battery that’s cycled every single day to chase arbitrage, with little reserve left for the rare four-day event, isn’t actually solving the same reliability problem a peaker solves. Comparing their cost per megawatt-hour as though they’re interchangeable products is a bit like pricing a taxi against a tow truck because they’re both vehicles that move things.
Where I’d push back gently on the industry’s own messaging, on both sides: gas proponents sometimes talk as if the fleet needs to keep growing in absolute terms, when the ISP modelling mostly has it holding roughly steady or shrinking slightly as batteries and pumped hydro pick up more of the shorter-duration load. And renewables advocates sometimes talk as if firming is a solved problem once enough batteries get built, when the duration maths plainly says otherwise, at least until something like Snowy 2.0 or Borumba actually delivers meaningful pumped hydro capacity, both of which have had their own well-documented schedule and cost blowouts worth remembering before leaning on them as the answer.
What this looks like on the ground over the next few years #
Practically, expect the pattern that’s already emerging to keep going: batteries and the Waratah Super Battery-style grid-scale projects soak up the short evening peaks and the frequency control work, gas plants get called less often but stay on the books as the insurance policy, and the genuinely hard problem (the week-long low-wind, low-sun stretch) stays unsolved by anything except gas, hydro where there’s water, and interconnection between regions, which is its own long and expensive story tied up in projects like Marinus Link and VNI West. None of that is a verdict on whether gas is “good” or “bad”. It’s just where the duration maths currently sits, and duration maths doesn’t move because of a press release.
I’ll admit my own view has shifted slightly on this over the years. I used to think the gas fleet’s role would shrink faster than it has. It hasn’t, mostly because nothing else cheap enough has turned up yet to do the four-day job. That might change with more pumped hydro, or it might not, and anyone who tells you with total confidence which way it goes by 2035 is spruiking, not forecasting.
Next time you hear a politician or a company announcement use “firming” as though it were a single, simple thing, it’s worth asking which duration they’re actually talking about. The answer tells you more than the announcement does.
– Tom Fitzgerald, Baseload & Fuels Correspondent
Photo by American Public Power Association on Unsplash