Australia’s Electricity Crisis: the 2028-2033 Stress Window Mapped

Two numbers define the next decade of Australian energy, and they are moving in opposite directions. Coal still generated 42.7% of the country’s electricity in calendar year 2025. Over the next ten years, total electricity demand is projected to climb more than 40%.

The problem is not the transition itself. It is the timing. Four major coal stations are set to close between 2028 and 2035, removing significant dispatchable capacity, while data centres are forecast to quadruple their share of the grid from 3% to 13% by 2035. Whether renewables, storage and gas can fill that gap on the schedule AEMO’s planning assumes is the contested question at the heart of Australia’s electricity crisis.

This piece maps the supply-demand math so the stress points become concrete. You will see which fuel sources are genuinely contested, where the shortfall risk is structurally likely, and why the window between 2028 and 2033 is the period that matters most for anyone watching Australian energy infrastructure or resource investment.

What Australia’s grid actually looks like right now

Start with the number that reframes everything that follows: coal is not a legacy fuel winding down in the background. It is still the workhorse of the entire system.

Coal supplied 42.7% of Australia’s electricity in calendar year 2025.

That single figure does more analytical work than any projection, because it establishes that coal supplies more than four times the electricity of any other single dispatchable source. Here is how the rest of the mix sat in 2025, according to the Australian Energy Statistics compiled by the Department of Climate Change, Energy, the Environment and Water:

• Coal: 42.7% of generation
• Solar: approximately 19.6% of generation
• Gas: approximately 16.2% of generation
• Other sources (wind, hydro and the remaining balance): approximately 21.5% combined

Total generation reached approximately 287 TWh across the year, which sets the scale of the system every replacement plan is measured against.

What this tells you is uncomfortable for anyone who assumes the energy transition is largely done. On the dispatchable side, the side that delivers power on demand regardless of weather, coal remains overwhelmingly dominant. Australia’s grid is not finishing a transition. It is largely yet to begin the hardest part of one.

The pressure compounds from the demand side at the same time. State electrification policies are actively shifting energy use away from gas and toward electricity, pulling more load onto the grid precisely as coal’s exit is being locked in. The supply source doing the most work is scheduled to leave the room, and the room is filling up.

Understanding this baseline is the precondition for evaluating any claim about reliability. Without it, the sheer size of the replacement task is easy to underestimate.

The retirement cliff: four stations, five years, 8 GW leaving the grid

The retirements are not a planning assumption. They are a dated schedule with named operators, and the weight accumulates fast.

First comes Yallourn in Victoria, operated by EnergyAustralia, with roughly 1.6 GW of capacity exiting in June or July 2028. The following year brings two closures almost back to back: Gladstone in Queensland, around 1.7 GW, shutting in March 2029, and Eraring in New South Wales, operated by Origin Energy and the largest coal station in the country at approximately 2.9 GW, closing in April 2029.

Add the first three together and roughly 6.2 GW of coal capacity is gone by mid-2029. Then Loy Yang A in Victoria, operated by AGL Energy, follows in 2035, taking the cumulative loss well beyond the first three closures. What reads as a managed phase-out on paper lands, when stacked in sequence, as a structural hole opening in the dispatchable fleet inside five years.

NSW coal transition policy is particularly consequential in this analysis because Eraring, the single largest closure in the retirement schedule at approximately 2.9 GW, sits inside the NSW grid, meaning that state’s approach to replacement capacity and worker transition timelines directly shapes how much of the 2029 shortfall risk materialises.

The fleet-level picture from AEMO’s 2024 Integrated System Plan (ISP) sharpens the point. Ten large coal stations have already closed since 2012, and retirements have now been announced for all but one of the remaining fleet, with about half scheduled to exit by 2035. Across the decade, AEMO expects roughly 13-15 GW of coal and gas-fired generation to leave the system.

Then there is the signal hiding inside AEMO’s own documents.

AEMO’s 2024 ISP modelled a complete coal exit by 2037-38. The Draft 2026 ISP now pushes that full exit out to 2048-49, reflecting slower-than-planned closures in NSW and Victoria and changed state roadmaps, according to Infrastructure Australia’s 2026 analysis.

That eleven-year slip tells you the transition is already running behind its own policy ambition. If you are using the earlier 2037-38 timeline as your base case, you are working from an outdated map. The retirement schedule is the most concrete forward signal in Australian energy: it sets the floor on how much replacement capacity must be operational and by when, and that floor has just moved.

AEMO’s Draft 2026 ISP projects that reaching net-zero electricity will require 120 GW of grid-scale wind and solar, 32 GW of grid-scale batteries, and 12 GW of pumped hydro by 2050, a build rate that sets a formidable benchmark against which current delivery timelines can be measured.

The demand shock that nobody planned for at this scale

Here is where the arithmetic of the transition breaks from the familiar script. Most analysis treats demand as roughly stable while supply shifts underneath it. That assumption no longer holds, because of one line item growing faster than anything else on the grid.

Data centres are set to increase their electricity consumption almost sevenfold in a single decade. The scale of the movement, drawn from AEMO’s 2026 Electricity Statement of Opportunities (ESOO) as reported by The Age on 24 August 2026, runs as follows:

1. National electricity demand today: approximately 176 TWh (2025-26)
2. Data centre share today: approximately 3% of the grid, or roughly 5 TWh
3. National electricity demand by 2035-36: approximately 250 TWh
4. Data centre share by 2035-36: approximately 13% of the grid, or roughly 34 TWh

This is not incremental growth. It is a structural recomposition of what the grid must supply. National demand rising more than 40% over the decade is significant on its own; data centres accounting for a disproportionate slice of that growth relative to their current footprint is the part that rewrites the risk model.

The driver sitting behind the number is a pipeline of approximately 225 data-centre projects in development, which AEMO links to the forecast climb. The analysis from the Clean Energy Finance Corporation (CEFC) suggests this demand growth could lift wholesale electricity prices by more than 20% in NSW and Victoria by 2035 if clean supply and network capacity fail to keep pace, though that figure remains unverified against primary sources.

A near-sevenfold rise in data centre consumption arriving in the exact window coal is departing means the grid must do two historically large things at once: replace its biggest existing supply source and absorb its biggest new demand category. There is no recent domestic precedent for that combination. What this means for you as an analyst is simple. Any read of Australian energy that treats demand as roughly flat is materially mispricing the risk.

Why AI and cloud infrastructure are accelerating, not slowing

The demand is structural, not cyclical, which is what makes it hard to wave away. Hyperscaler and AI workload growth is driving the connection pipeline, with aggregate capacity requests reported to have grown toward 67 GW in recent tracking, though attrition on those requests exceeds 40%, meaning realised demand could land anywhere across a wide band. Both figures remain unverified against primary sources and should be treated with caution.

That band matters for near-term planning. EY’s 2026 report “Powering the AI and digital surge” places data centre consumption between 8 and 13 TWh by 2029-30 depending on transition speed, a spread that captures how much genuine uncertainty sits inside even the five-year view. The direction is not in doubt. The pace is.

The connection pipeline and grid infrastructure regulation governing how data centres actually connect to the NEM introduce additional uncertainty: approval timelines, network augmentation requirements and cost allocation rules all affect whether the 225-project pipeline translates into realised load at the pace AEMO’s demand forecasts assume.

Gas, batteries and the firming gap: what actually fills the space between now and 2035

This is where the planning documents stop agreeing with each other. The question of what fills the space between coal’s exit and a fully renewable grid is not settled. It is a live disagreement between planning optimism and supply-constraint realism, and the gap between the two is where the investment risk concentrates.

Gas is cast as the bridge. AEMO’s 2025 Gas Statement of Opportunities (GSOO) identifies a growing need for gas-powered generation (GPG) as Eraring and Yallourn close, flagging seasonal supply gap risks from 2028 and structural annual gaps from 2029. Left unaddressed, those gaps could reach 90-140 PJ per year from 2033 without new supply and storage, according to the GSOO.

The trouble is that gas faces its own supply crunch in the same years it is most needed as a coal backstop.

The structural nature of the east-coast gas supply crunch extends beyond seasonal shortfalls; domestic reserve depletion, LNG export commitments and underinvestment in new fields have combined to narrow the margin available for gas to perform its intended bridging role through the coal retirement window.

The ACCC’s June 2025 Gas Inquiry interim report projects an east-coast supply gap of approximately 55 PJ in 2028, rising each year to approximately 543 PJ by 2034.

That is the structural contradiction at the core of the plan. Gas is simultaneously the most available firming option and the one facing the most acute supply constraints. And the fragility is not purely a 2030s concern. The Grattan Institute’s 2025 Gas Market Review warns of seasonal shortfalls in Victoria and possibly NSW within the next three to four years, driven by cold weather, low renewable output and coal plant unavailability, which places the first real stress test inside the late 2020s.

The planning-preferred alternatives are batteries, pumped hydro and demand response. AEMO’s ISP scenarios lean heavily on these as core firming tools. The caveat is the one that matters most: those scenarios assume the technologies are delivered on time and at scale. If delivery slips, gas becomes the only backstop available at the speed the coal retirement schedule demands, and gas is the very thing the ACCC says is running short.

For an investor, this is where the transition risk is most concentrated. The answer to “what fills the gap” determines whether electricity stays reliable and affordable through the 2030s. It also determines which infrastructure assets carry the most upside and which carry stranded-asset exposure, a tension the next section takes head-on.

What the supply-demand math means for energy investors watching the 2028-2033 window

Pull the threads together and one period stands out as the system’s genuine stress test. The window from 2028 to 2033 is where every pressure converges at once.

Eraring and Yallourn close in 2028-2029. Gas supply gaps begin opening from 2028-2029 on the ACCC’s own timeline. Data centre demand is in its steepest growth phase, climbing toward the 13 TWh EY models by 2029-30, with roughly 6.2 GW of coal already retired by mid-2029. The renewable and storage pipeline that is supposed to absorb all of this faces real delivery risk across the same years. This is not a 2030s theoretical. It is a late-2020s operational reality.

The investment thesis that follows is deliberately two-sided, and the research supports neither a pure bull nor a pure bear case.

• Gas infrastructure and peaking assets carry near-term upside given the firming gap, but face stranded-asset risk beyond the mid-2030s if battery and renewable deployment accelerates as the ISP assumes.
• Renewable and storage assets carry the long-term structural tailwind from a demand base rising more than 40%, but are exposed to execution and grid-integration risk in the near term.
• The ISP timeline shift from 2037-38 to 2048-49 adds a third dimension: extended coal operation could suppress the wholesale price signals that would otherwise pull renewable investment forward, creating a potential policy-induced investment trap.

Three variables are worth tracking more closely than any planning headline:

1. Coal retirement adherence: whether the named stations actually close on their dated schedule or slip further, as recent ISP revisions suggest they might.
2. Gas supply gap realisation: whether east-coast supply tracks toward the ACCC’s 543 PJ by 2034 projection or new supply and storage close the gap.
3. Renewable-plus-storage pipeline delivery: the rate at which firming capacity is actually commissioned versus assumed in AEMO’s scenarios.

What this tells you is that the supply-demand math does not just describe a system under strain. It identifies where shortfalls are structurally likely, which assets are most exposed, and which categories of infrastructure sit on a genuine physical demand base rather than policy assumption alone.

The stranded-asset question on gas infrastructure

The gas debate resolves to a timing mismatch between near-term necessity and long-term exposure. In the late 2020s, gas is the backstop the retirement schedule demands, and assets that can firm the grid through the gap carry real value. Beyond the mid-2030s, the picture inverts.

The CEFC’s framing is explicit in its preference for assets aligned with long-term net-zero pathways over new, long-lived fossil fuel infrastructure. Analysts echo the warning that new gas fields, pipelines and large GPG capacity built to cover short-term gaps risk underutilisation if batteries, demand response and renewable overbuild reduce peaker need through the 2030s. The risk is not whether gas is needed now. It is how long that need lasts, and whether assets with multi-decade lifespans can earn out before the system moves past them.

Battery cost trajectories are the variable with the largest swing effect on whether gas peakers earn out their capital over a multi-decade asset life: if GenCost’s projections hold and storage costs continue falling at their current rate, the business case for new long-lived gas infrastructure weakens materially through the 2030s.

Reading the grid’s stress signals before the market does

The core argument is not that Australia’s energy transition is failing. It is that the timing of three movements, supply exit, demand acceleration and replacement delivery, is converging into a discrete reliability and price risk window from the late 2020s through the early 2030s. The transition is happening. The question is whether it arrives fast enough in the years it is needed most.

That is why the policy uncertainty deserves active attention rather than deference. The ISP timeline shifting eleven years, the ACCC’s gas inquiries and the state-level roadmap changes are not background noise. They are the indicators that tell you which delivery assumptions are holding and which are slipping, and they move faster than any planning document’s conclusions.

The sharpest frame to carry forward is this. The supply-demand math already exists in public documents, visible to anyone willing to read them. The analytical edge lies in recognising that the planning-optimist and supply-constraint-realist views cannot be reconciled without knowing which delivery assumptions actually hold, and that answer will reveal itself station by station, gap by gap, between now and 2033.

This article is for informational purposes only and should not be considered financial advice. Investors should conduct their own research and consult with financial professionals before making investment decisions. Past performance does not guarantee future results, and financial projections are subject to market conditions and various risk factors.