AEMO's Data-Centre Forecast Doubles Down on a Grid Already Living Off Directions
AEMO's latest ESOO projects data-centre demand rising from 5 TWh to 34 TWh by 2035-36, even as its own reliability record shows how thin the margin has become.
The Australian Energy Market Operator published its annual Electricity Statement of Opportunities on Tuesday (2026-08-18), projecting data-centre electricity demand in the National Electricity Market will climb from roughly 5 TWh to 34 TWh by 2035-36, a near sevenfold increase over a decade.3
The same AEMO underwriting the reliability case for the next ten years is the operator that had to suspend the entire NEM in a recent stress event, issuing around 500 directions to more than 5 GW of generation plant to keep supply intact.1 The ESOO is a planning document, not a supply guarantee, and the gap between the two has already been tested.
The headline numbers in the ESOO are, on their face, reassuring. Around 40 GW of new generation and storage is now committed or anticipated, with a further 33 GW sitting behind government programs, against a total NEM of 77 GW.2 Roughly 9 GW was added over the past year, a yearly record, and 24 GW has moved into committed or anticipated status since last year's ESOO.2
But the ESOO stresses that timely delivery and operational availability remain the binding constraints, not the size of the queue.4 AEMO has made that point in successive annual reports, and it has not stopped ageing coal from setting the timetable. Nearly 40% of the NEM's coal fleet has retired since market start, and the average age of remaining stations is 38 years.1
The data-centre forecast carries the least developed track record of any figure in the report. AEMO's projection is built on disclosed and pipeline data-centre data, and hyperscaler construction schedules have a habit of slipping. If the 34 TWh arrives late, the reliability arithmetic shifts, but so does the investment case for some of the 33 GW of government-backed capacity that has not yet reached committed status.2,3
A second layer of the demand picture receives less attention in the ESOO's summary materials. More than 4 million rooftop generators now sit behind one in every three Australian homes, and at times those consumer resources meet more than 60% of NEM demand, with total capacity larger than the remaining coal fleet.1 Genuine supply-side change, but variable, largely invisible to AEMO's central dispatch, and concentrated in the middle of the day when data centres need flat, round-the-clock load.
The combination is uncomfortable. The NEM has a record build pipeline and record consumer-side capacity, yet the operator has already demonstrated it can be forced into a market suspension when weather, outages and fuel availability line up badly.1 Data centres do not flex down when a cold snap hits. They add firm, high-utilisation load that demands the same characteristics on the supply side, and storage in the current pipeline is not yet sized to that shape.2
Commodity prices visible on 2026-09-11 add context without resolving the tension. Newcastle thermal coal at $140.75/t keeps the remaining coal units commercially viable for now, which delays retirement decisions rather than accelerating them. Wallumbilla gas was priced at A$11.12/GJ, keeping gas-fired peaking expensive as a firming backstop. JKM Asian LNG stood at $24.81/MMBtu, elevated by historical standards, which limits the appetite for new gas-to-power commitments tied to spot LNG supply.3,1
The trade implication is not a directional NEM power call. It is a timing call. The ESOO's reliability case rests on 40 GW of committed capacity arriving on schedule and on the 33 GW of government-supported projects converting to committed status without slipping.2 AEMO will publish the next ESOO in 2027; between now and then, the data-centre connection queue and the conversion rate on government-backed capacity are the two figures that move the reliability assessment.
The unresolved risk is that both slip in the same year. AEMO has already run that stress test in real time, and the answer was 500 directions and a market suspension.1 A 34 TWh data-centre load added onto that grid, delivered on a slower timeline than the coal exit, is the scenario the ESOO's headline 40 GW does not fully cover.3,2