The Australian Energy Ledger · demand outlook
Australian Energy & Data Centres
Grid demand in the National Electricity Market has been flat for a decade. AEMO's 2026 forecast breaks that line, and the fastest-moving single component is data centres. This page sets what has actually been metered against what AEMO now forecasts to 2035–36, and separates the 67 GW that has been announced from the far smaller figure the market operator expects to be built and used.
- Source
- 2026 NEM ESOO, Australian Energy Market Operator
- Market
- National Electricity Market
- Metered
- 2019–20 → 2025–26
- AEMO forecast
- to 2035–36
- Updated
- 2026-08-30
Electricity drawn from the grid, net of what consumers send back. Rooftop solar generation consumed on site is not in here.
Flat for a decade, then up 42 per cent in ten years
Grid-supplied consumption in the NEM was 177.8 TWh in 2019–20 and 175.7 TWh in 2025–26: six years, no growth. Rooftop solar absorbed everything population and the economy added. AEMO's 2026 NEM ESOO forecasts that line breaking, with consumption reaching 249.4 TWh by 2035–36. Data centres are the fastest-moving single component, going from 5.1 TWh to 33.7 TWh over the same period, or 2.9% of consumption to 13.5%.
NEM grid-supplied consumption, data centres broken out
terawatt hours per financial year| Year | Total TWh | Data centres TWh | Share | DC average GW | Basis |
|---|---|---|---|---|---|
| 2019–20 | 177.8 | 1.8 | 1.0% | 0.21 | metered |
| 2021–22 | 175.0 | 2.5 | 1.4% | 0.28 | metered |
| 2023–24 | 174.2 | 3.3 | 1.9% | 0.38 | metered |
| 2025–26 | 175.7 | 5.1 | 2.9% | 0.58 | metered |
| 2027–28 | 182.8 | 9.3 | 5.1% | 1.07 | AEMO |
| 2029–30 | 194.5 | 15.5 | 8.0% | 1.77 | AEMO |
| 2031–32 | 209.5 | 20.8 | 9.9% | 2.38 | AEMO |
| 2035–36 | 249.4 | 33.7 | 13.5% | 3.85 | AEMO |
What network businesses have been asked to connect, against what AEMO's forecast actually lets through.
67 GW announced, about 8.0 GW in the forecast
Network businesses reported 225 data centre projects in the connection process, carrying 67 GW of proposed capacity between them, up from 38 GW a year earlier. AEMO's central forecast lets through the equivalent of about 8.0 GW, roughly 12% of the queue. Three things drive the discount: 36% of last year's projects have been cancelled, a mature fleet draws only 45% to 51% of its contracted capacity, and operators say a facility takes 5 to 10 years to reach capacity.
Announced connection capacity against forecast need
gigawatts of connection capacity| Measure | GW | Note |
|---|---|---|
| Proposed, all development stages | 67 | 225 projects |
| Proposed a year earlier | 38 | 76% growth in twelve months |
| Implied by AEMO's 2035–36 forecast | 8.0 | at 48% mature utilisation |
| Operating today | 2.2 | drawing 27% of it |
Annual energy divided across every hour of the year. Data centres run at about a 90 per cent load factor, so their average sits close to their peak.
A 3.8 GW block of demand that never steps off
Data centres are close to the only load on the system whose average and peak are nearly the same number. At a 90% load factor, the 33.7 TWh AEMO forecasts for 2035–36 is 3.8 GW sitting on the grid in every hour of the year, against the NEM's present average of 20.1 GW. If the whole announced queue were built and reached mature utilisation it would be about 32 GW, or 282 TWh a year, which is more than the entire NEM consumes today.
The same numbers as round-the-clock demand
gigawatts, average across the year| Case | GW | TWh a year |
|---|---|---|
| Data centres operating today | 0.58 | 5.1 |
| Data centres, AEMO 2029–30 | 1.77 | 15.5 |
| Data centres, AEMO 2035–36 | 3.85 | 33.7 |
| Whole announced queue, if built | 32.2 | 282 |
| Whole NEM today | 20.1 | 175.7 |
This forecast has been raised in each of the last three years. Worth knowing before treating the current one as settled.
Raised in each of the last three years
The 2024 ESOO put data centres at about 5 TWh by 2033–34 in its central case. The 2025 ESOO forecast 21.4 TWh by 2034–35. The 2026 NEM ESOO forecasts 33.7 TWh by 2035–36, a level the 2025 vintage did not expect until the middle of the following decade. The scenario spread is wide in the same direction: 52.5 TWh by 2035–36 under Accelerated Transition, 15.1 TWh under Slower Growth.
Data centre consumption by scenario and forecast vintage
terawatt hours per financial year| Year | Slower Growth | Step Change | Accelerated | 2025 ESOO |
|---|---|---|---|---|
| 2029–30 | 10.5 | 15.5 | 20.1 | 11.0 |
| 2034–35 | 14.3 | 30.0 | 45.5 | 21.4 |
| 2035–36 | 15.1 | 33.7 | 52.5 | 22.2 |
AEMO does not plan on a baseload basis and publishes no series under that name. Three quantities carry the meaning between them.
What "baseload" means once the load is flat
AEMO retired the word. No published series is called baseload, and the three quantities that carry the meaning between them now point in different directions.
Annual energy is what generation must supply across the year, and it is rising: 175.7 TWh now to 249.4 TWh by 2035–36.
Minimum operational demand is the daytime trough, the closest thing to a floor under the system. Rooftop solar has been driving it toward zero, and South Australia reached 57 MW in the June quarter of 2026. AEMO's 2026 forecast raised its minimum-demand outlook for the first time in years and names data centres as a key reason: they run straight through the solar peak.
Maximum demand is still set by hot evenings and, increasingly, cold winter mornings, when rooftop solar contributes nothing at all.
Data centres lift all three at once, which is unusual. That makes them easier to serve in one sense, since there is no peak to chase, and harder in another: the energy has to come from somewhere in every hour of the year, over a decade in which about 15 GW of coal and gas is scheduled to retire.
Western Australia runs a separate grid, forecast separately.
Western Australia, and where the load lands
Western Australia's South West Interconnected System is a separate market with a separate forecast. It is projected to grow from 20.4 TWh to 32.1 TWh by 2035–36, of which data centres are only about 0.8 TWh: strong investor interest, few well-progressed projects.
Within the NEM the load is concentrated. New South Wales and Victoria account for more than 85% of forecast data centre consumption through 2035–36, and about 66% of what is consumed today sits in New South Wales alone. AEMO notes it has since seen more Queensland interest than these figures carry.
The figures above, stated plainly.
Common questions
How much electricity do Australian data centres use?
Data centres connected to the National Electricity Market consumed about 5.1 TWh in 2025-26, which is 2.9% of grid-supplied electricity. AEMO forecasts that reaching 33.7 TWh, or about 13.5%, by 2035-36.
How much power does Australia's grid supply in total?
The NEM supplied 175.7 TWh in 2025-26. That figure has been broadly flat since 2019-20, because rooftop solar has offset growth from population and the economy. AEMO forecasts 249.4 TWh by 2035-36, an increase of about 42 per cent.
How many data centre projects are waiting to connect in Australia?
Network businesses reported 225 projects at various stages of the connection process in 2026, carrying 67 GW of proposed connection capacity between them, up from 38 GW the previous year. Around 36% of the projects on the previous year's list had been cancelled, including more than 30% of those classified as committed.
Will data centres consume 67 GW of Australian electricity?
No. The 67 GW figure is proposed connection capacity in the queue, not demand. A mature data centre draws 45% to 51% of its contracted capacity, many projects are cancelled, and those that proceed take 5 to 10 years to ramp. AEMO's central forecast implies about 8.0 GW of connection capacity by 2035-36, roughly 12% of the queue.
Are data centres baseload electricity demand?
In shape, yes. AEMO measures currently operating Australian data centres at a load factor near 90%, with only minor variation between working and non-working hours. That is flatter than any other significant load class on the grid, and it is why data centres raise minimum demand as well as annual energy.
Does data centre growth threaten grid reliability?
AEMO tested it. Its high-growth sensitivity brings forward the point at which New South Wales, Victoria and South Australia exceed the reliability standard by about a year, to 2032-33, and concludes that reliability holds if the committed and anticipated projects are delivered on time and no coal exits early. Roughly 15 GW of coal and gas is scheduled to retire over the decade.
Where in Australia are the data centres?
New South Wales holds about 66% of current NEM data centre consumption and Victoria about 30%, with the remainder in South Australia, Queensland and Tasmania. Those two states account for more than 85% of forecast consumption through 2035-36. Western Australia's separate grid expects only about 0.8 TWh by then.
Where this comes from, and what it can't tell you
This page carries forecasts, and says so. The other data pages on this site are compiled from published statistics and contain nothing modelled. This one is different, and the distinction is marked on every figure: metered history to 2025–26, then AEMO's published forecast to 2035–36. The line stops there rather than running on to 2050. The machine-readable file covers exactly the same span and carries a basis column on every row.
What is being counted. AEMO's delivered consumption for the NEM: electricity drawn from the grid by consumers, net of what they send back. This is the basis AEMO quotes its data centre shares against. It excludes rooftop solar consumed on site. Underlying consumption, which includes it, runs 204.2 TWh in 2025–26 rising to 308.4 TWh by 2035–36.
The history is metered, then rebased. AEMO publishes its consumption actuals as charts rather than tables, so the history here is built from AEMO's own aggregated price and demand files, which give metered operational demand for every settlement interval in each region. Those are summed to financial years and scaled by a constant factor of 0.9570 to meet AEMO's published 2025–26 delivered figure of 175.7 TWh; the constant stands in for auxiliary load and network losses. The settlement interval is not constant across the period, moving from 30 minutes to five in October 2021, so each file's interval is read from its own timestamps. The shape, a decade of flat demand, is metered rather than modelled.
The forecast is AEMO's, interpolated between published anchors. AEMO publishes anchors, not a year-by-year table: 175.7 TWh rising to 249.4 TWh in total, and 5.1, 15.5 and 33.7 TWh of data centre load in 2025–26, 2029–30 and 2035–36. Intermediate years are smoothed between those anchors, with growth weighted to the 2030s as AEMO describes. Data centre history before 2025–26 is interpolated between Oxford Economics' estimates of 1.4 TWh in 2017-18 and 3.9 TWh in 2024-25.
Why the line stops at 2035–36. The series stops at 2035–36 because that is where AEMO's published forecast stops. The 2026 ISP does carry a 2050 figure, but its demand outlook was settled before this year's data centre revision and already sits about 20 TWh below the ESOO by 2035–36. Running the line out to 2050 would therefore mean publishing a decade of our own arithmetic alongside AEMO's, which is a different kind of claim. Every number on this page is traceable to a published source.
Gigawatts here are average load, not plant capacity. Every gigawatt figure on this page is average load, annual energy spread across the year, not installed plant capacity and not peak demand. For data centres the two are close because the load is flat. For the grid as a whole they are not: the NEM's 20.1 GW average sits well below the capacity needed to meet its evening peak. Connection capacity is converted at AEMO's mature utilisation assumption of about 48%.
Source terms and attribution. AEMO holds copyright in its publications and grants use of that material for any purpose with attribution, which is why this page and its data file name AEMO on every figure drawn from it. That permission is compatible with the Creative Commons Attribution 4.0 licence the compiled file carries, but it is not itself a Creative Commons licence: if you reuse the underlying figures, attribute AEMO rather than this page. Data centre history is from Oxford Economics Australia's report for AEMO. AEMO asks that conclusions drawn from its data not be attributed to AEMO, and none of the analysis here is theirs.
The queue is not a forecast. Connection capacity advised by network businesses is what developers have asked for, not what will be built. AEMO records that 36% of the projects on its previous list have since been cancelled, including more than 30% of those the networks had classified as committed. Treat the queue as an upper bound on interest, never on demand.
Revisions. AEMO republishes the ESOO each August and the ISP every two years. This page is rebuilt from those editions rather than edited by hand, so the anchors move together.
- 2026 Electricity Statement of Opportunities, Australian Energy Market Operator (2026-08-25)
- 2026 Integrated System Plan, Australian Energy Market Operator (2026-06-30)
- Data Centre Energy Demand, Final Report, Oxford Economics Australia for AEMO (2025-07-25)
- 2026 WEM Electricity Statement of Opportunities, Australian Energy Market Operator (2026-06-11)
- Aggregated price and demand data, Australian Energy Market Operator
- Machine-readable series (CSV) and full series (JSON), compiled for this page