Australia’s energy market operator (AEMO) is, for the first time, breaking out data centres as a distinct category in its national electricity demand forecasts – a shift that could have real implications for how operators plan, expand and position themselves in the years ahead. The move signals that data centre electricity demand is no longer just a footnote in the broader commercial and industrial load mix.
Until now, data centres were modelled alongside a range of unrelated loads, obscuring their individual impact on the grid and on the National Electricity Market’s planning horizons. In its newly finalised methodology, AEMO has committed to measuring the sector with the same attention it applies to heavy industry, but using a bespoke mix of modelling tools to reflect the diversity of facilities and the volatility of the market.
At the heart of this new approach is a split between commercial-scale data centres – those under 10 MW – and industrial-scale facilities of more than 10 MW. The smaller category will be forecast in aggregate, starting with short-term econometric models built from historical consumption data matched to economic activity, and then transitioning to a techno-economic model for the medium and long term. This blend is essentially designed to capture the inertia of existing demand patterns while also accounting for step changes in technology adoption.
The large-scale end of the market will be treated more like other major industrial loads. Here, AEMO is leaning on a combination of direct information from data centre builders and operators, responses to its large industrial load surveys, standing information requests to network service providers, market research and even media monitoring.
Real-world data
It reckons this allows the forecasts to incorporate real-world project data like expected ramp-up periods from connection to full load, load realisation factors that temper rated capacity to more realistic operating levels and load factors that describe the ratio of average to peak demand.
By modelling these parameters, AEMO can produce consumption forecasts that better reflect the staged nature of facility commissioning, the technical realities of operations and the likelihood that announced projects will actually be built. Projects will be assigned a status – committed, anticipated, or proposed – based on defined criteria, and only certain categories will be included in each forecast scenario.
For example, under its “central” and “high” economic growth scenarios, anticipated projects with a grid connection application are included up to the reliability obligation threshold; beyond that point, all anticipated projects are included, but proposed projects with a connection application appear only in the “high” scenario. This is a deliberate tightening of the rules compared to other industrial loads, reflecting the high uncertainty and speed of change in data centre development.
Eye-watering numbers
The numbers AEMO is working with are eye-watering. The operator estimates that data centres consumed around 4 TWh in FY 2025, roughly 2.2 percent of total NEM demand. In its Step Change scenario, demand could grow at around 25 percent a year, hitting 12 TWh by 2029-30 – about 6 percent of NEM consumption. That’s a tripling of the sector’s share in just five years, and it sets the stage for data centres to become one of the most influential demand drivers in the grid’s near-term evolution.
This sharper focus on the sector did not emerge in a vacuum. Victorian electricity transmission network operator and energy infrastructure company AusNet, in its submission to AEMO’s consultation process, argued strongly for a dedicated data centre forecasting framework. It proposed a survey-driven model with clearly defined project stages and milestone criteria – ranging from land acquisition to financial close – to ensure that only projects with a realistic chance of proceeding were baked into the numbers. The goal was to avoid speculative load inflating forecasts and, by extension, potentially distorting grid planning and investment decisions.
Hybrid approach
AEMO’s final methodology takes some of these ideas on board. The decision to treat data centres as a separate segment mirrors AusNet’s position, as does the use of survey data for large-scale projects. But the operator stopped short of adopting AusNet’s full staging model. Instead, it opted for a hybrid approach that blends econometric and techno-economic modelling with project-specific intelligence, using its own criteria for assigning project status.
This approach gives AEMO more flexibility to adapt as the sector evolves, but it may also mean that operators will need to be proactive in supplying credible, detailed project information if they want their developments to be accurately reflected in official forecasts.
For data centre operators, the takeaway is clear: the sector is now on AEMO’s radar in a way it never has been before. Forecasting will increasingly be grounded in specific project data, and the growth numbers suggest the industry will play a pivotal role in shaping the NEM’s generation mix, transmission planning, and reliability settings over the next decade.
In practical terms, this could influence not just how quickly new capacity can connect to the grid, but also the policy and market mechanisms that frame those connections. In short, the way the grid sees data centres is changing – and operators who engage early and transparently in the forecasting process are going to be better equipped to navigate that change.
Event details: Sydney International Convention Centre, 21 August 2025, 8:00am–8:30pm.
Register here: https://clouddatacenter.events/events/sydney-cloud-datacenter-convention-2025/
