Quick insights
- Data centres are fundamentally different to traditional large industrial loads connected to the electricity system due to their volume, size and pace of development.
- AusNet has developed a survey-based forecasting framework that includes criteria tailored to data centre load so the electricity sector has visibility of projects in the pipeline that are genuinely likely to proceed.
- The Australian Energy Market Operator (AEMO) is considering our framework in a formal submission to its Demand Forecasting Methodology Consultation.
- Closer collaboration is required between the electricity and data centre sectors to remove barriers to grid connection and enable Australia to benefit from this significant economic opportunity.
Data centres are a critical part of our digital lives. Every time you search the internet, upload an image or video to social media, or ask AI to perform a task – you are accessing secure data storage and computing services facilitated by data centres.
There are three common types of data centres, each with their own grid-connection requirements:
- Hyperscale data centres require connection into our high-voltage transmission network because they need to progressively scale their electricity consumption over time and have very high reliability requirements.
- Edge data centres are smaller facilities that provide time sensitive information, often with a degree of flexible demand. As a result, they typically require connection to the lower-voltage electricity distribution network.
- Enterprise data centres sit in between these two and are typically connected to the transmission network.
Figure 1: Common types of data centres:
At the start of 2023, transmission-connected data centres were an emerging trend being considered alongside other more well-established electricity demand drivers, such as electrification of transport and the gas sector. Most enquiries were edge distribution-connected facilities built by established data centre players. Since then, transmission-connected hyperscale data centres have experienced a surge in investment, driven by an exponential increase in demand for cloud-based digital capacity and artificial intelligence.
Data centre connections are an enormous economic opportunity for Australia. As a general rule it costs between $9 million to $14 million per megawatt (MW) of commissioned IT load to build a complete data centre.1 This means a 300 MW hyperscale data centre development would result in between ~$3-4 billion in construction capital spent.2 A recent report projected Australia will have 3,100 MW of deployable data centre capacity by 2030, generating $26 billion in direct investment.3 This figure appears conservative given the level of data centre interest in Victoria alone.
AusNet has a dedicated team to support its load customers through the connection process. We are working with data centre proponents who are investing heavily to secure sites, planning approval and grid connection as quickly as possible.
Alex Berce, Head of Business Development – Load Connections
Taking advantage of the economic benefits from hyperscale data centres requires the electricity sector to maximise existing network hosting capacity and address electricity capacity challenges. A key step in this process is accurately forecasting electricity demand driven by data centre connections so networks can appropriately plan for load growth, including generation.
AusNet has been exploring improvements the electricity sector can make to increase confidence in longer-term data centre load forecasts. We learnt that data centres are fundamentally different to traditional large industrial loads (LILs) that AEMO and networks forecast and plan for. This is due to:
These differences require us to ask questions we would not have previously contemplated with other LIL customers. For example:
- What data centre projects should be included in central demand forecasts that inform network planning investment decisions?
- When should they be included in the project life cycle?
- Should the electricity sector accept the ‘ultimate load’ and ‘ramping’ proposed by data centres into the future?
- Should the expected aggregate annual consumption be delayed or reduced?
- How might data centres impact the long-term electricity supply-demand balance?
AEMO is aware of these challenges and sought feedback on data centre forecasting in its 2024 Electricity Demand Forecasting Methodology (EDFM) Consultation. In November 2024, as part of its annual load consumption forecasting survey (questionnaire) process, AEMO proposed to separately forecast data centres as its own customer segment independent from other LILs. It also proposed to expand the criteria used to determine whether LIL should be included in demand forecasts (i.e. the load commitment criteria). These criteria were proposed to reflect the different development stages of a LIL project lifecycle, such as proposed, anticipated, and committed.
While we support the intent of these reforms, for the reasons explored earlier, we see benefits in developing a survey-based forecasting framework that includes commitment criteria tailored to data centre load rather than adopting the same criteria for all LIL projects.
We outlined this framework in detail within our formal submission to the EDFM Consultation submitted in January 2025, which is available for download above. It introduces five stages of development to capture a data centre’s progression, starting with ‘publicly announced’, where there is no demonstrable progress through to ‘committed’, which aligns with the current definition used within AEMO planning documents. This will allow improved long-term load forecasting which will enable network solutions to be implemented.
Figure 2: Draft framework with a set of inclusion criteria dedicated to data centre load
Projects will be assessed as having achieved a stage of development (i.e. status) if they achieve all five criteria - land, grid connection, planning approvals, construction and finance - in the relevant column or better. For example, a data centre load is only ‘anticipated’ if evidence confirms the project has:
- a land options agreement (or ownership)
- submitted a connection application (or obtained a 5.3.4A letter)
- a submitted or approved planning permit
- finalised site plans or construction commenced
- executed early works contracts or (financial investment decision).
Our forecasting framework was developed following consultation with a sample of data centre industry participants. This engagement allowed us to better understand market dynamics and test what objective evidence is readily available to demonstrate a data centre project's development status.
Jason Jina, Policy & Reform Manager
Our forecasting framework has a higher threshold to reach ‘anticipated’ status than the criteria proposed in AEMO’s EDFM. As a result, we are better placed to provide visibility of developments that are genuinely likely to proceed and improve the accuracy of electricity demand forecasts. Meeting all five criteria to reach ‘anticipated’ status under our framework would require a data centre proponent to invest tens of millions of dollars in development funds, which provides comfort they are unlikely to cancel their development plans. The additional stages of development status (i.e. ‘publicly announced (1) and (2)’, and ‘in development’) provide additional visibility and granularity about the data centres likely to mature over the next 4-5 years and assists the electricity sector to respond to the pace of data centre development.
The electricity sector is grappling with how to tailor their plans and processes to connect data centre projects. Closer collaboration between the two industries is critical to addressing barriers to connection and Australia benefiting from this significant economic opportunity.
Laura Walsh, GM Network Management
The framework offers some flexibility around how project status could inform data centre forecasts for various growth scenarios – for example, AEMO’s Central scenario in its Integrated System Plan. Our view is that including data centre projects with a ‘committed’ or ‘anticipated’ status will produce a forecast that is closer to the demand that is “most likely to eventuate.” Only including ‘committed’ projects in the Central economic growth scenario would underestimate forecast demand from data centre connections.
For more information, see Data centres.
Glossary
- Large industrial loads (LILs) – these are industrial electricity customers who are connected to our transmission network or are above a demand threshold of 10 megawatts (MW) on the electricity distribution network. These customers are typically involved in aluminium and steel production, or mining. We are seeing new and emerging LILs from hydrogen electrolysers and data centres.
- Data centre – this is a physical facility storing computing and networking equipment and their related hardware used to collect, process, and store digital data, as well as to distribute and enable access to digital resources.
Footnotes
1 AusNet internal.
2 Construction costs is predominantly electrical systems (up to 45%), followed by building fit-out, land and mechanical equipment.
3 Mandala, Empowering Australia’s Digital Future, 2024.