Energy considerations
Increasing generation for increasing demand
Data centres represent a significant demand load—one likely to fundamentally influence how Australia's electricity market operates in coming years. In 2024, data centres already consumed ~5% of Australia's electricity, and this consumption is expected to reach anywhere from 8%-20% by 2030. This increase in total required electricity will need be matched by a corresponding increase in electricity generation (most likely renewable as the push for decarbonisation continues). Analysis by Morgan Stanley predicts that a bull market case represents 20% compound annual growth, a base market case represents 13% compound annual growth, and a bear market case represents 8% compound annual growth.
Stabilising the grid
Whilst data centres do perform some interruptible computing tasks, they largely require a continuous supply of energy to operate. This demand increase will result in greater strain on the grid during periods of peak demand. However, the continuous demand may provide stability for the grid during periods of high generation currently unmatched by demand (eg during the midday solar and wind peak). These high generation windows are traditionally problematic for non-dispatchable generation projects which face the risk of negative electricity prices and/or being curtailed by AEMO in order to stabilise the grid.
Co-location with generation assets
The co-location of data centres with hybrid renewable energy projects (solar or wind farms connected to battery storage) may be a promising solution for multiple stakeholders. From the perspective of a data centre owner, co-location offers a practical solution to meet a range of operational considerations, including energy needs, emissions targets and financial optimisation.
Co-location also poses benefits to potential developers of new generation assets because supplying electricity directly to a data centre behind the meter can alleviate curtailment risk, thereby increasing the profitably of the project and supporting the developers case for financing support. Another consideration is that co-location may offer a competitive advantage for projects in the highly competitive renewable energy zone (REZ) access rights tenders. Given the majority of the generation capacity will be consumed by the co-located data centre, these projects may be able to argue they will have lower transmission capacity requirements (thereby allowing the Government to award the limited transmission capacity to more projects and thus incorporate greater total generation capacity within the REZ).
Planning and zoning
Not all jurisdictions are created equal
The approval strategy for data centres differs depending on location. Jurisdictions take varying approaches to permissibility, assessment against planning controls and infrastructure charges, and the availability of approval pathways. Aside from planning approval, the extent of other approvals and licences required to construct and operate a data centre also varies by jurisdiction. From an ease of securing approvals perspective, this makes some jurisdictions more attractive or easier to navigate than others for data centre development.
In NSW and Victoria for example, data centres are specifically recognised as a class of development with local or state-assessed planning approval pathways depending on certain thresholds. The state planning approval pathway in NSW comes with industry-specific assessment requirements, making it clear from the outset what level of impact assessment is required. An advantage of a state pathway is having a sophisticated and well-resourced consent authority assess the development application, since it understands the broader economic benefit to the economy. It can be more difficult to navigate the assessment process with local councils.
In other jurisdictions like Queensland, for example, there is no specific definition for data centres and there are no standard assessment requirements. The level of assessment for data centres in Queensland is not consistent across the state and depends on the planning scheme for each local government area.
Because data centres do not operate like standard industrial or commercial uses, it often proves difficult to apply some aspects of the local controls, such as minimum car-parking rates, which are calibrated to an assumption about staff numbers that reflects more traditional warehousing or office developments. Likewise, data centre proposals in built-up commercial zones often face added scrutiny in terms of visual and amenity impacts.
Infrastructure charges are typically levied by reference to gross floor area or cost of development. Neither method is a neat fit for data centres. While the built form of data centres may be relatively inexpensive, the internal fit out and plant/equipment cost is typically significant and can distort the infrastructure charges payable if they are calculated by reference to the cost of development.
ESG considerations
New risks rear their heads
Despite a quieting of the external narrative surrounding ESG, the area remains a priority for regulators and civil society. Data centres should remain alert to the potential for sustainability and human rights impacts in their operations and supply chains, and have in place appropriate risk mitigation and management systems.
- The sustainability implications of generative AI mean there is growing pressure on data centres and tech companies to make the housing of data as 'green' as possible.
- Companies should also be alert to potential human rights impacts—eg when it comes to the training of AI, the sourcing of critical minerals in supply chains, and human rights and privacy considerations.
- To mitigate enforcement and litigation risk, it is important for boards, legal and sustainability teams to be assessing and monitoring potential impacts, to have in place tailored systems and controls to respond to issues, and make sure commitments are understood and implemented effectively.
Regulator and civil society focus on ESG has significance for data centres in the context of heightened scrutiny of energy and water usage. It is reported that, despite increases in energy efficiency, data centre energy consumption worldwide accounted for approximately 3% of global emissions in 2023,1 and data centre energy demand is projected to double by 2030 due to the additional load posed by AI.2 Google's 2024 Environment Report suggested its data centre water consumption has increased by nearly 88% since 2019.3 Tech companies have also come under scrutiny for emissions arising from the use of in-house and third-party data centres.
The training and use of data for generative AI purposes can also potentially present risks for workers in the technology supply chain, including due to the potentially exploitative, labour-intensive work of training AI models and the risk of workers being exposed to harmful content. AI systems themselves can be discriminatory, eg as vulnerable individuals and communities may be impacted by emerging issues like facial recognition discrimination.4 The trend towards data localisation can give rise to human rights risks if data centres are set up in states with poorer human rights and privacy records.
Footnotes
-
https://kpmg.com/dp/en/home/insights/2024/01/going-green-data-centres.html
-
https://www.theguardian.com/technology/2024/sep/15/data-center-gas-emissions-tech
-
https://sustainabilitymag.com/articles/the-real-cost-of-meta-google-microsofts-ai-investments
-
https://www.wearehumanlevel.com/content-hub/the-ai-threats-to-climate-action-and-human-rights