On Monday, the Australian Government published its Expectations of Data Centres and AI Infrastructure Developers. The expectations set out the government's position on what it expects from the sector. They effectively codify the social licence to operate for data centres, while leaving implementation to future processes.
The expectations are:
Expectation 1 — National Interest: prioritise Australia's national interest, including national security and data sovereignty.
Expectation 2 — Energy transition: support Australia's energy transition, including securing new clean energy generation, covering grid infrastructure costs, and enhancing demand flexibility.
Expectation 3 — Water: use water sustainably and efficiently, including investing in non-potable water sources and efficient cooling technologies.
Expectation 4 — Skills: invest in Australian skills and workforce capability, including apprenticeships, structured training pathways and collaboration with education providers.
Expectation 5 — Local R&D and capability: support research, innovation and local capability, including enabling access to compute for Australian start-ups, researchers and not-for-profits on favourable terms.
Without analysing each expectation in detail, our key observations for the current market include:
- Trade-offs
The expectations involve inherent trade-offs. National interest, location, energy and water dynamics and social and economic goals cannot all be optimised independently. Developers will need to strike an appropriate balance.
- Effectively mandatory
Many of the matters were already expected in practice. The expectations now make this clearer. While they are not yet legally mandatory, they are effectively immediately mandatory from a policy perspective. Projects that do not address power and water self-sufficiency, efficiency technology, national security and local capability are unlikely to progress.
The levers of approval remain to be developed by governments of different tiers, but the Expectations are effectively mandatory already. It is not an option to not meet or exceed these expectations – other projects will not happen (although there is some scope to position the message).
- Integrated issue, system design and compliance
Compliance is complex and multi-factorial, and compliant solutions should ideally be architected as a holistic scheme. Retrofitting is possible, but requires urgent and material action, and may change project economics, design choices and market deployment.
- Public procurement mindset
Although datacentres are not publicly procured, developers should adopt a similar mindset. This means proactively presenting solutions which demonstrate how the expectations are met or exceeded. Stronger alignment is likely to lead to more favourable the regulatory and approval tailwinds for the development – these need to be crafted to be as demonstrably compliant and straightforward as possible.
- AI turbo charges energy transition
AI offers a solution for energy security, not just a problem. Used effectively, AI can significantly reduce energy and water use in data centres when combined with CPUs. Moreover, the datacentre industry buildout represents opportunity to step-change into grid stablisation role. This requires holistic consideration.
- Siting, timing and flexibility
Location decisions will need to balance competing factors, including proximity to energy sources, delivery timeframes, cost, reliability and data latency. A more flexible approach may better meet the thrust of the expectations, AI deployment, security, the energy transition and the grid.
- Engagement
Early and constructive engagement with government and regulators will be important. Developers who demonstrate intent to comply can help shape any future formal guidance in ways that are workable for industry and achieve firm social licence and regulatory acceptance and expedition.
There is much at play here, and no silver bullet. Trade-offs are inevitable and flexibility and iteration essential as we create the infrastructure of the future while flying it. From here, this is the slowest we will ever move.
G+T is at the leading edge of datacentre and AI development and the energy transition. We will share more related insights in this space shortly.