Australia's AI Infrastructure Consultation: Proposed Standards For Data Centres And Frontier AI

The Australian Government has released a consultation paper proposing national mandatory minimum standards for data centres (Consultation Paper), as part of its broader AI standards initiative. The Consultation Paper establishes a tiered framework governing energy, water, community and workforce obligations, and separately seeks feedback on conditions for frontier AI training in Australia.

Several large data centre projects currently progressing through approvals may be captured by the standards before becoming operational. 

Submissions close at 5pm AEDT on 9 October 2026.

Background

On 15 July 2026, the Prime Minister announced that the Government will introduce AI laws in Australia's interests.

The AI standards will build on the national expectations of data centres and AI infrastructure developers released in March 2026 (explore our article here), and align with National Cabinet's decision delivered in August 2026. The Commonwealth will legislate the national AI standards for large data centres, while preserving state and territory responsibility, and the standards will complement existing approvals processes.

This national push reflects a broader shift towards greater regulation of large data centre sustainability – a trend already moving at the state level. In September 2026, the Victorian Government proposed to introduce measures relating to renewable energy sourcing, water usage and the development of new data centres in the State. These measures are broadly aligned with the proposed national standards.

Key takeaways:

• The Australian Government is proposing a tiered approach that excludes small facilities below 30MW from the national AI standards, introduces baseline requirements for facilities with an ultimate electricity network connection capacity between 30–100MW, and establishes more substantial requirements for large data centres above 100MW, or a cumulative portfolio equivalent of 150MW.
• The Commonwealth Government proposes to legislate the national AI standards for large data centres, while preserving state and territory responsibility for local planning and approvals.
• Large data centres could face significant new obligations relating to energy consumption, water use, community engagement and workforce development.
• Frontier labs authorised to undertake large-scale AI training in Australia could be subject to minimum security and safety expectations, such as disclosing reportable AI incidents to relevant Australian authorities.
• Data centre operators, developers, investors and AI developers should carefully review the proposals and consider making a submission before the 9 October 2026 deadline.

The tiered framework

The proposed framework establishes three tiers based on a facility's ultimate electricity network connection capacity.

Pipeline projects

The Government is considering how the standards should apply to projects not yet operational at commencement of the legislation. Options include requiring facilities above a threshold to submit compliance plans at a reasonable point in their development pathway, legislating obligations to confirm necessary approvals are in place, and applying differentiated requirements depending on an entity's ability to influence the delivery and operation of a facility. These questions have significant implications for developers, investors and financiers with projects already in the pipeline.

What are the proposed requirements for large data centres?

The obligations proposed for large data centres span four areas. 

1. On energy, facilities would need to bring forward new renewable generation to fully offset their energy demand and retire renewable electricity certificates equivalent to their annual consumption. They would also need to manage their consumption during peak periods and ensure they do not increase transmission, network or other utility costs for other users. The Consultation Paper contemplates renewable electricity certificates being sourced through power purchase agreements, the spot market or electricity retailers.
2. On water, facilities would need to prioritise efficient cooling technologies, use non-potable or recycled water, and demonstrate operational resilience to drought and climate change. 
3. On community engagement, facilities would need to demonstrate meaningful engagement with affected communities, First Nations peoples and councils, show that they have drawn on local suppliers and are appropriately sited away from sensitive areas. 
4. On skills and training, facilities would need to demonstrate active support for workforce training. There is a real risk that large data centre projects draw labour away from other critical sectors. 

Conditions for AI training

Frontier labs authorised to undertake large-scale AI training in Australia could be required to meet minimum safety and security expectations, including disclosing reportable AI incidents to Australian authorities. 

The Government is consulting on how conditions can keep pace with technological change while providing the industry with sufficient certainty. It is also seeking feedback on the information AI developers should be required to provide to the Government. In addition, the Government has highlighted concerns that Australia may host AI infrastructure and training activities while the associated intellectual property and economic benefits are accrued offshore. Copyright has been identified as a key component of the proposed AI standards, with the Attorney-General to undertake further stakeholder consultation on how best to balance incentives for AI development and training with strong copyright protections.

What next?

We will monitor the progress of this consultation. 

In the meantime, entities should consider whether there are any practical issues they wish to raise on the Consultation Paper by the due date of 9 October 2026.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.