Thank you for the invitation to speak at what is fast becoming an essential topic - not just for the data centre sector, but for anyone who cares about Australia's economic future.
The transformation of Australia's digital and energy infrastructure is not a forecast. It is happening right now. The decisions - or the failures to decide - we make in the next three to five years will undoubtedly shape what this country looks like for the next thirty.
We are standing at the intersection of a massive economic opportunity and hard infrastructure constraints that demand real answers.
My call today is about achieving collaboration, not competition. To get there - to build the infrastructure we need while meeting our energy, water and land use demands - we need to move beyond the familiar talking points on all sides and have some genuinely honest conversations.
We need data centre and AI deployment to give Australia's productivity a much-needed boost. We cannot afford to dilly-dally - our economic needs are pressing and being at the forefront of data centre development and deployment can, if we play our cards right, give us a massive leg up in our drive for productivity growth.
Put simply, these facilities can become the fuel stations for our future productivity engines.
In our recent collaborative report with the National Artificial Intelligence Centre, we examined how Australian businesses are integrating AI into their daily operations. The findings are compelling: when implemented well, AI amplifies human capability, supports worker safety, and delivers results. Not in spite of people - but for them.
Our economy's digitalisation trajectory demands this. As a modern, advanced economy, we cannot afford to be passive consumers of technology developed elsewhere. We need to be at the cutting edge - capable of adopting, adapting, and innovating. That requires physical infrastructure. And that means data centres.
So let me be unambiguous: Australian Industry Group supports the growth of data centres. We want them built. We want them to succeed. But we want them built well - and we want the broader economy carried along with them.
Late last year, we wrote to Energy Ministers expressing both our support for data centres and our concern that energy supply growth might fail to keep up with demand. We proposed urgent work on three fronts to lower that risk. Today, we are encouraged to see genuine progress on all three.
Our first call was for better information from the energy market authorities. With so many data centre announcements - many of them speculative or duplicative - it is easy to imagine astronomical growth rates. It is also hard to know which figures to trust. Better information leads to calmer, better-focused discussion.
Energy Ministers made a similar request of officials and AEMO. One result is the data AEMO now publishes in its Quarterly Energy Dynamics report, showing the pipeline of solid projects formally seeking connection to the National Electricity Market. In the first quarter of 2026, that figure stood at 4.1 gigawatts of data centres in the connection queue - not yet guaranteed to proceed - with a further 1.3 gigawatts in project implementation but not yet commissioned.
That is quite substantial. For context, average demand from existing data centres across NSW and Victoria was just 0.6 gigawatts in Q1 this year. The coming wave is large and will require careful management. But it is less alarming than what we might imagine without the data.
Our second ask was to require new and expanded data centres to bring new electricity supply into the market to meet their expected energy needs. A similar expectation has since been expressed in national guidelines, and Energy Ministers have requested policy options to turn that expectation into concrete requirements.
That is good progress. But I want to be clear: the intent is not to throw sand in the gears of growth. Expectations must be reasonable, flexible where appropriate, and compatible with rapid expansion.
There is also a practical mechanism emerging that can help. I am talking about the Electricity Services Entry Mechanism - the E-SEM - which Energy Ministers are moving to legislate later this year, with broad support from energy users and suppliers. In short, E-SEM will contract with new electricity resources on behalf of the whole market, underwriting their later years of operation where projects already have commercial offtake for their first few years.
It bridges the gap between the 15-to-20-year contracts electricity investors need and the 5-year horizons that major buyers are willing to sign up for. And it offers an elegant way to moderate the ask on new data centres. Yes, they would need to contract with new resources for their expected energy needs - but only for their first few years. E-SEM covers the rest. That seems like a fair deal.
Our third ask concerned peak demand. The cost of our grid is largely determined by the peaks we build it to meet. The higher those peaks, the more the system costs, and the higher the price per unit of energy. But if we grow average demand faster than peak demand, we spread costs across more units - and prices can fall. Smart, flexible demand management is a potentially significant source of downward cost pressure.
Again, Energy Ministers are asking for options to require that flexibility from data centres. Again, that movement is welcome.
And here is what is genuinely encouraging: the evidence is growing that AI-oriented data centres can meet this expectation without compromising their core business. In the United Kingdom, a trial in late 2025 demonstrated that an AI training data centre could flex its demand rapidly enough to support the National Grid on demand. Halftime in a big football match - when the grid groans beneath the sudden surge of a million teakettles switching on - is far easier to manage when another large load turns down at the same moment. This trial showed that with the right technology, AI data centres can respond rapidly, flexibly, and without forewarning.
The technology involved is not simply a giant dimmer switch. It is a sophisticated system that monitors and prioritises every job the data centre is running, redistributing workloads across time and between facilities to meet all requirements simultaneously. The trial met 100% of the grid operator's requests while delivering over 99% of high-priority jobs on time.
This flexible capability is now being built into major new data centres in the United States, and NVIDIA is including it in the reference design for facilities using its chips. There is good reason to be optimistic that, with the right policy framework, new data centres will strengthen the grid - not strain it.
But as these facilities become more capable and more complex, a deeper issue is starting to emerge that deserves attention. Data centres are becoming some of the largest and most complex loads in the system. The way power is delivered inside these facilities is shifting toward direct current power distribution - it is more efficient and reduces losses.
But it also brings new challenges, especially in standards and integration and those need to keep pace. Without clear and nationally consistent standards that underpin safety, reliability and interoperability, we risk creating uncertainty and making it harder to integrate these critical facilities into our energy system.
Water impacts of data centre growth are also attracting attention, and no wonder given Australia's history with water scarcity. This is a topic that deserves careful attention and its own thoughtful policy response.
Based on benchmarks from MIT and the US Department of Energy, 10 gigawatts of mooted data centres would increase National Electricity Market demand by 20% to 40%, while only increasing water demand by 1.4% to 2.8% in NSW and 2.3% to 4.6% in Victoria.
That rough and ready estimate could be too pessimistic for several reasons.
The rate of data centre growth may be much lower. 5.4 gigawatts of data centre capacity is currently in AEMO's connection process.
Data centre energy efficiency has major room to improve, and better energy efficiency means less waste heat per unit of services delivered - which means less demand for cooling - which means less demand for water.
That said, complacency would be a mistake. Water resources are local, not just a statewide abstraction. Intense local uses need to be managed well even where they look straightforward on a market-wide basis.
Recycled water can be provided in the right place with investment and notice. Victoria has led the way, enabling recycled water for data centre cooling. That is practical, forward-looking policy, and we encourage other states to bring forward their solutions.
Let me turn to what I consistently hear described as the biggest practical bottleneck facing data centre developers in Australia: planning and approvals.
This is not a data-centre-specific problem. Ask any housing developer, any renewable energy project, any major infrastructure proponent, and they will tell you the same thing. Australia's approvals systems are slow. They involve multiple levels of government, multiple statutory processes, overlapping requirements, and enormous uncertainty about timeframes.
The solution is not to lower standards. I am not here to argue that environmental assessments should be less rigorous, or that community consultation should be bypassed. These requirements exist for good reasons.
