Data Centres in Australia: Bring Your Own Power, Bring Your Own Answer

By Max van Someren

In March, the federal government published its Expectations of Data Centres and AI Infrastructure Developers. It is four pages long, it is not regulation, and it may be the most consequential document in Australian data centre development this year.

The teeth are in the implementation section: "Energy-intensive data centre proposals not closely aligned with the expectations will not be prioritised by Commonwealth regulatory assessments." Alongside it, the Energy and Climate Change Ministerial Council is moving the states toward a nationally consistent Bring Your Own Power framework, and the direction is plain: large new loads should bring clean energy, secure firm capacity, be flexible, and pay their way.

Read those together and something important has happened: the power question has moved desks. It used to belong to the utility — the load applied for a connection and the grid worked out how to serve it. Now the developer is expected to arrive with the answer: new and additional clean generation or storage to offset demand, a share of network costs, demand flexibility, no upward pressure on prices. With AEMO forecasting data centres growing from around 2% of grid-supplied electricity to 6% by 2030, that's not a courtesy — it's the condition of the queue moving at all. Power strategy is no longer a procurement detail. It is the foundation of your approval pathway and achieving the social licence to operate.

Three things about that shift are underappreciated.

The expectations are hourly, even though they never say so. "Secure new and additional clean energy to offset demand" sounds like an annual matching exercise — contract enough MWh of wind and solar to cover consumption and the spreadsheet balances. But the other expectations cannot be answered annually. Whether you put upward pressure on prices depends on which hours you draw from the grid. Whether you improve system stability depends on what your load, storage and firming do in the tight hours. Peak-load management is, definitionally, about specific hours. A capacity factor is an average — and the questions regulators, communities and offtakers now ask are not average questions.

The cost of getting it wrong is asymmetric and large. Overshoot on firming and you carry stranded capital measured in hundreds of millions. Undershoot and you miss the commitments that anchored your connection agreement and your community's consent, with your anchor tenants watching. And the deciding is done under time pressure, because for most projects time-to-power, not land or fibre, is the constraint that determines whether they exist.

The analysis feels complex, but it no longer needs to be slow. Screening sites and supply options against cost, reliability and hourly carbon-free energy is a well-posed optimisation problem. The barrier was never the mathematics; it was that the proper version — full-year, hourly, whole-system — lived inside consulting engagements and enterprise software licences. Weeks per option doesn't work when the shortlist has twenty.

That is why we chose this moment to launch Levelised (levelised.com - that's with an 's'). It is the modelling engine behind our advisory work at Bivios. Full-year, hourly optimisation of any energy or mass flow, built on the open-source PyPSA framework used in grid studies worldwide: now free, in the browser.

There is a solved, grid-connected data centre example ready to open: buy from the grid, or build on-site wind, solar, storage and gas, and see what the hours actually say. No account needed.

And there is an honest circularity in the timing. The demand wave that created BYOP — AI — is the same technology that let a small team build and ship this platform, and it is changing how the modelling itself is done. Internally, we already build and iterate models directly from AI chat, sitting on top of the optimisation engine — a capability we're preparing to roll out. Chat is a good way to ask. A real solver is what makes the answer trustworthy. The future of AI and the future of energy are not just the same conversation. Increasingly, they are the same toolchain.

If BYOP has moved the power question onto your desk: open the data centre example at levelised.com and change the numbers to yours. And if the decision is live and the stakes justify support, the Bivios team does this work every week — tell us the decision you're trying to make, and we'll build the model with you.