Tag: New South Wales

  • The Office of AI: expectations are about to become obligations

    The Prime Minister has announced an Office of AI inside his own department and a single national AI framework, with the March data centre expectations to be legislated. It is the strongest confirmation yet of the argument we have been making all year: the era of voluntary good behaviour is ending and the question now is what happens at the assessment table.

    Speaking in Sydney this week, Anthony Albanese announced the immediate establishment of an Office of AI within the Department of the Prime Minister and Cabinet, coordinating what he described as a world first: a single national framework covering AI’s impacts on energy, copyright, productivity, education and labour rights. For data centres the substance is concrete. The national expectations released in March are to become law, expected in early 2027, including legal obligations for large facilities to underwrite their own new power supply, pay their full share of grid connection costs and meet energy and water efficiency standards. On copyright the government was blunt: there will be no exemption allowing AI firms to mine Australian creative work with impunity, even as one major developer has reportedly tied a A$21.6 billion investment to clarity on exactly that question.

    What it gets right

    Three things stand out. First, expectations are becoming obligations. When the March framework was released we noted that its five benchmarks pointed in the right direction but were not binding and that the gap between aspiration and obligation is precisely where communities lose out. Legislating the energy and water requirements closes that gap for the issues communities feel most. Second, the government has embraced the argument that guardrails attract investment rather than repel it: the Prime Minister’s case that clear rules deliver “greater clarity and speed for approvals” is the same case South Australia made with “new energy for new demand” and it is the right one. Third, someone now owns coordination. An Office of AI at the centre of government is an answer to the problem our governance frameworks name on their first page: the gap in most jurisdictions is not law but coordination.

    The vindication of the state that moved first

    Underwrite your own new power supply. Pay your full connection costs. Be as efficient as the technology allows. That is South Australia’s strategy, nationalised. It is also, in different dress, NSW’s developer-funded infrastructure principle and Victoria’s sustainable integration goal. The convergence we described in our three-state analysis has now reached Canberra; states that were waiting to see which way the Commonwealth would jump have their answer. For Queensland, Western Australia and the territories, all yet to publish a data centre position, the calculation just changed: a national statute is coming and jurisdictions with their own assessment machinery will shape how it lands; jurisdictions without will have it land on them.

    What legislation cannot do

    A statute passed in Canberra still has to work at a council assessment table in Western Sydney, a state referral desk in Brisbane and a pre-lodgement meeting in Hobart. Who verifies that a facility’s contracted supply is genuinely additional? Which metrics establish “as efficient as possible” for a liquid-cooled AI factory in a subtropical climate against an air-cooled one in Tasmania? How do national obligations mesh with three different state models and the planning Acts of eight jurisdictions? Former minister Ed Husic’s critique this week, that social licence without enforcement is doomed to failure, sharpens the same point from the other direction: rules only bite where somebody independent measures compliance and publishes the result. The legislation will set the standard; the assessment layer has to deliver it. That layer is measurement, verification and consistent practice; it does not yet exist.

    The bottom line

    This is the most significant week for digital infrastructure policy since the March expectations and it moves the debate exactly where it needed to go. It also raises the stakes on the unglamorous work: between now and early 2027, the detail of how these obligations are defined, measured and enforced will be written; the Senate inquiry taking submissions until September is one place that detail will be shaped. Independent, jurisdiction-neutral machinery for assessing compliance is no longer a nice-to-have; it is what the new law will need to function. That is precisely what the Digital Infrastructure Institute builds and we will be engaging with the Office of AI’s work from day one.

    A topical commentary from the Digital Infrastructure Institute. For the underlying arguments, see our State & Territory Frameworks and our series on planning Australia’s digital infrastructure.

  • Victoria’s action plan: three states, three models, one question

    Victoria wants to be Australia’s AI capital and its Sustainable Data Centre Action Plan is the most coordinated state response yet. With South Australia legislating and NSW consulting, the country now has three distinct models for managing the data centre boom. They are converging on the same unanswered question: what happens at the assessment table?

    Victoria’s Sustainable Data Centre Action Plan (February 2026) is a genuinely different animal from what the other states have done. It is a whole-of-government framework led by the industry portfolio with energy, planning and government services alongside, built through engagement with operators, utilities, transmission companies and planning agencies. It covers five areas: giving investors clear information on land, energy, water and transport; integrating facilities into the energy system sustainably; managing water demand system-wide; building the workforce through TAFE and tertiary partnerships; and coordinating planning. The government puts the potential pipeline at more than A$25 billion and the state has demonstrated the fastest major data centre approval in the country. The ambition is explicit: Australia’s AI capital.

    Three models in the space of five months

    Australia’s three biggest data centre states have now shown their hands and no two answers match. South Australia chose legislation: a strategy built on “new energy for new demand”, with a dedicated Act to give its requirements statutory force. New South Wales chose consultation: five principles through Infrastructure NSW, now being converted into policy while a parliamentary inquiry runs alongside. Victoria chose coordination: a non-statutory action plan that gets agencies, utilities and investors working from the same page without new law. Each model fits its state. SA is protecting the world’s most advanced renewable grid and can afford to set firm terms. NSW is managing the largest and messiest pipeline and needed to hear from everyone first. Victoria is competing for investment and optimised for speed and certainty.

    What Victoria gets right

    Coordination is the thing most jurisdictions never achieve and Victoria has built it deliberately: the energy system, water system, land supply and workforce pipeline treated as one problem rather than four portfolios. The plan’s instinct on water, a system-wide approach to demand rather than project-by-project improvisation, is exactly right for a state whose supply is drought-sensitive. The workforce partnerships acknowledge what our research keeps finding: the constraint on this industry is increasingly people, not power. Speed, done properly, is also a virtue; a 75-day approval with a strong evidence base beats a two-year approval with a weak one.

    What’s still missing, everywhere

    Here is the convergence: an Act, a consultation and an action plan all still need the same missing layer. When a hyperscale application lands in front of a council planner or a state assessor, what standard applies? Which metrics? What counts as genuine additionality, real water stewardship, enforceable community benefit? None of the three models yet gives its assessors a published, consistent, project-level standard. Victoria’s version of the risk is sharpest because its approvals are fastest: speed without a published standard will eventually be read as light scrutiny, fairly or not. A 9 gigawatt pipeline that industry expects to deliver less than a gigawatt by 2030 also means Victorian assessors will spend much of their time on projects that never get built, which makes consistent, efficient assessment machinery more valuable, not less.

    The bottom line

    Three states, three models and the differences are instructive; the gap is identical. Principles, plans and even statutes change outcomes only when they reach the assessment table as practical tools. That layer, consistent assessment standards, honest demand numbers and enforceable benefit, is jurisdiction-neutral by nature and it is exactly what the Digital Infrastructure Institute builds. We will be watching how each model performs and comparing notes across all three, because the state that closes the gap first will set the national norm.

    A topical commentary from the Digital Infrastructure Institute. For the underlying arguments, see our series on planning Australia’s digital infrastructure.

  • New South Wales and the scale test: principles are the easy part

    NSW hosts the largest data centre pipeline in the country and has just finished consulting on how to manage it. The five principles in the government’s consultation paper are sound. The hard question is the one South Australia is also facing, at a tenth of the scale: how do principles become consistent practice at the assessment table?

    New South Wales is where Australia’s data centre build-out actually lives. At March 2026 the state’s development pipeline held 44 projects totalling 11.4 gigawatts and around 60 per cent of the 5.4 gigawatts in AEMO’s transmission connection queue. The Investment Delivery Authority has endorsed 15 Sydney-region projects worth A$51.9 billion for fast-tracked approval support. In March the government released a Data Centre Consultation Paper through Infrastructure NSW, setting out five principles for managing growth; consultation closed in May and a parliamentary inquiry is underway. The direction is being set now.

    What the paper gets right

    The principles have the right instincts. Developers should fund their own infrastructure requirements, over and above what is already planned, so households do not pay for the sector’s growth: that is additionality in all but name and it is the single most important discipline in this debate. Approvals should be evidence-based. The paper is also honest about a problem others gloss over: Sydney Water’s demand forecasts and industry-commissioned modelling diverge materially; the paper calls for shared data and standardised demand assumptions. You cannot plan a water or electricity network around numbers nobody agrees on.

    The scale problem nobody else has

    NSW’s challenge is not direction; it is volume. A connection queue several times larger than any plausible build-out makes planning hard for everyone. AEMO estimates that six of every seven megawatts in connection requests nationally may never materialise; NSW holds most of them. Phantom projects clog the queue, distort forecasts and can crowd out the real ones. At the same time the fast track raises the opposite risk: 15 endorsed projects moving through prioritised approvals will test whether speed and scrutiny can travel together. Prioritisation must never mean lighter conditions, or fast-tracking becomes under-conditioning at exactly the moment volumes peak.

    The question every consent authority will ask

    Principles change outcomes only when they reach the assessment table. Will a planner in Western Sydney, weighing a hyperscale application against housing and employment land, have the same tools and the same standard as an assessor in a regional REZ council? Will “evidence-based approvals” mean a published assessment standard, or a phrase in a paper? South Australia is confronting the same implementation questions with its dedicated Act; NSW faces them at ten times the scale and without bespoke legislation. That makes consistent, practical assessment machinery more important in NSW, not less.

    The bottom line

    The consultation was the easy part and NSW did it well. What follows determines whether the principles bite: a published assessment standard every consent authority applies the same way, honest shared demand numbers and enforceable local benefit through the planning agreement machinery the state already has. The states are converging on the same questions from different directions. Independent, cross-jurisdictional analysis of what actually works at the assessment table is becoming more valuable by the month; that is precisely where the Digital Infrastructure Institute focuses.

    A topical commentary from the Digital Infrastructure Institute. For the underlying arguments, see our series on planning Australia’s digital infrastructure.