In one paragraph
On 15 July 2026 the Prime Minister committed Australia to being “the first country in the world to bring these issues into a single, national framework” for hyperscale and AI data centre infrastructure (PM’s address, University of Sydney, 15 July 2026). The framework rests on five vectors already articulated in the March 2026 Expectations document (DISR, March 2026): national interest, energy transition, water security, skills, and research and innovation. The Prime Minister has since confirmed on 7.30 that the mandatory obligations will apply to new proposals only, with facilities already under construction expressly exempt (Sunshine Coast News, 20 July 2026). The Office of AI, established inside PM&C from 15 July 2026, is coordinating the design of the standards as a coordinating body across departments rather than a standalone regulator (Department of the Prime Minister and Cabinet). National Cabinet is scheduled to consider framework direction in August 2026, with legislation targeted for early 2027, as set out in the Prime Minister’s address of 15 July 2026 and the accompanying ArchitectureAu report, 16 July 2026. The Senate Environment and Communications References Committee is running an inquiry with submissions accepted until 1 September 2026 and reporting on 16 November 2026 (Parliament of Australia). This paper argues that the framework as announced leaves a systematic evidence gap on the existing operational fleet and the ninety-plus facilities already in the pipeline, which cause the majority of the harms the framework names, and that closing this gap requires a mandatory annual reporting obligation on large-scale existing and under-construction facilities against a published Commonwealth reference methodology, structured on the standard Australian precedent chain of the National Greenhouse and Energy Reporting Act 2007, the Modern Slavery Act 2018, and the Treasury Laws Amendment (Financial Market Infrastructure and Other Measures) Act 2024 climate disclosure regime. Two connected recommendations follow: a mandatory reporting layer on the existing fleet, and open Commonwealth procurement of the reference methodology that serves both the reporting layer for existing sites and the compliance layer for new proposals.
1. The recommendation, in three sentences
Extend the national AI infrastructure framework with a mandatory annual reporting obligation for large-scale existing and under-construction data centre facilities, against a published Commonwealth reference methodology, structured on the NGER, Modern Slavery Act, and climate disclosure precedents. Procure that reference methodology through open Commonwealth procurement rails under a multi-supplier open-panel structure, and use the same reference methodology as the compliance layer for new proposals. One reference methodology, two applications, one competitive procurement.
2. The evidence gap the framework leaves on the existing fleet
The Prime Minister’s 15 July 2026 speech named four specific harms the framework is designed to address: grid stress that pushes up power prices for households, water competition, competition with new housing for land, and community amenity impacts on where facilities are built. Every one of these harms is being generated right now by facilities the framework will not touch. There are ninety-plus data centres already in the pipeline in Australia (Australian Greens, 17 July 2026). The NEXTDC SC2 facility on the Sunshine Coast, currently under construction, has been publicly confirmed by the Prime Minister on 7.30 as exempt from the incoming rules.
A framework that mandates net-generator behaviour on new sites and requires no disclosure at all from the existing fleet is fragile on three fronts. It leaves a large loophole that community and environmental groups have already flagged, with the Australian Greens’ communications spokesperson Senator Sarah Hanson-Young stating that “with over 90 datacentres already in the pipeline, we cannot permit a free-for-all in the interim,” and the Lane Cove Responsible Planning Group calling for no new approvals until the framework is in place (The Guardian Australia, 16 July 2026). It produces an evidence vacuum that prevents the Commonwealth from knowing what harms are being caused, by which facilities, at what scale. And it exposes the framework to state-level bypass, in which a state government can underwrite a facility outside the framework’s design intent without any Commonwealth instrument to make the departure visible. The Beetaloo and Project Ares proposals set out in Section 4 illustrate that risk in live time.
Australian regulatory practice has a well-established answer to problems of this kind. Commonwealth mandatory-reporting regimes are the standard instrument for extending oversight to an existing fleet without displacing state operating regimes. Three precedents establish the pattern.
The National Greenhouse and Energy Reporting Act 2007 requires companies meeting defined thresholds to register with the Clean Energy Regulator and report emissions, energy production, and energy consumption annually, using the Emissions and Energy Reporting System (Clean Energy Regulator). The Safeguard Mechanism leverages this data for the largest emitters. NGER applies to existing operations, imposes no operating conditions beyond disclosure, and has operated without material constitutional or cooperative-federalism friction since 2007.
The Modern Slavery Act 2018 requires Australian entities and entities carrying on business in Australia with annual consolidated revenue of at least A$100 million to publish an annual modern slavery statement (Attorney-General’s Department). The Act commenced 1 January 2019. It imposes no operating conditions, only mandatory disclosure of steps taken to address modern slavery risk in operations and supply chains.
The Treasury Laws Amendment (Financial Market Infrastructure and Other Measures) Act 2024 requires certain Corporations Act entities to apply AASB S2 Climate-related Disclosures for annual reporting periods beginning on or after 1 January 2025 (Australian Accounting Standards Board). The regime applies to existing entities and imposes disclosure obligations only, covering climate-related risks, opportunities, transition plans, and Scope 1, 2, and 3 emissions.
The pattern is consistent across all three: Commonwealth-scale, single-national-framework instruments that apply to existing operations by design, impose disclosure rather than operating conditions, sit on top of state operating regimes without displacing them, and are ratified across successive governments of both major parties. Mandatory annual reporting for large-scale data centre facilities against a published reference methodology is a direct precedent match to this pattern. The March 2026 Expectations already name the five vectors around which a reporting instrument can be structured. What is missing is the reporting instrument itself.
The market for the disclosed information is already there. Insurance underwriters are increasingly pricing physical climate risk and water-catchment risk into facility premiums; disclosed site-level scoring lets them do this without proprietary alternatives. Corporate procurement of AI infrastructure, including Commonwealth agencies, financial institutions, and universities, increasingly requires ESG disclosure at facility level; disclosed reporting closes this gap without operator-by-operator negotiation. Community groups, local government, and state parliamentary submissions are asking for facility-level transparency the current voluntary regime does not provide; the City of Melbourne unanimously passed a motion in February 2026 calling on state and federal governments to establish clear monitoring standards on energy and water consumption for data centres, and Lord Mayor Nicholas Reece has publicly warned that if all currently proposed Melbourne data centres proceeded, they could consume up to 20 gigalitres of water annually, equivalent to around four per cent of Melbourne’s drinking water supply (ABC News, 4 February 2026). Mandatory reporting operationalises the framework’s own transparency logic across the fleet, at zero additional operating burden on operators beyond disclosure itself.
3. What mandatory reporting would specify
Design elements the framework should include. Each is drawn from the precedent chain above and adapted to the site-level data centre category.
Scope. Mandatory annual reporting for all data centre facilities in Australia above a defined threshold. Threshold options include nameplate power draw above a defined MW figure (illustratively 20 MW, consistent with common industry cut-offs for “hyperscale”), floor area, or contracted grid connection capacity. The framework’s own consultation process can settle the exact threshold; the design of the reporting instrument does not depend on it. Facilities under construction on the date of commencement are captured on a phased basis, with the first report due 24 months after commissioning. Facilities in the approvals pipeline are captured from a defined trigger point.
Reference methodology. Reporting is against a published Commonwealth reference methodology covering, at minimum, grid connection capacity, generation-to-draw ratio on the Prime Minister’s net-generator definition, and grid-import mix at the facility connection point; water consumption, cumulative-catchment demand, non-potable water utilisation, and any additional water infrastructure funded by the facility; facility location relative to residential and community amenity, planning-scheme zoning, and heritage overlays; skills and workforce composition indicators; and fossil-gas alignment indicators covering facility-adjacent gas infrastructure and gas dependency of the grid-import baseline. Each of these maps to one of the five vectors named in the March 2026 Expectations, and each has data sources already in the public domain including AEMO, state water corporations, state planning schemes, ABS, and BOM. The reference methodology binds these data sources to a single facility-level report.
Cadence. Annual, on a fixed reference year, whether financial year or a Commonwealth-nominated reference year. Version-stamped to a specific methodology version so year-on-year comparisons are defensible.
Publisher. The Commonwealth publishes the report by default. Confidentiality treatment is available for information that would prejudice commercial position, on the precedent of NGER’s publication-threshold rules. The default posture is disclosure.
Assurance. Third-party assurance on the disclosed values, on the precedent of AASB S2 climate disclosures. This creates a market for professional-services assurance under a published standard, a role distinct from methodology supply.
Interaction with state instruments. The mandatory reporting layer does not displace state planning, water, or environmental determinations, and does not create a Commonwealth operating condition. It sits alongside state instruments on the standard NGER precedent. State agencies may use the disclosed data as an input to their own determinations without any harmonisation legislation.
Sanctions. Failure to report attracts civil penalties on the standard Modern Slavery Act and NGER structure. Publication of a report that materially misrepresents the reported values attracts additional sanctions on the standard corporate-disclosure precedent.
Legal instrument. A stand-alone Commonwealth Act on the Modern Slavery precedent, or a schedule to a broader AI Standards Act. Either can carry the reporting obligation without any change to state operating regimes; the drafting choice is open.
4. Beetaloo and Project Ares: the state-level bypass problem in live time
The state-level bypass problem is not hypothetical. On 22 July 2026, seven days after the Prime Minister’s speech, the Northern Territory government granted Beetaloo Digital, a subsidiary of ASX-listed Beetaloo Energy, an exclusive “not-to-deal” arrangement over 185 hectares at Weddell, approximately 30 kilometres south of Darwin, for a proposed A$40 billion, two-gigawatt hyperscale AI data centre powered by fracked gas from the Beetaloo Basin (ABC News, 22 July 2026; Renew Economy, 22 July 2026; Australian Financial Review, 22 July 2026). The NT Minister for Lands, Planning and Environment described the arrangement as demonstrating the Territory is “open for business.” Greenpeace Australia Pacific has called the proposal a “disaster proposal.” Beetaloo Energy has stated it intends to overbuild generation and export surplus power to the Darwin-Katherine grid (Sydney Morning Herald, 22 July 2026).
A second live example is Energy North’s Project Ares at Murranji Station in the Barkly region, described in The Saturday Paper of 18 July 2026. Project Ares proposes an initial one-gigawatt AI data centre with expansion capacity to five gigawatts, requiring up to four billion litres of groundwater per year from the Wiso Basin aquifer, powered by a 7,000-hectare solar farm with 16 GWh of battery storage plus 1,038 MW of backup gas generating capacity. Twelve companies are reported to be in advanced discussion with the NT government on data centre proposals in the Territory.
Both proposals are relevant to the argument of this paper for three reasons.
They are proposed on a scale that is unambiguously “large” under any threshold definition the framework is likely to adopt, and they are proposed under state underwriting. If the framework applies only to new proposals and imposes no reporting obligation on the existing fleet, and if a state government underwrites facilities of this scale in a manner that pushes against the framework’s design intent, the Commonwealth has no instrument to make the departure from framework logic visible in a defensible way.
Each of the vectors named in the March 2026 Expectations has a public, defensible, first-order finding on both proposals. A two-gigawatt gas-fired hyperscale facility performs materially differently on the energy vector from a co-located renewable-plus-firming facility; four billion litres of annual groundwater extraction from the Wiso Basin aquifer is a scoring input on the water vector, not a rhetorical claim. Whether either proposal qualifies as “net-generator” under the framework depends on the technical definition of net-generator adopted, which is a scoring question rather than a rhetorical one.
Both proposals name the Darwin-Katherine grid as the beneficiary of the surplus power the facilities intend to export. The interaction between state underwriting, cross-border grid effects, and the Commonwealth’s framework commitments is precisely the interaction that a mandatory reporting layer, threshold-independent by construction, makes visible without displacing state planning, water, or environmental determinations.
The Beetaloo and Project Ares examples together are the single strongest argument for the mandatory-reporting-plus-reference-methodology combination this paper recommends. Without both layers, the framework can be substantively bypassed by state underwriting of facilities that fall outside its scope. With both layers, every large facility above the reporting threshold produces a versioned, audit-anchored composite score that Commonwealth, state, parliamentary, and community stakeholders can all reference.
5. Anticipated objections and responses
Seven objections are foreseeable to the recommendation this paper makes. Each is engaged on substance below.
On the concern that a private supplier is proposing a policy that expands its own market. The recommendation is not that any single supplier’s product be mandated. The recommendation is that the Commonwealth publish a reference methodology, whether procured or developed in-house, and require reporting against that methodology. Suppliers compete under the methodology on the same basis that assurance firms compete for audit tenders under standards they did not write, and on the same basis that carbon assurance providers compete under NGER methodologies they did not write. The recommendation reads identically when the vendor is anonymised: substitute “any qualified site-intelligence methodology supplier” for every reference to a named supplier in this paper and the argument is unchanged.
On the concern that mandatory reporting duplicates existing regimes. Hyperscalers will argue that a new reporting instrument is duplicative of NGER, AASB S2, and state planning-scheme reporting. On substance, this is incorrect. NGER covers Scope 1 and 2 emissions and energy production and consumption at company level, disaggregated to facility level for the largest emitters under the Safeguard Mechanism; it does not cover water consumption, cooling architecture, grid-import mix at the facility connection point, non-potable water utilisation, catchment cumulative-impact contribution, or fossil-gas alignment. AASB S2 covers entity-level climate-related risks, opportunities, and transition plans; it does not require facility-level performance data on the vectors the March 2026 Expectations name. State planning-scheme reporting covers zoning, buffer, and heritage compliance for a specific proposal; it does not produce a versioned facility-level composite over the framework’s five vectors. There is no existing Australian regime that covers the five Expectations at facility level for a specific data centre facility. Where inputs are already reported under NGER or AASB S2, the reference methodology pulls them from the existing disclosure rather than requiring re-reporting, which is the standard design pattern for stacked reporting regimes in Australia.
On the concern that siting, water, and planning are state functions. Data centre siting, water, and planning are state functions under the Australian constitutional distribution of powers. Commonwealth mandatory operating conditions on these matters would face constitutional and cooperative-federalism friction, and this is one of the reasons the framework as announced is structured around new proposals rather than existing facilities. The recommendation in this paper does not create operating conditions. It creates a disclosure obligation. Commonwealth mandatory-disclosure regimes on state-regulated activities are a well-established precedent: NGER on state-regulated emissions activities, the Modern Slavery Act on state-regulated corporate operations, and climate disclosure on state-regulated corporate financial performance. None have faced material constitutional or cooperative-federalism challenge. The reporting recommendation sits in exactly this pattern and is the least-friction path to fleet-wide Commonwealth visibility.
On the concern that mandatory reporting on existing facilities is retroactive regulation. Disclosure of ongoing performance is prospective, not retroactive. NGER did not require reporting on past emissions; it required annual reporting on ongoing emissions from the point of commencement forward. The Modern Slavery Act did not require statements about past supply chains; it required annual statements about ongoing supply chains from the point of commencement forward. AASB S2 climate disclosure does not require disclosure of past climate risk; it requires annual disclosure of ongoing climate risk from the point of commencement forward. A data centre reporting obligation follows the same precedent: annual disclosure of ongoing facility performance from the point of commencement forward, with phased application to facilities currently under construction. This is not retroactive regulation. It is prospective reporting under the standard Australian pattern.
On the concern that the “large” threshold is undefined. The Prime Minister's 15 July 2026 speech refers to the "next generation of large-scale data centres" without defining the class, and the framework's threshold will be a live consultation issue through the National Cabinet drafting window. A recommendation dependent on that threshold being resolved risks stalling on the same consultation. The reference methodology is threshold-independent by construction: it scores any parcel at any size. The threshold at which reporting becomes mandatory is a separable policy question. The Commonwealth can adopt a conservative first-year threshold and adjust in subsequent years without changing the methodology. This insulates the framework against the need to reopen the methodology each time the threshold is revisited.
On the concern that adding a reporting layer extends the drafting timeline. The reporting instrument is a shorter drafting exercise than the operating-condition standards. The Modern Slavery Act 2018 moved from introduction to Royal Assent in approximately six months; NGER 2007 moved on a similar timeline. The drafting complexity for a mandatory disclosure regime is materially lower than for an operating-condition standard. If the operating-condition standards for new proposals are on track for early 2027, a reporting schedule or companion instrument can be included on the same timeline without material impact on the overall drafting cadence.
On the concern that extending obligations to existing operators is inefficient in a regulatory-economics sense. The strongest steelman of the “new proposals only” carve-out is a Coasian efficiency argument: operators built existing facilities under the rules that existed at the time, and imposing new obligations on them transfers costs to parties who priced their investments against a different regulatory regime. On substance this argument overstates the case for the carve-out on three grounds. First, it treats a mandatory disclosure obligation as if it were a mandatory operating-condition obligation; the two are different both economically and legally. Disclosure imposes an administrative cost of preparing and lodging a report; it does not impose an operating cost of altering the facility. NGER, the Modern Slavery Act, and AASB S2 climate disclosure all impose disclosure obligations on existing operators without material efficiency loss, and none has generated the kind of investment-deterrence effect the Coasian argument predicts. Second, the argument proves too much: on the same logic, no Commonwealth mandatory-disclosure regime could ever be extended to an existing fleet, which is not the position Australian regulatory economics has taken across four decades of disclosure-regime enactment. Third, the Coasian analysis assumes existing operators internalised the externalities the framework is designed to address at the point they built. The framework’s own premise, articulated in the Prime Minister’s 15 July 2026 speech, is that these externalities were not internalised. Disclosure is what operationalises the externality-internalisation gap, and on any first-year efficient-taxation analysis of the externality is welfare-improving relative to the counterfactual of no disclosure.
6. The reference methodology to serve both layers
The mandatory-reporting recommendation and the compliance layer for new proposals both require the same underlying reference methodology. If the framework used two different measurement systems, one for existing facilities under reporting and one for new proposals under compliance, operators would face inconsistent measurement obligations and neither Commonwealth nor state agencies would have a single dataset to work from. A single reference methodology, versioned and audit-anchored, is the only design that produces internal consistency across both layers.
No published Commonwealth or state instrument currently produces a site-level composite score across all five vectors on a consistent basis at the granularity operators, water corporations, planning authorities, and community stakeholders each need. Individual pillars have public instruments: AEMO’s ISP for grid, DEECA’s Victorian Water Accounts and state equivalents, and state planning overlays. The composite layer that ties the five vectors to a specific parcel with a specific water catchment and a specific grid connection point does not exist inside government today.
Three structural properties of the problem determine the correct instrument for building it. The product is a data service, not physical infrastructure: its value refreshes monthly and its credibility rests on multi-buyer neutrality. Bespoke Commonwealth entities designed for 20- to 30-year capital deployment are the wrong vehicle for a SaaS data product. The score must be an input to regulatory decision-making, not the decision itself: procured from an arm’s-length data service, the score remains evidence and the regulatory decision remains a departmental function, avoiding administrative-law exposure on every scored proposal. And timeline compression is binding: building a Commonwealth-owned site-intelligence capability from scratch historically takes 24 to 48 months, which the mid-2027 commencement target does not allow.
The correct instrument is procurement of the reference methodology through open Commonwealth procurement rails, including the DTA Digital Marketplace, the DTA Cloud Services Panel, or a purpose-scoped Standing Offer arrangement, under a multi-supplier open-panel structure. Multi-supplier procurement is important on its own terms: a single-verifier structure, even under an open panel, reintroduces the regulatory-capture risk that multi-supplier procurement is designed to mitigate. Carbon accounting has a single fungible unit and can tolerate single-verifier structures; site-level composite scoring is multi-dimensional and requires triangulation across suppliers to be regulatory-decision-quality.
What the Commonwealth would procure is a site-intelligence data service with the following published characteristics.
Coverage. Every large-load-capable land parcel in Australia currently occupied by, proposed for, or previously proposed for a data centre or AI infrastructure facility. A composite score on a 1-to-10 scale across seven pillars aligned to the five Commonwealth Expectations plus two additional operational pillars: Power and Grid, Water Security, Fibre and Connectivity, Planning and Location, Physical Climate Risk, Fossil-Gas Alignment, and Social-Licence Context. Weights transparent, versioned, and adjustable per jurisdictional use case.
Alongside the seven pillars, the methodology accumulates data assets across six modules: Module A (Power / Grid), Module B (Water and Climate), Module C (Community and Social Licence), Module D (Biodiversity and TNFD), Module E (Emissions and PPA), and Module F (Waste Heat and Urban Thermal). Modules are the mechanism through which the methodology produces derived proprietary composites from public data inputs; each is designed to interleave with the pillar structure without replacing it. Module D in particular is designed to align to the Taskforce on Nature-related Financial Disclosures (TNFD) framework and to the C40 Global Urban Data Centres Pact Pillar 1 (Environmental Standards), signed on 23 June 2026, both of which are scheduled to become load-bearing for Australian institutional-investor and municipal reporting on data centre and infrastructure assets from 2028 to 2029. Keeping the modules architecturally distinct from the seven core pillars reflects that Module D in particular serves a separate mandatory-disclosure clock from the DISR Expectations and Commonwealth framework, and it allows the reference methodology to service both the operator-and-government scoring layer and the investor-facing nature-disclosure layer without conflating their evidence bases.
Methodology. Published, versioned, peer-reviewed. Independent methodology review across each pillar and across each of the six modules under discussion with competent experts across academia and industry, including a competent expert covering biodiversity, nature, and greenfield-versus-brownfield assessment for Module D. The energy pillar codifies the Prime Minister’s net-generator language as a specific technical definition: gross facility renewable generation minus gross facility grid draw, at facility boundary, over a defined 12-month rolling window, additional to grid, verified against NEM data.
Data outputs to government. Score, sub-scores, confidence flags, cumulative-impact catchment and cluster views, provenance and methodology version on every value. Machine-readable feeds under the Australian Government API Design Standard.
Community-facing layer. An anonymised, site-non-identifying public register of what the score would be if a hypothetical data centre were proposed on a given zoned parcel. This is the transparency function that Senate inquiry submissions, state-level Hansard records, and the City of Melbourne’s February 2026 motion all implicitly demand.
Explicit exclusions. The tool does not replace planning decisions, water corporation servicing determinations, or Commonwealth compliance findings under the standards once legislated. It is an evidence base, not a decision-maker.
Data acquisition posture. All data assets are contractually acquired under NDA where they are not open public data. No unauthorised collection of copyrighted content occurs. The reference methodology should not require any supplier that trains foundation models or uses artist, journalist, or unlicensed content in any part of its data pipeline, and this constraint can be embedded in the procurement specification directly.
Five governance obligations follow from procurement that are stronger than under a bespoke Commonwealth entity. Peer-reviewed methodology publication, revised annually, with independent external review across each pillar. An open-data feedback obligation, under which the supplier contributes derived, non-commercially-sensitive layers back to data.gov.au, Geoscience Australia’s National Digital Twin, and state equivalents where feasible. Right of methodology audit: ANAO may audit the methodology and data pipeline under the Auditor-General Act 1997, without any new statutory instrument. Contract-exit rights: a standard Commonwealth procurement contract is terminable and re-competable, with no lock-in on either side. And a living-methodology commitment: annual methodology refresh, with review-panel sign-off before publication.
7. What is not being requested
This paper does not ask the Commonwealth for grant funding, equity, or a bespoke structural instrument. It does not ask for endorsement of any single supplier ahead of a competitive procurement, for a sole-source arrangement, for an accreditation role, or for a nominated-verifier status. It does not ask for expedited framework treatment for any specific site.
It asks the Commonwealth to include a mandatory annual reporting obligation on large-scale existing and under-construction data centre facilities against a published reference methodology, on the NGER, Modern Slavery Act, and climate disclosure precedents, with phased thresholds and ANAO methodology audit rights. It asks that the reference methodology be procured through open Commonwealth procurement rails under a multi-supplier open-panel structure rather than built inside a department, and that the same methodology serve both the reporting layer for existing sites and the compliance layer for new proposals. And it asks the Office of AI for a 30-minute methodology briefing on the seven-pillar reference methodology and its mapping to the five Commonwealth Expectations, at the Office’s convenience, before the National Cabinet paper is drafted.
8. Closing note
The single national framework the Prime Minister committed to on 15 July 2026 is the most important AI infrastructure policy commitment Australia has made in the past five years. As announced, it applies to new proposals only and imposes no disclosure obligation on the fleet already generating the harms the framework names. Adding a mandatory annual reporting layer on the existing fleet against a published reference methodology, on the standard Australian mandatory-disclosure precedent, closes that gap without displacing state operating regimes. Procuring the reference methodology through open Commonwealth procurement rails under a multi-supplier open-panel structure ensures the reference is delivered on the mid-2027 legislative timeline and remains defensible against the regulatory-capture risk that single-verifier structures concentrate. Two recommendations, one reference methodology, one competitive procurement. SIIQE is a candidate under that competitive procurement, offered as one among any qualified suppliers under the multi-supplier open-panel structure this paper recommends.