Submission to the Inquiry into AI and Data Centres

interior view of a data centre. There are rows of servers on either side of a wide corridor.

Capturing publicly created land value while supporting investment in Australian AI infrastructure

September 2026
Dr. Michael D’Rosario
Director of Research and Policy
Prosper Australia

1. Executive summary

Australia is entering a phase that will determine the success of the national sovereign AI strategy and have major ramifications for international competitiveness, as it decides how to facilitate the substantial expansion in the physical infrastructure required to support artificial intelligence. Training and inference at scale require data centres, electricity generation and transmission, telecommunications networks, water, and appropriately located land. As these requirements increase, locations that can provide them concurrently acquire an economic advantage that can be reflected in land values as well as in returns to the productive capital installed on the site.

Australia has strong reasons to support investment in domestic compute (and inference) capacity, including general productivity, research capability, technological resilience and the provision of infrastructure for Australian firms and public institutions. However, the return required to induce investment in servers, buildings, cooling systems, and other produced assets should be distinguished from increases in land and locational value arising from scarcity, public infrastructure, and planning decisions. Acknowledging this distinction, which separates the socially generated value of land from the value of improvements made by its owner, has served many economies. It also follows Prosper Australia’s longstanding work on land value capture, which has examined how public infrastructure and changes in development rights can be capitalised into private land values.1 2

The appropriate policy response given this opportunity is to employ a coordinated value capture process. Most relevant instruments sit with state and local governments through land taxation, planning, development contributions and infrastructure provision. The Commonwealth nevertheless has a material coordinating role because compute capacity, electricity transmission, telecommunications, foreign investment, and interstate competition have national consequences.

A cooperative framework should identify publicly created land-value uplift, establish baseline valuations before major infrastructure announcements, assess the adequacy of existing state mechanisms, and reduce the risk that interstate competition transfers publicly created rents without increasing national investment.

Our primary contention

Australia can support large-scale private investment in AI infrastructure while recovering a reasonable proportion of the land and locational value created by public infrastructure and public decisions. The relevant base is not the productive capital installed in data centres, but the economic rent attached to scarce, publicly enabled locations. This is logical, given that sites require significant public investment and/or draw heavily upon public resources, while benefiting from value uplift.

Principal findings

  • AI infrastructure is unusually location-sensitive because large facilities require electricity, transmission capacity, fibre, planning permission and, in some cases, substantial water access at the same location.
  • Public investment and regulatory decisions can materially increase the productive capacity of particular sites, with part of that benefit capitalised into land values rather than retained as a return to the infrastructure provider or the wider public.
  • Infrastructure cost recovery and land value capture are not equivalent. A developer may meet the attributable cost of an infrastructure augmentation while the resulting scarcity value of the site increases by a larger amount.
  • Prosper Australia’s earlier work on infrastructure and rezoning provides an established Australian basis for separating returns to productive investment from publicly created land-value uplift.3
  • Timing matters because expected infrastructure benefits can be incorporated into land prices following credible announcements, before the relevant infrastructure is completed.

The appropriate policy conditions

The Commonwealth should coordinate a national framework to identify and measure land and locational value associated with major AI infrastructure, while leaving the principal fiscal instruments with states and territories. States should assess existing land tax, betterment, windfall gains and infrastructure contribution arrangements before creating any datacentre-specific charge.

Baseline land valuations should be established before major planning changes, preferred precinct announcements, or infrastructure commitments materially alter market expectations. Public assistance for large datacentre projects should be assessed on a national basis, including its effect on underlying land values and whether it changes total Australian investment or principally its location between states.

Where public infrastructure creates material private land-value uplift that is not captured under existing arrangements, states should recover a reasonable proportion using the least distortionary available land-based instrument.

2. AI infrastructure, land and economic rent

The expansion of AI is frequently described as a digital transformation, but the infrastructure supporting it is physical. Large-scale training and inference require substantial capital investment, together with access to electricity generation, transmission and distribution networks, high-capacity telecommunications, water, and land. The scarcity of locations that can combine these inputs creates a locational component of value that is analytically distinct from the value of the data centre equipment itself. Differences exist between inference- and compute-focused centres, but the economic dynamics remain broadly consistent.

A data centre operator creates productive assets through investment in servers, buildings, cooling systems, energy equipment, and software. The value of the underlying location is determined partly by factors external to that investment. Access to high-capacity electricity connections, transmission infrastructure, fibre and appropriate development rights can make one site considerably more valuable than another. Where these advantages are scarce, and particularly where they arise from public investment or regulatory decisions, their value can be capitalised into land prices.

This distinction is consistent with the Georgist treatment of land and improvements. K’Akumu (2025) places contemporary land value capture within this tradition, focusing on the socially generated value of land and the case for separating that value from improvements produced by private investment. For AI infrastructure, the practical implication is that a fiscal instrument directed at unimproved or publicly created land value has different incentive effects than a tax on servers, buildings, or other produced assets. The objective of this mechanism is not to appropriate the return to AI investment (there may be valid reasons to do so beyond the focus of this submission), but to recover a proportion of economic rents arising from scarcity and collective investment.4

International experience also indicates that the distribution of land rents associated with rapid development is shaped by institutions rather than fixed by the development process itself. Phang’s (1996) analysis of Singapore describes a system in which extensive public land ownership and planning powers allowed the state to capture a substantial share of land rents associated with economic development. Singapore’s institutional structure is not directly transferable to Australia, but it demonstrates that land-rent capture can form part of a broader development strategy rather than operate solely as an additional revenue measure.5

3. Public infrastructure and land value capture

Prosper Australia has addressed the relationship between infrastructure finance, planning and rezoning over a long period. Our land value capture research argues that public infrastructure generates increases in surrounding land values and that part of this uplift should appropriately contribute to the public cost of creating it. Our proposition is not that all property appreciation should accrue to government, but that public expenditure can change the productive and locational advantages of land in ways that are subsequently capitalised into private asset values.6

Prosper Australia’s historical work has identified Australian precedents associated with the Sydney Harbour Bridge, the Glen Waverley railway extension and earlier infrastructure provision in New South Wales. These arrangements arose in different institutional settings and should not be replicated mechanically for contemporary data centres. Their relevance is that Australian governments have previously treated increases in land value as part of the economic return generated by public infrastructure rather than regarding infrastructure solely as a cost to be financed from general taxation.7

The same reasoning appears in our longstanding work on rezoning. Consider the distribution of the sources of benefits and burdens and the beneficiaries. The Rezoning Honeypot: Evidence from Fisherman’s Bend (2021) examined large gains that can follow changes in development rights and the incentives for speculative acquisition and rent-seeking that accompany those gains. Electricity connections, transmission augmentation and other enabling investments are not identical to rezoning, but both can materially change the feasible intensity and profitability of land use through decisions external to the productive investment ultimately undertaken on the site.8

Prosper’s 2025 submission to Infrastructure Victoria further distinguishes infrastructure cost recovery from capturing the wider land-value uplift generated by infrastructure. This distinction is particularly relevant to datacentres because the cost of augmenting a network and the economic benefit conferred by access to scarce network capacity need not coincide. A developer may make an appropriate contribution towards a transmission upgrade while the resulting access increases the underlying value of the site by a larger amount. Cost recovery may therefore be complete while a material economic rent remains privately capitalised.9

4. Applying value capture to data centres

Consider industrial land where development potential is initially constrained by electricity availability, transmission capacity or planning conditions. A subsequent combination of network augmentation, planning approval and complementary infrastructure may permit the same site to accommodate a large compute facility. Any increase in the underlying value of the land following those decisions cannot be attributed wholly to additional capital invested by the landowner. To the extent that it reflects improved infrastructure access, development rights and the scarcity of comparable sites, it represents an increase in the economic value of the location itself.

The distinction has direct implications for tax incidence, which has been largely absent in this debate thus far. A tax imposed on servers, buildings, or other produced assets raises the cost of capital formation and can affect the scale or location of investment, but where warranted may still be necessary for reasons beyond the scope of this submission (crowding out effects impacting creative industries, labour market demands for trades, equitable distribution of a societally derived dividend), but nonetheless warrant consideration.

Timing is central to implementation because market expectations can, and do, capitalise future infrastructure benefits before construction occurs. If governments identify preferred data centre precincts, announce major transmission investment or signal that particular locations will receive substantially greater electricity capacity, land prices may respond before the infrastructure is delivered. Baseline valuations should therefore precede major public announcements wherever practicable, supported by counterfactual comparisons with similar industrial land that does not receive the same infrastructure or planning advantages.

Existing state land taxation should be the starting point for any reform. Where recurrent land tax already captures a reasonable proportion of increased unimproved value, an additional charge could duplicate the existing system. Supplementary mechanisms are more defensible where a discrete government decision creates a substantial uplift that is weakly captured under recurrent taxation, or where substantial public expenditure is undertaken specifically to enable a development. Available instruments include infrastructure contributions, betterment or windfall gains mechanisms, recurrent land taxation and, where land remains publicly owned, long-term leasing.

5. Commonwealth and state responsibilities

Responsibility for most of these instruments lies with state and local governments. Land taxation, planning, rezoning, development contributions and many infrastructure decisions fall principally within state jurisdiction, so the case advanced here does not require a Commonwealth land tax or a federal charge on data centres. The development of AI infrastructure nevertheless has a national economic dimension that provides a basis for Commonwealth coordination.

Compute capacity has implications for national productivity, research, technological capability, and resilience. Large data centres also interact with electricity generation and transmission across the National Electricity Market, while telecommunications, foreign investment, industry policy and aspects of energy policy involve substantial Commonwealth responsibilities. Decisions by one jurisdiction concerning major electricity users can consequently have effects extending beyond that jurisdiction, particularly where new transmission, generation, or network capacity is required.

Interstate competition provides an additional reason for coordination. States have legitimate reasons to compete for productive investment, but concessions offered to influence the location of an investment within Australia do not necessarily increase aggregate national investment. Where a facility would have been constructed in Australia irrespective of whether it was located in Victoria, New South Wales, or Queensland, competition through tax concessions, subsidised infrastructure or other assistance can redistribute the investment geographically while reducing the public return associated with it. The problem is more pronounced where government assistance is subsequently capitalised into privately owned land.

A cooperative Commonwealth-state framework could establish common principles for identifying the sources of economic rent associated with large compute and inference facilities, measuring changes in underlying land values, assessing the contribution of public infrastructure to those changes and evaluating the extent to which existing state mechanisms recover them. Implementation could remain with states and territories, allowing instruments to reflect differences in land markets, electricity systems and existing fiscal arrangements.

6. Recommendations

For the Commonwealth:

  1. Initiate an intergovernmental AI infrastructure value capture framework with states and territories, focused on common valuation, attribution and disclosure principles rather than a new Commonwealth tax.
  2. Require national assessments of material public assistance for major datacentre developments to consider whether assistance increases total Australian investment or principally changes its location between jurisdictions, and to identify associated land-value effects.
  3. Incorporate land-value and infrastructure-benefit assessment into relevant Commonwealth AI, energy and major infrastructure processes where Commonwealth expenditure or policy materially affects datacentre locations.

For state and territory governments:

  1. Establish baseline land valuations before major planning changes, preferred precinct announcements or infrastructure commitments likely to alter the development potential of datacentre sites.
  2. Assess existing recurrent land tax, infrastructure contribution, betterment and windfall gains arrangements before adopting sector-specific charges, and use supplementary land-based mechanisms only where material publicly created uplift remains uncaptured.
  3. Publish the principal forms of infrastructure support, concessions and land-related assistance provided to major data centre developments, together with an assessment of the expected public return.

For local government:

  1. Where local rating and planning powers are relevant, maintain consistent valuation and development contribution practices so that local concessions do not inadvertently transfer publicly created location value without a corresponding public benefit.

7. Conclusion

Australia need not choose between attracting investment in AI infrastructure and recovering economic rent associated with its development. Competitive returns to servers, buildings, technology, and other produced capital that support investment should be preserved, albeit with acknowledgement of broader claims justifying an equitable distribution; while increases in the value of scarce locations resulting from public infrastructure (most notably but not exhaustively) transmission/grid access, planning decisions, and wider economic development have a different economic basis.

Prosper Australia’s earlier research on infrastructure finance and rezoning, together with the wider literature, provides a coherent basis for separating these components of value and for recovering a proportion of publicly created uplift without treating productive AI investment itself as the tax base.

The principal instruments remain state-based, but the national significance of compute, electricity infrastructure and interstate investment competition gives the Commonwealth a legitimate coordinating role. Establishing common principles, baseline valuations and transparent assessment before major infrastructure commitments become fully capitalised into land prices would allow Australia to support the build-out of AI infrastructure while retaining a reasonable share of the value created through public action.

Note – The report distinguishes produced capital from underlying land and locational value. References to value capture concern increases in unimproved or site value associated with public infrastructure, planning or scarcity, not the gross value of the data centre development. Any future quantitative application should establish pre-announcement baseline valuations and use comparable sites or other counterfactual methods to separate general market movements from uplift plausibly associated with specific public decisions.

  1. Prosper Australia, Land Value Capture: The Rational Way to Finance Infrastructure, Prosper Australia, Melbourne. ↩︎
  2. K’Akumu, O.A. (2025), ‘The principles of land value capture in the perspective of Georgist political economy’, Planning Perspectives. https://doi.org/10.1080/02665433.2024.2386685 ↩︎
  3. Prosper Australia (2021), The Rezoning Honeypot: Evidence from Fishermans Bend, Prosper Australia, Melbourne. ↩︎
  4. K’Akumu, O.A. (2025), ‘The principles of land value capture in the perspective of Georgist political economy’, Planning Perspectives. https://doi.org/10.1080/02665433.2024.2386685 ↩︎
  5. Phang, S.Y. (1996), ‘Economic development and the distribution of land rents in Singapore: a Georgist implementation’, American Journal of Economics and Sociology, 55(4). ↩︎
  6. Prosper Australia, Land Value Capture: The Rational Way to Finance Infrastructure, Prosper Australia, Melbourne. ↩︎
  7. Prosper Australia, Value Capture: A Historical Perspective, Prosper Australia, Melbourne. ↩︎
  8. Prosper Australia (2021), The Rezoning Honeypot: Evidence from Fishermans Bend, Prosper Australia, Melbourne. ↩︎
  9. Prosper Australia (2025), Infrastructure Victoria Draft 30-Year Infrastructure Strategy: Submission, Prosper Australia, Melbourne. ↩︎
Tags
Avatar photo
Michael D'Rosario

Michael D'Rosario is Prosper Australia's Director of Research and Policy.