Submission to building AI infrastructure
The Australian Government via the Department of the Prime Minister and Cabinet has invited submissions for its consultation paper on building AI infrastructure. The focus is on "the future of artificial intelligence (AI) training and large data centres in Australia". Inputs have been provided for through fields attached to preformed questions (all optional).
5 October 2026
Which topics will you or your organisation be responding to?
Managing growth in large data centres and the AI training opportunity, Requirements for data centres, Conditions for AI training.
Is a single threshold such as an energy capacity (MW) sufficient to identify the data centres that should be subject to the AI standards? If not, what additional thresholds should be considered?
No. The metrics need to be more comprehensive, especially since the current situation is only the beginning. See "The current aversion
to nuclear energy...", https://www.otoom.net/dontreadthis.htm#thcuav
Does a tiered approach sufficiently capture different types of facilities and ensure proportionate obligations relating to Australia's energy transition, water management, and communities?
A tiered approach (as described in the Government's accompanying material) would be of the top-down kind, that is a form of presumption. Using energy
coefficients as in https://www.otoom.net/dontreadthis.htm#thcuav the sequencing would suggest itself, a far more natural approach.
What can Australia learn from jurisdictions that regulate large data centres, particularly in designing obligations that are proportionate to potential community impacts while maintaining incentives for investment?
Designing the obligations in the above way would be optimal, but I do not have the relevant examples from elsewhere at my disposal.
At what point in development processes should mandatory requirements apply for large data centres?
Since the standards are still under development:
The first step in the development would be establishing the contextual radius from the top down, rather like a flowchart with its nodes that branch off. Using already established data centres as a guide, the rate of growth and/or complexity
over time can be observed and used as additional margins for those that are still under development. That way, the requirements, suitably paired with their respective contextual radii, can be mandatory from the start. Which would ensure
the greater certainty as mentioned above. And by the way, such is the nature of data centres and their functionality, they can be scaled up or down (at least in principle).
What can be learned from approaches taken in other jurisdictions to ensure very large data centres currently in development pipelines are appropriately captured?
I do not have the relevant examples from elsewhere at my disposal.
Should there be a transition period for data centre proponents to meet their obligations, for example submitting a plan within a certain timeframe after commencing operations? If so, at what point in the development cycle should this occur?
The transition periods would suggest themselves as per the above mentioned 'flowchart', in fact they can cover the very beginning of the planning process. To insert construed timeframes in the form of snapshots here and there would only create
confusion and uncertainty (and quite possibly endless litigation).
How should governments anticipate and respond to large data centre energy loads which may become operational faster than additional energy generation can be completed to offset their demand?
The question implies the possibility that energy generation is caught short in the face of the eventual data centre loads. To avoid this, there always needs to be a safety margin built into existing and/or planned energy generation. Energy
coefficients as per https://www.otoom.net/dontreadthis.htm#thcuav would take care of unwanted eventualities.
Following the passage of the proposed legislation, when should compliance start? For example, immediately, or staggered for different requirements, recognising practical constraints around planning and approvals, and to ensure regulation is
proportionate. How should certainty for, and impacts on, the Australian community be balanced with these considerations?
No matter how precise and diligent the preparations are, the actual situation on the ground would require real time adjustments as the compliance process is applied in each case. To set down a presumed template would be contrary to its intent.
Do you have any additional comments on this topic?
In general, data centres and their use are very much part of nonlinear systems, in terms of themselves and their surrounds. To apply an approach that is linear creates substantial problems.
How could obligations be phased in to account for the time needed to build data centres and new renewable projects?
Energy generation (whatever the source) should be viewed in a modular fashion. Notwithstanding the time lag related to planning and building the infrastructure, taking care of modularity (as far as practical) would align the evolving needs and
purposes of both sides of the equation - the data centre and its energy supply. The phasing in process suggests itself in terms of the above.
How should 'new' renewable capacity be defined in relation to data centre investments?
In terms of the evolving modularity - see above.
At what capacity threshold should an obligation to retire or surrender renewable energy certificates apply?
To set a predefined threshold is impractical given the nonlinearity of such systems (including data centres, energy supply, surrounding infrastructure, social requirements).
Should the information on REGO certificates be used to time-match electricity use in certain periods? If so, what should these periods be?
Ideally, the system (including the parameters on REGO certificates) should have an inbuilt flexibility to respond to momentary fluctuations. It's a matter of hardware and the code to run it.
Are there any complexities in how data centres could acquire certificates to meet a renewable electricity obligation?
Of course. The complexities exist due to their very nature. I am not sure whether the conventional methodology (reading between the lines here) can account for complex dynamic systems.
What incentives or obligations would encourage data centres to invest in on-site generation batteries or other flexible demand measures, to minimise energy usage during peak pricing?
By baking in extra margins from the very beginning as part of the certification process.
Where electricity is being purchased from a state-owned operator, what should the AER consider when assessing whether to modify a data centre's renewable electricity obligation?
In such a case the utility radius would be extended to encompass state-related parameters, keeping in mind that in all likelihood there would be more than one centre.
What unintended consequences or implementation challenges should the government consider in designing a renewable electricity framework for data centres?
The energy framework answers not only to data centres, it also relates to the rest of the economy and its outlets. Therefore, changing markets, superceding technology, political decisions (eg, tariffs).
Should the Commonwealth establish national baseline energy and water efficiency metrics for data centres? What might be the benefits and/or unintended consequences of this approach?
What is needed is a coefficient as per https://www.otoom.net/dontreadthis.htm#thcuav. Then there won't be any unintended consequences.
Should all data centres be required to meet energy and water performance thresholds? If not, how should requirements change or scale for different data centres?
Yes. You wouldn't want a data centre to take over, as it were.
How frequently should data centres be required to transparently and publicly report their energy use, efficiency, and emissions as part of plans demonstrating compliance with mandatory requirements?
In terms of the timelines applicable to any remedial action.
Should a nationally consistent hierarchy of water sources be adopted for data centres, prioritising non-potable/recycled water with potable water as last resort?
Yes, provided the hierarchy incorporates local specifics, such as climate, rainfall, surrounding river systems.
Where recycled or non-potable water supplies are not reasonably available, what alternative approaches might be promoted to achieve equivalent water security outcomes? For example, water offsets within the same catchment, investment
in recycled water infrastructure with an associated transition timeframe, desalinated water supply, and local water efficiency programs.
All of the above, with the additional option of actually having no data centre.
Should data centres be required to demonstrate that proposed water use will not adversely affect local water security, drought resilience, environmental water requirements, or existing water users as part of a plan outlining their compliance?
That would be part of the planning process.
What gaps (if any) exist in electricity and water infrastructure charging and cost allocation arrangements that could result in consumers bearing costs associated with data centre-driven network or infrastructure augmentation?
The gaps would only exist if the initial parameters did not cover the overall, particular economy relevant to that data centre.
What are the barriers to data centres investing in sufficient new renewable energy and storage capacity to mitigate impacts to wholesale electricity costs?
Too narrow definitions centred on the data centre only.
What are the barriers to data centres investing in initiatives that would mitigate impacts to other water users, such as using recycled water?
Too narrow definitions centred on the data centre only.
What requirements should be placed on data centres to operate flexibly to support the operation of the electricity grid and improve system efficiency while protecting consumers from increased energy costs? This could include restrictions
on the use of non-renewable back-up generation, and scenarios under which flexibility should be required, for example during specified system stress events and scarcity events?
The requirements can be identified by observing the operation within its sustainability envelope.
What barriers would prevent data centres from adopting demand flexibility measures to support the wider electricity system?
There should be no barriers, to the point where a data centre simply would not be feasible.
What tools could be used by governments to require data centres to adopt demand flexibility?
The certification process, featuring consistent metrics.
Aside from infrastructure and network augmentation costs attributable to their energy and water demand, are there other significant potential costs from data centre activity, and how might accountability be provided for managing these?
Depends on what is meant by 'infrastructure' (access? transport? traffic density? displacements?).
What does best practice community engagement look like for data centres, including with nearby businesses?
Engage in community participation (which includes social as well as commercial activities), support local clubs, ideally the higher energy use would enable some of it to be branched off to the locals.
What role should the Commonwealth's mandatory requirements play in relation to location and community engagement? For example, the Commonwealth could promote transparency around where new data centres are being developed or plans for
community engagement.
Not being dictatorial, yet allowing local concerns etc to be taken seriously.
How can communities and industry work together to ensure data centres deliver lasting community benefits? How can this upfront commitment be reflected in state and territory planning processes or community engagement?
Let the local community take care of such matters, in line with the subsidiarity principle practised by the European Union.
Should data centres be required to locate a minimum distance away from sensitive sites such as schools or residential homes? If so, what should the mandatory minimum distance be? Which level of government is best placed to issue and administer
location requirements?
Again, observe the subsidiarity principle!
How important is proximity to population centres for data centre operators given improvements in connectivity and advances in fibre optics?
Not so much in terms of data transfer, but human movement/accessibility.
How should competing land uses be managed and prioritised, for example where land is also suitable for housing or agriculture?
What's the use of a data centre if you don't have a roof over your head?
Should states and territories work with energy, water and telecommunications companies to proactively identify data centre zones that could support future builds away from populated areas? What efforts are already underway?
In general, a resource relationship map would make things so much easier once it has been created.
Should data centre proponents be required to explore options with government and industry to uplift community telecommunications connectivity alongside their own investment?
The telecommunications infrastructure, including its current latency, should have been established by now.
How should data centre operators be required to provide skills and training opportunities, including for under-represented cohorts in skilled workforces, during construction and into ongoing operation?
It should be left to them (they are the experts), with the incentive that training would be somewhat compensated in terms of having lessened the load on publicly funded entities.
What broader workforce development obligations should apply to data centre operators during ongoing operations?
Data centres are necessary, let's not overload them.
Do you have any additional comments on this topic?
Several types of analyses have been mentioned here (coefficients, water related metrics, etc). In themselves they do not represent ground-breaking methodologies. Therefore they would be a good (and incentivising) exercise to be performed
by students in colleges and universities as a substantial assignment. So why not invite the appropriate institutes to contribute. Apart from the informational value the additional benefits would be social awareness and cohesion (and it
would be cheaper).
What international approaches, including regulatory and non-regulatory frameworks, should inform the type of safety, transparency or security conditions Australia applies to developers of frontier AI models?
Science does not follow borders.
How should safety and security conditions for AI training be designed to keep pace with rapid technological change, while providing sufficient legal certainty for investors and operators?
AI training needs to be situated next to those who drive that change, otherwise they won't be in sync with each other.
What information should AI developers be required to share with the Australian Government and how?
A literal 'sharing' cannot necessarily be relied upon. Instead, a new version/module must be made available to the government so it can be tested.
What safeguards should apply?
The safeguards will be identified once the module etc has been tested.
What contributions would add the greatest value to Australia's future industries and research and innovation system? For example: sourcing local AI solutions, access to compute capacity, locally based research and engineering
staff, investing in Australian research presence, university and public research organisation partnerships, investing in the VET system, and commercialisation activity.
All that is desirable, but can only be made to work if any red tape attached to such initiatives is pared away as much as possible. Partnerships etc can already be evolved through the incentive programs (some) universities have.
What mechanisms should the government consider for securing these contributions? For example: negotiated in-kind contributions, a compute reservation scheme, a public-interest research payment - or other approaches.
For the government the best thing is to stay out of them as much as possible. Governments should not be doers, they should be enablers, in other words, catalysts.
If frontier AI developers or other actors reserved some compute for Australian research, innovation and public-interest purposes, what conditions or terms should apply?
The conditions should be worked out amongst themselves, notwithstanding elements related to national security.
What indicators could best demonstrate contribution to Australian research, innovation and public interest purposes? For example: locally based research and engineering staff, university and public research organisation
partnerships, research outputs and commercialisation activity, or other approaches.
If an activity remains with its sustainability envelope, that activity is contributive by definition.
How should frontier AI developers measurably support skills, training and education opportunities beyond what is required of data centres operators (as outlined earlier in this form) to create a sustainable workforce and contribute
to the required workforce adaptations?
If an activity remains with its sustainability envelope, that activity is contributive by definition. It's not a glib statement: whatever the details, if all sides are happy (intellectually and commercially) there is no problem.
What additional regulatory reforms or incentives would encourage frontier AI training in Australia?
Streamline regulatory processes, get rid of red tape, automate what is left as much as possible.
Do you have any additional comments on this topic?
Although the items here are certainly important, they do not cover the full spectrum of AI-related contingencies. Please see "The world according to the
machine" at https://www.otoom.net/worldmachine.htm for more, from a perspective of cognitive dynamics as they pertain to both, human and machine, all under the
auspices of complex dynamic systems.
© Martin Wurzinger - see Terms of Use