- Scope
- Australian AI infrastructure; Commonwealth, Tasmanian and Melbourne submissions
- Article assessment confidence
- High on the principle; individual projects need their own evidence
I support substantial public investment alongside private investment where it secures a clear public benefit. If Australia wants lasting AI capability and control, governments and their corresponding citizens should be prepared to help pay for them.
The starting point is the society we need that capability to sustain.
Why the capacity matters
Australia’s fertility rate fell from about 1.80 births per woman in 2014 to 1.48 in 2024. Meanwhile, whole-economy labour productivity was 0.2% lower over the year to June 2026. Low fertility and weak productivity growth make it harder to sustain rising living standards as the population ages. ABS births data and Productivity Commission.
Treasury’s 2026 Intergenerational Report projects that the number of Australians aged 85 and over will triple over 40 years. A smaller working-age share of the population will support more older people, increasing pressure on care services and public finances. These are projections, but they describe a serious planning challenge. Treasury report.
Neither a quick reversal of fertility trends nor a sustained productivity recovery can be taken for granted. My view is that technological enhancement is increasingly necessary if we want to provide good care, maintain living standards and sustain the income from which taxes are paid. AI deserves serious investment as part of that response.
Early promise is a reason to prepare
I do not regard today’s AI as a solution to these pressures at national scale. But early research shows promising gains in particular applications.
A 2025 observational study of 218 US nursing homes associated AI-enhanced clinical records with improvements in several care-quality measures. The study was industry-funded and cannot establish that AI alone caused those improvements, or that Australian facilities would achieve the same results. These are early findings about a particular clinical support system. Nursing-home study.
Separately, a 2025 study of 5,172 customer-support agents estimated a 15% average increase in issues resolved per hour with AI assistance. That result concerned a specific workplace deployment, rather than an economy-wide productivity gain. Workplace study.
If useful AI raises productive incomes, it could also strengthen the tax base that funds care and other services. Treasury identifies that possibility while noting uncertain timing and offsetting costs. Who receives the income, how it is taxed and what the transition costs all matter. Higher output does not automatically become public revenue. Treasury’s fiscal assessment, page 21.
AI requires computing infrastructure to develop and operate. The case for building that capacity is therefore important, but each proposed Australian facility must still explain what useful capability it adds.
Be clear about what the public receives
Construction and operating jobs matter. So do useful services, practical skills and credible Australian control. Each proposal should specify which benefits it will deliver, to whom and on what terms.
Where control is part of the public case, the assessment should examine whether an Australian operator could keep an agreed model running if an overseas partner withdrew access. A local building alone cannot establish that ability, or assure access to successive generations of capable models.
Investment should also help businesses put the capability to work. Practical support means developing workflows, judging outputs and understanding responsibilities. Infrastructure and the ability to use it need to develop together.
Make the costs and risks visible
I accept that some supporting electricity and water costs may be shared through taxes or household and business utility bills. Those contributions should be explicit and justified, with a fair developer contribution.
The assessment should compare alternatives, including what would happen without public support, and identify who carries the risk if demand or promised benefits fall short. Communities should be able to distinguish forecasts, firm commitments and achieved results.
Public investment should establish infrastructure and practical skills. I do not support general subsidies for AI consumption. Users should pay for the services they consume, including a possible premium for Australian control and continuity.
Make responsible development achievable
I strongly support additional renewable generation to match data-centre demand, coordinated with network connections and credible arrangements where timing differs.
Water requirements should reflect actual demand, cooling design and local supply. Greater Western Water’s June 2026 announcement of planned recycled-water supply to a future AWS facility illustrates one possible arrangement, rather than savings already delivered or a solution for every site. Greater Western Water.
Assessment should identify genuine constraints and establish what would resolve them. Staging, shared infrastructure or a different location may provide a workable path. Unnecessary delay also carries an economic cost.
The demographic and productivity pressures strengthen the case for acting. Whether AI helps us meet them depends on the capability we build, how we use it and how carefully we distribute its costs and gains.
Why this edition changed
Rewrote the argument to connect AI infrastructure with ageing, low fertility and weak productivity. Adds official demographic and fiscal evidence and qualified early AI results, while retaining project-specific public-benefit tests and transparent contributions from governments and citizens.
Before: Substantial public investment can be justified when it secures a clear public benefit. Contributions through taxes or utility bills should be explicit and justified, with a fair developer contribution and practical capability for businesses.
Now: Substantial public investment can be justified when it secures a clear public benefit. Contributions through taxes or utility bills should be explicit and justified, with a fair developer contribution and practical capability for businesses.
Confidence: High on the principle; individual projects need their own evidence
Practical effect: Readers can distinguish the strategic case for developing AI capacity from evidence that particular facilities or funding arrangements will deliver the promised public benefit.
Edition history
v2 · 19/10/2026 · Edition source
Sources and limits
Births, Australia, 2024
2024 period fertility 1.481 births per woman, compared with 1.795 in 2014. Not completed family size or a deterministic population forecast. Registered births increased in 2024, so falling fertility does not mean fewer birth registrations that year. Migration also affects population outcomes.
Australia’s productivity performance
Dashboard updated September 2026 reports whole-economy labour productivity 0.2% lower from June 2025 to June 2026. A dated change, not evidence of an uninterrupted decline. Labour productivity, multifactor productivity and sector measures are distinct. Dynamic page checked 19/10/2026.
2026 Intergenerational Report
Official full report and factsheet reviewed. Printed pages 21, 83 and 127 cover conditional AI fiscal channels, ageing and pressure on the tax base. The 85-plus population is projected to triple over 40 years; working-age population continues to grow modestly. These are projections, not observed AI productivity or tax gains. Fiscal effects depend on income distribution, prices, transitions and other offsets.
Impact of an Artificial Intelligence and Machine Learning Enhanced Electronic Health Record System on Quality Measures in Nursing Homes: A Difference-in-Differences Analysis
Publisher abstract reviewed, not full text. 218 US nursing homes, 94 with AI-enhanced electronic records and 124 with electronic records alone. Industry-funded by Resmed. Adjusted baseline differences; 11 of 18 quality measures met the parallel-trends assumption. Associations do not establish that AI alone caused improvements, Australian transferability, staffing substitution, generative-model efficacy or national productivity effects.
Generative AI at Work
5,172 customer-support agents serving one company; estimated 15% mean increase in issues resolved per hour, with heterogeneous effects. Purpose-built historical conversational assistance, not a current-model benchmark or a universal business return. Non-randomised rollout; no aggregate employment, wage or Australian SME effect established.
Melton plant leads the way with first recycled water connection for a Victorian data centre
Recycled cooling-water supply to a planned AWS facility from the start of operations. Demonstrates a proposed arrangement, not achieved drinking-water savings or a solution suitable for every site.