Data Centres Australia chief executive Belinda Dennett and Anthropic's ANZ head of policy, David Masters, both welcomed the federal government's new consultation paper on national standards for AI and data centers, released today, Friday, September 18, 2026. Their endorsements signal that the proposed framework aligns with the interests of major tech capital, even as the government claims to set terms for investment. The paper outlines a strategy to attract significant investment from top AI companies, aiming to legislate national AI standards in early 2027.
The federal government's stated objective is to grow AI capability within Australia. Assistant science minister Andrew Charlton declared, "The Australian Government has been clear about the important objective of growing AI capability in Australia. We welcome investment, but on Australia's terms." These "terms" include minimum safety and security requirements for companies undertaking large-scale AI training. However, the paper does not define what constitutes a "reportable incident," leaving crucial details open to interpretation. It asks for feedback on whether developers should provide information proactively, on request, or through public disclosures.
Who Profits from the AI Boom
The consultation paper explicitly aims to attract "major investment from top AI companies in a sustainable way." This commitment to capital accumulation is central. Companies like Anthropic, which recently signed a deal with the Queensland government for a proposed $32 billion data center, are already positioning themselves within this state-backed expansion. The framework also suggests that data centers already under development could be retrospectively required to comply with new rules, ensuring a consistent regulatory environment for future corporate operations. Helen Wilson, Department of Industry, Science and Resources deputy secretary, asserted at an AI parliamentary hearing that hosting AI training would grant Australia greater influence over the technology, stating, "By having training here, that does give us the ability to help shape the development and deployment of AI." This framing positions the state as a facilitator of corporate power, rather than a check on it.
Managing the Commons for Capital
The proposed standards address the massive resource demands of these data centers, particularly energy and water. The government wants large data centers to bring forward enough new renewable energy generation to offset their electricity demand. This "renewable" energy, however, can be "backed by gas, batteries or hydro to keep supply reliable," indicating a continued reliance on fossil fuels and large-scale infrastructure projects. The paper also floats the idea of requiring electricity use to be matched with renewable generation annually and during some periods, potentially through renewable energy certificates – a market-based mechanism that allows for continued consumption without fundamental shifts in energy production. Data centers could also be required to adjust their electricity demand to support the grid and operate in ways that minimize costs for households and businesses, shifting the burden of energy management onto the public infrastructure while ensuring corporate operations remain viable.
For water, the proposed standards mandate minimizing consumption, prioritizing efficient cooling, and using recycled or non-drinking water. Data center providers would also have to prepare for drought and water disruptions, meet efficiency standards, and report water use. They would also be required to "pay their share of infrastructure and servicing costs." These measures, while appearing to address environmental concerns, primarily function to manage the externalities of capital's resource extraction, ensuring the long-term viability of profitable data center operations without challenging the underlying consumption model. Companies training advanced AI in Australia would also be expected to support Australian research, either by direct payment or by reserving computing capacity, a minor concession within a vast profit-generating enterprise.
Responses to the consultation paper are due in less than 1 month, by October 9, 2026. OpenAI, Google, and Meta did not immediately respond to a request for comment, perhaps finding the proposed framework already sufficiently accommodating to their interests.