AI Inquiry Atlas

An unofficial, independent guide to the public records of the Joint Select Committee on Artificial Intelligence. Not connected with the Parliament of Australia or the committee. Its makers also made a submission (no. 481; see About). Summaries and scores are written by AI and can be wrong: check the original before relying on them.

Submission 275

National Computational Infrastructure and the Pawsey Supercomputing Research Centre

Submitter group: Research institutes.

An AI model identified the submitter group and wrote the summary, key points and recommendation summaries. Counts do not measure importance or merit.

Open in the interactive guide

Summary

Read the original submission for the full text.

Pawsey and NCI argue public supercomputing is core national AI infrastructure alongside private investment. They report demand around three times capacity and warn current funding means a smaller Gadi replacement and no Setonix replacement. They seek urgent funding and a ten-year program estimated at $1.8–$2.1 billion, subject to review, plus nationwide access, specialist skills and independent safety-evaluation capacity.

Key points

Quotes are the submitter’s own words. Text taken from documents, including scanned pages read by text-recognition software, can differ from the PDF. Check the original. Listed in the same order as on the submission’s page in the interactive guide.

  1. Argues that public compute is a core national capability, not simply a complement to private investment.

    Public compute is not simply a complement to private investment, it is a core national capability. At the heart of Australia’s public computing capability are Australia’s two large scale research supercomputers: Gadi at the National Computational Infrastructure (NCI) in Canberra and Setonix at the Pawsey Supercomputing Research Centre (Pawsey) in Perth.

  2. Says private investment is welcome and essential but is only one piece of the puzzle, and the market will not provide all the computing power Australia needs.

    This investment is welcome and essential to a flourishing AI ecosystem. However private investment is only one piece of the puzzle, and the market will not provide all of the computing power Australia needs.

  3. Argues that classified and sensitive government AI workloads, and safety assessments of private infrastructure, require a trusted sovereign compute option.

    Other AI operations are not suitable for private compute at all, and must only be deployed on public infrastructure, such as national security research, Government or health services, classified and sensitive AI workloads. A trusted, sovereign compute option is essential for classified and sensitive government AI workloads, or for safety and security assessments of private infrastructure.

  4. States that demand for national public supercomputing is already about three times available capacity.

    Requests through its national allocation process were close to three times the computing capacity of available. Pawsey is also experiencing record demand. In other words, Australia’s demand for public compute is already approximately three times available capacity.

  5. States that on current funding the Gadi replacement would be lower-capacity and Setonix would have no replacement.

    Increased prices for computing components caused by the AI boom mean that on current funding levels, the replacements for Gadi and Sedonix will be smaller with less capacity. In short, existing funding levels will result in a lower-capacity replacement for Gadi and no replacement for Setonix at all.

  6. Argues that having a data centre in Australia does not by itself create sovereign capability.

    Having a data centre located in Australia is economically valuable, but location alone does not create sovereign capability. Australia also needs the ability to decide which work receives computing capacity, reserve capacity for nationally significant compute demands, and continue important work if commercial priorities, prices or access conditions change.

Recommendations

Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.

  1. Recognise public computing as a national AI and productivity priority and fund NCI and Pawsey long term to meet capacity constraints and growing AI demand.

    Recommendation: Recognise public computing infrastructure as a national AI and productivity priority, and establish long-term funding for NCI and Pawsey sufficient to address existing capacity constraints and growing demand from AI.

  2. Provide urgent top-up funding so Gadi and Setonix can be replaced without reducing national computing capability.

    The Committee recommends urgent immediate top-up funding to ensure that Gadi and Setonix can be replaced in a way that does not reduce Australia’s existing national computing capability.

  3. Establish a National AI Plan ten-year public compute program funded in the 2027–28 Budget, estimated at $1.8–$2.1 billion, including approximately $1 billion in the first five years.

    The Committee recommends that the Government establish a ten-year national national public compute investment program as part of the National AI Plan, funded in the 2027-28 Budget cycle. This is expected to cost $1.8–$2.1 billion over ten years, including approximately $1 billion in the first five years.

  4. Review the precise investment over time as technology, demand and market conditions change.

    The precise investment will need to be reviewed over time as technology, demand and market conditions change.

  5. Establish a decade-long funding roadmap in consultation with NCI and Pawsey.

    The Committee recommends that a decade-long funding roadmap be established in consultation with NCI and Pawsey as part of implementation of the National AI Plan.

  6. Consider whether the Australian AI Safety Institute's independent technical evaluation would benefit from a standing compute allocation on national research infrastructure of the kind operated by Pawsey and NCI.

    For example, the Australian AI Safety Institute has a security and safety mandate, including evaluating and testing frontier models. We encourage the Committee to consider whether this independent technical evaluation would benefit from a standing compute allocation on national research infrastructure of the kind operated by Pawsey and NCI.

  7. Ensure long-term investment in national computing infrastructure maintains geographically distributed capability and enables access by researchers, businesses and public institutions across Australia, including regional areas.

    Recommendation: Ensure that long-term investment in Australia's national computing infrastructure maintains geographically distributed capability and enables researchers, businesses and public institutions across Australia, including in regional areas, to access that capability.

  8. Recognise national public computing infrastructure as part of the enabling infrastructure for priority industries, including sectors combining AI with robotics, automation and advanced modelling.

    Recommendation: Recognise national public computing infrastructure as part of the enabling infrastructure for Australia's priority industries, including sectors in which AI is increasingly combined with robotics, automation and advanced modelling.

  9. Consider establishing dedicated access pathways through NCI and Pawsey for Australian SMEs developing and commercialising AI applications, supported by appropriate technical expertise.

    Recommendation: Consider establishing dedicated access pathways through NCI and Pawsey for Australian SMEs developing and commercialising AI applications, supported by appropriate technical expertise.

  10. Support use of national computing infrastructure to develop and test AI on significant Australian datasets, with governance suited to data sensitivity, including Indigenous data-governance principles where relevant.

    Recommendation: Support the use of Australia's national computing infrastructure to develop and test AI using significant Australian datasets, with governance arrangements appropriate to the sensitivity and significance of the data, including Indigenous data-governance principles where relevant.

  11. Include advanced computing and research-infrastructure specialists in national AI workforce and skills planning, and fund long-term development and retention of the specialist workforce.

    Recommendation: Include advanced computing and research-infrastructure specialists in national AI workforce and skills planning, and ensure long-term investment in national computing infrastructure includes funding to develop and retain the specialist workforce required to operate it.

  12. Consider the infrastructure and jurisdictional requirements of sensitive Commonwealth AI applications in the AI governance framework, and assess NCI and Pawsey's potential role in providing trusted computing capacity.

    Recommendation: The Australian Government should consider the infrastructure and jurisdictional requirements of sensitive Commonwealth AI applications as part of its AI governance framework, and assess the potential role of NCI and Pawsey in providing trusted Australian computing capacity for those applications.

  13. Alongside regulatory settings for AI infrastructure, establish a long-term investment framework for national public computing capability, including NCI and Pawsey.

    Recommendation: Alongside consideration of appropriate regulatory settings for AI infrastructure, the Australian Government should establish a long-term investment framework for Australia's national public computing capability, including NCI and Pawsey.

  14. Incorporate CARE alongside FAIR principles, engage Indigenous communities and governance bodies where relevant, and ensure Indigenous authority over Indigenous data and cultural knowledge in national AI, data and computing initiatives.

    This includes recognising the CARE Principles for Indigenous Data Governance alongside FAIR data principles, and ensuring appropriate engagement with Indigenous communities and Indigenous data-governance bodies where relevant. Recommendation: Ensure that national AI, data and computing initiatives incorporate appropriate Indigenous data-governance principles, including the CARE Principles, and mechanisms for appropriate Indigenous engagement and authority over the use of Indigenous data and cultural knowledge.

  15. Recognise AI safety, detection and harm prevention as important public-interest uses of national computing infrastructure when planning future capacity.

    Recommendation: Recognise AI safety, detection and harm prevention as important public-interest uses of Australia's national computing infrastructure when planning future capacity.

  16. Recognise computing security and resilience in AI capability planning; fund NCI and Pawsey’s security, specialist workforce, systems and supporting infrastructure alongside new equipment.

    Long-term investment must therefore provide for the security, specialist workforce, systems and supporting infrastructure needed to maintain that capability, not simply the purchase of new computing equipment. Recommendation: Recognise the security and resilience of Australia's national computing infrastructure as part of national AI capability planning, and ensure long-term investment in NCI and Pawsey includes the resources required to maintain secure and resilient national capability.

  17. Assess the AI Safety Institute's future computing requirements and the role NCI and Pawsey could play in providing independent computing capacity for AI safety research, testing and evaluation.

    Recommendation: Assess the Australian AI Safety Institute's future computing requirements and the role NCI and Pawsey could play in providing independent Australian computing capacity for AI safety research, testing and evaluation.

Read the original

Open in the interactive guide