Capability 1.1.2.1.1
Cloud Training Compute Infrastructure as a Service (public cloud)
Definition
Large-scale compute clusters made available locally as Infrastructure as a Service (IaaS). Individuals, companies or organisations can rent computing capacity remotely and on demand for AI model training, often using specialised chips (accelerators) such as Graphics Processing Units (GPUs) and Tensor Processing Units (TPUs).
Examples
- Multinational hyperscalers (AWS, Google Cloud, Azure, Oracle Cloud). Local/regional AI cloud providers (e.g. Sharon AI).
Indicators
- Number and scale of local cloud clusters
- Availability and cluster sizes of advanced chips known as accelerators (such as GPUs like NVIDIA H100 or equivalents)
- Approximate number of H100-equivalent accelerators
National AI Plan
PLAN COMMITMENT
Significant commitment
General: Commits to mapping compute infrastructure to identify gaps and guide future investment (Action 1), but unclear if or how mapping exercise will capture the different types of compute articulated in this layer and no details on when mapping will occur. The Data Centre Expectations (a deliverable under Action 1 of the plan) include an expectation that 'enabling access to compute for Australian start-ups, innovative small businesses, researchers and not-for-profits on favourable terms' but the type of compute is not specified. Specific: Australian Government and Anthropic memorandum of understanding (MOU) (signed 1 April 2026) includes a provision that 'Anthropic intends to take steps towards ensuring benefits to local industry and researchers through access to its infrastructure ...'. In the Australian Government and Microsoft MOU (signed 23 April 2026), Microsoft commits to meeting the Australian Government's expectations for data centres and AI infrastructure developers, which include the expectation that (unspecified) compute be provided for Australian start-ups, innovative small businesses, researchers and not-for-profits on favourable terms. Significant commitment in the plan strongly aligns with recommended action to prioritise building agaency in this enabling capability.
Stocktake
‘AI Compute Sovereignty: Infrastructure Control Across Territories, Cloud Providers, and Accelerators'. Hawkins et al. SSRN. June 2025. http://dx.doi.org/10.2139/ssrn.5312977
‘Australia’s artificial intelligence ecosystem: growth and opportunities'. National AI Centre, Department of Industry, Science and Resources. 25 June 2025. https://www.industry.gov.au/publications/australias-artificial-intelligence-ecosystem-growth-and-opportunities
‘Made in Australia: Our AI opportunity'. Australian Academy of Technological Sciences & Engineering. 22 August 2025. https://www.atse.org.au/what-we-do/strategic-advice/made-in-australia-our-ai-opportunity/
Is the Stocktake missing evidence?
Share additional information for potential future Assessments
