Sovereign AI
Sovereign programme record · Australia · 2025
Australia: ~$26M in government funding
Australia: National AI Plan + GovAI Sovereign Compute
- Announced—
- Government funding~$26M
- Committed—
- Disbursed—
Operating compute / delivered hardware. Reported operational or delivered: 120 NVIDIA H200 (physical hardware); 16 NVIDIA A100 (physical hardware); 640 NVIDIA V100 (physical hardware); 768 AMD Instinct MI250X (physical hardware).
Hardware on record
| Model | Count | Status | Organisation / role |
|---|---|---|---|
| NVIDIA H200 | 120 | Operational | NCI Gadi (government, ANU) |
Physical hardware: Gadi’s H200 partition: 30 nodes with four GPUs each. Primary source | |||
| NVIDIA A100 | 16 | Operational | NCI Gadi (government, ANU) |
Physical hardware: Gadi’s A100 partition: two nodes with eight GPUs each. Primary source | |||
| NVIDIA V100 | 640 | Operational | NCI Gadi (government, ANU) |
Physical hardware: Gadi’s V100 partition: 160 nodes with four GPUs each. Primary source | |||
| AMD Instinct MI250X | 768 | Operational | Pawsey Setonix (government) |
Physical hardware: Setonix’s 192 GPU nodes each contain four physical MI250X modules; each module has two compute dies. Primary source | |||
Empanelled partners · 6
| Partner | Accelerators | Model | Stage | Source |
|---|---|---|---|---|
| National Computational Infrastructure (NCI) | — | NVIDIA H200; NVIDIA A100; NVIDIA V100 | Commissioned | Primary source · 2025-12-02 |
Operates Gadi at the Australian National University. Its GPU partitions comprise 30 four-H200 nodes, two eight-A100 nodes and 160 four-V100 nodes. Researchers use the system for simulation, data analysis and AI.
| ||||
| Pawsey Supercomputing Research Centre | — | AMD MI250X; NVIDIA GH200 | Commissioned | Primary source · 2025-12-02 |
Operates Setonix in Perth, an HPE Cray EX system with 192 GPU nodes. Each node contains four physical MI250X cards, each with two software-visible GPU dies. The separate Setonix-Q pilot provides GH200-based quantum simulation and access to quantum computers through a merit allocation scheme.
| ||||
| HPE Cray | — | — | Commissioned | Research database · 2022-01-01 |
Supplied the HPE Cray EX architecture used by Pawsey’s Setonix supercomputer.
| ||||
| Firmus (Project Southgate) | — | NVIDIA (model unspecified) | Announced | Press or analyst · 2025-11-14 |
Project Southgate is a commercial AI infrastructure partnership with CDC Data Centres and NVIDIA. Firmus’s October 2025 announcement specifies an initial A$4.5B investment in Melbourne and a planned national expansion to 1.6 GW through 2028.
| ||||
| Microsoft | — | — | Announced | Research database · 2026-04-23 |
Announced A$25B (US$18B) in Australian digital infrastructure investment on 23 April 2026, with capital and operating expenditure through 2029. The announcement builds on its October 2023 investment and follows the December 2025 National AI Plan.
| ||||
| Amazon Web Services | — | — | Announced | Research database · 2025-06-01 |
Announced A$20B for 2025–2029 to expand, operate and maintain Australian data centres. The investment serves AWS cloud and AI customers; it is a corporate infrastructure budget.
| ||||
External dependence
NVIDIA hardware: High · US technology: High. These are editorial assessments based on the programme’s procurement disclosures.
Named partners: CSIRO, Department of Industry, Science and Resources, OpenAI, NextDC, AWS, Microsoft, NCI, Pawsey Supercomputing Research Centre.
The December 2025 National AI Plan brings together existing funding for AI research and adoption with measures to expand local infrastructure. It describes GovAI as an Australian-hosted service through which public agencies can develop AI applications. NCI and Pawsey operate research computing systems; Microsoft, AWS and Firmus have announced separate commercial infrastructure investments.
Data-quality caveats on this record · 4
- The plan combines more than A$460M of existing AI and related funding. Its A$39.9M measure supports AI policy and capability, including the National AI Centre.
- Gadi has 120 H200, 16 A100 and 640 V100 GPUs. Setonix has 768 physical MI250X cards across 192 nodes; software exposes their 1,536 GPU dies separately. Both systems support research workloads, including AI.
- Microsoft, AWS and Firmus announce investment in their own commercial infrastructure. Their budgets have different dates and scopes, including construction, operations and expansion.
- The A$39.9M policy and capability measure is approximately US$26.2M at the Federal Reserve H.10 rate of US$0.6556 per Australian dollar on 2 December 2025.
The evidence trail
15 cited sources- Australia National AI Plan (official page) — Primary source · accessed 2026-04-07
- National AI Plan launch announcement — Primary source · accessed 2026-04-07
- Official DISR announcement — Primary source · accessed 2026-03-28
- 2024 Federal Budget AI allocations breakdown — Press or analyst · accessed 2026-03-28
- Sovereign Australia AI private cluster — Press or analyst · accessed 2026-03-28
- OpenAI/NextDC deal (private) — Press or analyst · accessed 2026-03-28
- Amazon Web Services: A$20 billion DC investment announced June 2025 — PRIVATE, not sovereign — Primary source · no accessed date on record
- National AI Plan: GovAI hosting service and scope of existing public AI funding — Primary source · accessed 2026-09-08
- NCI: Gadi GPU node and accelerator specifications — Primary source · accessed 2026-09-08
- Pawsey: 192 Setonix GPU nodes, four physical MI250X cards per node and two dies per card — Primary source · accessed 2026-09-08
- Pawsey: Setonix uses the HPE Cray EX architecture — Primary source · accessed 2026-09-08
- Firmus: Southgate investment denominated in Australian dollars and planned national expansion — Primary source · accessed 2026-09-08
- Microsoft: April 2026 Australian investment amount, currency and capital-plus-operating scope — Primary source · accessed 2026-09-08
- Amazon: A$20B data-centre investment plan for 2025–2029 — Primary source · accessed 2026-09-08
- Pawsey: physical MI250X cards, software-visible dies and separate GH200 allocation — Primary source · accessed 2026-09-08
Machine accessThe same record, machine-readable: the evidence-trail feed and the derivation-chain export.