Data 2026-07

Sovereign AI

Announced commitments against government budgets and reported disbursements, by country.

Sovereign programme record · Australia · 2025

Australia: National AI Plan + GovAI Sovereign Compute: $280M announced · disbursement not on record

Every figure below derives from the same record the dashboard reads — this page is the programme as a citable object: the reconciliation with its denominator stated, the decomposition with its residual, and the dated evidence behind each claim.

Announced~$280M
Government-only floor$24M
Committed~$24M
Disbursed—
No disbursement figure on record.

The reconciliation. Four figures on one dollar scale, each bar a share of the largest. The reality ratio above divides disbursed by the government-only floor ($24M), and the field it divides by is stated so the division is reproducible. A hatched track is an absent figure carried with its reason — a government that does not publish disbursement is a finding, not a zero. Tildes mark estimated figures.

Pipeline stage: Deployed (primary-verified evidence) — 2,312 accelerators tracked. The stage is a Scrutica derivation (tier 4), computed from the figures above, never asserted by the source.

Hardware on record

Model	Count	Status	Owner
NVIDIA H200	120	operational	NCI Gadi (government, ANU)
NVIDIA A100	16	operational	NCI Gadi (government, ANU)
NVIDIA V100	640	operational	NCI Gadi (government, ANU)
AMD Instinct MI250X	1,536	operational	Pawsey Setonix (government)

Empanelled partners · 6

Partner	GPUs	Model	Stage	Award
National Computational Infrastructure (NCI)	—	NVIDIA V100; NVIDIA DGX A100	commissioned	Primary source · 2025-12-02
Pawsey Supercomputing Research Centre	—	AMD MI250X; NVIDIA GH200	commissioned	Primary source · 2025-12-02
HPE Cray	—	—	commissioned	Research database · 2022-01-01
Firmus (Project Southgate)	—	NVIDIA (model unspecified)	announced	Press or analyst · 2025-11-14
Microsoft	—	—	announced	Research database · 2026-04-23
Amazon Web Services	—	—	announced	Research database · 2025-06-01

External dependence

NVIDIA hardware: high · TSMC fabrication: high · US technology: high · BIS jurisdiction reach: low. Dependence levels are Scrutica classifications (tier 4) from the programme’s own disclosures — which hardware it buys, where it fabricates, and whose export jurisdiction its supply chain crosses.

Named partners: CSIRO, Department of Industry, Science and Resources, OpenAI, NextDC, AWS, Microsoft, NCI, Pawsey Supercomputing Research Centre.

National AI Plan is primarily governance, adoption, and ecosystem strategy. No dedicated sovereign compute program comparable to Gulf states. GovAI is a hosting platform, not a GPU procurement program.

What this record flags about itself · 5

A$460M consolidates existing programs rather than new spend; new dedicated AI compute ~A$39.9M/5yr
Australia has no dedicated sovereign AI GPU cluster, but national HPC systems have significant GPU resources: NCI Gadi (776 GPUs: 120 H200 + 16 A100 + 640 V100) and Pawsey Setonix (1,536 MI250X). These are general HPC, not AI-optimized.
A$100B+ figures in press refer to private DC investments (OpenAI/NextDC A$7B, AWS A$20B, Microsoft A$5B)
Government direct compute spend is in the low tens of millions AUD
'Sovereign Australia AI', a Blackwell GPU cluster, is a private entity despite the name.

The evidence trail

7 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

Machine surfacesThis record is one object under one id: the evidence-trail feed and the derivation-chain export serialize the same programme this page renders.