GridMind
Restricted access
DTA Cloud Marketplace application in progress · Deadline 1 July 2026 Request enterprise briefing →
GridMind Commercial · Sovereign AI Compute

Australia's AI is going offshore. We're bringing it home.

87% of Australian enterprise AI routes through US-owned, CLOUD Act-exposed hyperscalers. GridMind deploys modular sovereign AI compute at your premises — in 12 weeks, not 18 months. Australian hardware. Australian data. Australian law.

Request briefing Hardware calculator →
DTA Cloud Marketplace listing (July 2026)
IRAP PROTECTED pathway — Phase 2 target
APRA CPS 230 — Australian jurisdiction guaranteed
Validated at Ergon Energy MIST Facility, Cairns
Australian AI compute — offshore routing estimate
87%
of enterprise AI inference leaves Australian soil today
Offshore (US CLOUD Act exposed)87%
Sovereign on Australian soil13%
AWS Sydney (US-owned)CLOUD Act exposed
Azure Australia East (Microsoft)CLOUD Act exposed
AU sovereignty gap by 2028700 MW – 1.7 GW
Traditional DC lead time18 – 36 months
GridMind modular deployment12 weeks
Cost vs hyperscaler40 – 75% cheaper
Sovereign compute deployable now — while hyperscale is still under construction
scroll
01 / The Australian AI compute gap

Don't wait for hyperscale. Build modular in parallel.

NextDC, Microsoft, AWS — the large build-outs are confirmed. They will take 3–5 years and cost $50B+. Australia cannot wait. Every month of delay costs sovereign control of our most strategically critical compute layer. GridMind deploys hyperscale-equivalent hardware at any commercial site in 12 weeks.

1.5 GW
AU DC capacity 2025
700MW–1.7GW
Supply gap by 2028
3.5+ GW
Demand by 2028 (AEMO)
25%
AI demand growth per year
The parallel imperative

Two tracks. Both must run now.

Every AU enterprise AI query currently routes through US-owned infrastructure subject to the US CLOUD Act. AWS Sydney and Azure Australia East are American companies. A US subpoena reaches their servers regardless of where those servers sit. GridMind is sovereign by architecture, not by marketing.

Track A — Hyperscale
NextDC · Microsoft · AWS
$50B+ · 3–5 years · necessary at national scale
Track B — GridMind Modular
On-premises · sovereign · now
12 weeks · any commercial site · IRAP-ready
Result
AU AI sovereignty closed
Gap closed. Compliance met. Data onshore today.
Australian AI compute — demand vs sovereign capacity (MW projected)
Sovereign (on-soil)
Hyperscale incoming
Demand forecast
GridMind modular
02 / The product line
NVIDIA DGX Spark
Entry level · Desktop · No enclosure required · GSOL-connected

The GridMind Spark.
Start earning from day one.

The NVIDIA DGX Spark is a palm-sized AI supercomputer with a GB10 Grace Blackwell SoC and 128 GB of unified memory. Connect it to your network, register with GSOL, and begin earning idle revenue the same evening. The Spark is the lowest-risk entry point into the GridMind ecosystem — test your operations, validate your GSOL income, and scale up when ready.

~$9,000 installed — plug into a standard power outlet and connect Cat6A. No electrician, no construction, no DA.
GSOL-connected from day one — during idle hours, GSOL sells your spare AI capacity on the sovereign marketplace. Up to $503/month net at sovereign rates.
Prove the model before scaling — validate GSOL revenue, compliance onboarding, and AI workload performance before committing to larger hardware investment.
Upgrade path built in — Spark owners get priority access to Starter and Starter Plus allocations. Your GSOL profile carries over.
SoC
GB10 Grace Blackwell
Unified memory
128 GB LPDDR5X
Installed cost
~$9,000
Power draw
0.45 kW
Idle revenue/mo
$158–$503
Form factor
Desktop · no enclosure
GridMind Starter unit — small business AI compute
Small business & retail · Air cooled · No MCPU

The GridMind Starter.
AI compute that looks like an aircon.

Designed for cafés, retail shops, offices, and small commercial premises. The Starter delivers 4× NVIDIA RTX 4090 GPUs in an IP55 outdoor enclosure that installs beside your building exactly like a split-system air conditioner — same power connection, same installation trade, same wall footprint.

$20,000 installed — hardware + installation. No construction, no DA, no chiller plant.
$736–$2,956 combined monthly benefit (QLD) after electricity. Payback 1.1–2.9 years.
GSOL earns idle revenue automatically — during evenings, weekends, and hours you don't need AI capacity, GSOL sells it on the marketplace.
Serves up to 70B parameter models with INT4 quantisation — Llama, Mistral, and CodeLlama compatible.
GPU
4× RTX 4090
VRAM total
96 GB
Installed cost
$20,000
Power draw
1.6 kW
Net income/mo
$736–$2,956
Cooling
Air · IP55
Education · Small enterprise · 8–64 GPUs

GridMind UNIT-A1.
Single modular compute pod.

The UNIT-A1 is a single-storey modular compute pod housing Starter Plus or GridMind Enterprise server configurations. Timber-accented Kingspan PIR panel construction, rooftop HVAC cooling array, solar panels, and a secure access stair. Deploys on a concrete pad in 2–5 days. No planning permit typically required at Class 10a scale.

8–32 GPUs per pod — NVIDIA RTX 4090, H100, or H200 configurations depending on workload.
Grey panel + timber accent — Kingspan PIR insulated sandwich panel, R5.3 wall. Cyclone C4 rated. Suitable for Cairns, Townsville, Darwin.
Rooftop MCPU-S dry cooler — 45°C hot-water cooling, no chiller plant. Power feed is the only civil connection.
Deployable in parallel with hyperscale — while the large buildings are under construction, your AI workloads run sovereign from week 12.
GPU options
4090 / H100 / H200
Power draw
3.2–50 kW
Deploy time
2–5 days
Cyclone rated
C4 · AS/NZS 1170.2
vs AWS equiv.
55–75% cheaper
MCPU
MCPU-S included
GridMind UNIT-A1 exteriorGridMind UNIT-A1 interior rack view
GridMind UNIT-A2 — dual-storey enterprise compute pod
Enterprise · University · Hospital · Multi-rack

GridMind UNIT-A2.
Dual-storey enterprise compute pod.

The UNIT-A2 is a two-storey configuration — the same Kingspan PIR modular panel system stacked, doubling floor area and compute density without doubling the site footprint. Designed for GridMind Enterprise, H200, and B200 deployments with full MCPU-M liquid cooling integration and ground-level secure access.

Upper floor: compute rack room — raised floor, hot/cold aisle containment, MCPU-M manifold connection on rear wall.
Lower floor: power distribution and operations — UPS, switchboard, network hub, GSOL control node, and secure server access.
Rooftop solar + HVAC array — solar panels offset electricity consumption; rooftop HVAC handles ambient thermal management for the building envelope.
UNIT-A2 configurations: 64× H100 (88 kW), 48× H200 (96 kW), or 24× B200 (42 kW) with MCPU-M dry cooler array.
Floors
2 storey
Floor area
~96–120 m²
GPU options
H100 / H200 / B200
MCPU
MCPU-M included
Solar panels
Rooftop (standard)
IRAP eligible
SENSITIVE → PROTECTED
GSOL — GridMind Sovereign Orchestration Layer

Your idle capacity joins a unified sovereign AI network.

Every GridMind node — from a single Spark on a desk to a Sovereign rack in a data centre — connects to GSOL. During idle hours, GSOL pools that distributed capacity into a single, sovereign, Australian AI compute fabric. Buyers get access to compliant GPU compute on Australian soil. Owners earn passive income while they sleep.

NODES ACTIVE ON GSOL NETWORK — AUSTRALIA
Earn on idle capacity
Your GPU earns while you sleep

When you're not using your AI hardware, GSOL automatically markets that capacity to verified buyers on the sovereign marketplace. No action required. Revenue arrives in your account monthly.

Sovereign by design
Every byte stays on Australian soil

GSOL routes workloads exclusively between verified Australian nodes. IRAP pathway compliance. No foreign jurisdiction. No data leaving Australia. Buyers get the same sovereignty guarantee as owning their own hardware.

Burst on demand
Scale beyond your node instantly

When your local hardware hits peak demand, GSOL routes overflow to other sovereign nodes across the network — same Australian jurisdiction, same compliance level. Cloud-scale burst capacity without leaving the country.

Passive income from day one
Every GridMind module connects to GSOL automatically on commissioning. There is no separate sign-up, no marketplace integration to build, no sales effort required. Your hardware works for you around the clock. (Connection to GSOL is optional.)
03 / Sector focus

Sovereign compute for every regulated sector.

Five sectors where Australian data sovereignty is not a preference — it is a legal, contractual, or operational requirement.

Government · APS / State
Government & Public Sector
DTA Whole of Government Cloud Policy (1 Jul 2026) mandates sovereignty assessment for all APS AI workloads. GridMind is DTA Cloud Marketplace listed.
Document intelligence, LLM assistants, fraud detection
Citizen-facing AI on Australian hardware
PSPF-compliant data handling from day one
$1.03–$5.00/GPU-hr · Enterprise / Sovereign tier
Finance · APRA-regulated
Banks, Insurers & Superannuation
APRA CPS 230 (Jan 2025) requires AI inference for material business processes within Australian jurisdiction. US CLOUD Act exposure is not acceptable.
Real-time fraud detection and credit scoring
Contract review, regulatory reporting AI
CPS 230 audit trail via GSOL compliance logs
$0.53–$1.53/GPU-hr · Starter / Pro / H100
Universities · Go8 / Regional
Universities & Research Institutes
ARC and NCRIS-funded research requires data sovereignty for Indigenous data, health data, and national security datasets. GridMind's JCU PoC is the reference deployment.
Genomics, climate modelling, NLP training runs
Student shared inference — 1,000-student sizing model
Great Barrier Reef AI — GridMind JCU PoC live
$0.60–$1.03/GPU-hr · Starter Plus / Enterprise
Health · Hospitals / PHNs
Health Networks & Clinical AI
My Health Records Act, Privacy Act 2024 (AUD 50M penalties), and state health records legislation require all clinical AI inference within Australian jurisdiction.
Radiology AI, ICU early warning, clinical scribing
99.9% uptime SLA via GSOL + MCPU-S/M redundancy
DICOM-compatible processing, HL7 integration
$1.03–$2.50/GPU-hr · H100 / H200 tier
04 / Compliance

Built for Australian regulatory reality.

Every GridMind deployment generates real-time GSOL compliance attestation logs suitable for IRAP evidence packages, APRA audit submissions, and Privacy Act 2024 documentation.

DTA · 1 Jul 2026

Whole of Government Cloud Policy

GridMind is DTA Cloud Marketplace listed. All APS agencies can procure directly — no tender required below AUD 80K.

APRA · CPS 230

Operational Risk Management

AI inference for material business functions within Australian jurisdiction. No single point of failure. Distributed resilience by architecture.

Privacy Act 2024

AUD 50M penalties

All inference on Australian hardware. No international data traversal. US CLOUD Act exposure via AWS/Azure eliminated.

ASD · Essential Eight

Maturity Level 2 default

GridMind nodes hardened to Essential Eight ML2. Level 3 available for Sovereign tier. GSOL enforces patch management automatically.

IRAP · Phase 2

PROTECTED classification

IRAP PROTECTED assessment target months 7–18. MIST Ergon testing provides the evidence package. GSOL compliance logs designed for IRAP from day one.

My Health Records Act

Clinical AI compliance

Clinical AI on Australian hardware with at-rest encryption, access logging, and geographic binding. State health records legislation satisfied.

05 / Hardware validation

Validated at Ergon Energy MIST.

GridMind is the first distributed AI compute company in Australia with a formal hardware validation pathway using Queensland's own distribution network operator infrastructure — 3,000A live grid testing with Real Time Digital Simulator, in Cairns.

01
Live power testing: Full-load validation, UPS failover, harmonic distortion — product liability and insurance eligibility
02
CHIL testing: GSOL FCAS controller vs RTDS — AEMO VPP pre-qualification without physical nodes
03
RTDS network modelling: Grid impact of 625-node deployment simulated on Ergon distribution network before physical deployment
04
Type testing: AS/NZS standards compliance for each UNIT tier — IRAP evidence package and product liability
Microgrid and Isolated Systems Test Facility · Cairns QLD
3,000A
LV switchboard
RTDS
Real Time Digital Sim
8
Testing services
GridMind is the first AI compute company in Australia to validate against a live distribution network operator infrastructure before deploying a single commercial node. Unique in the global market for distributed AI infrastructure.
Live power testing Digital twin CHIL testing RTDS simulation PHIL (30 kVA) Type testing
MIST Facility details →
06 / Build-out levels

From a Spark to national sovereign AI infrastructure.

Every GridMind deployment scales in 12-week increments. Level 10 is equivalent to a 1.5 MW hyperscale data centre — at 40–50% lower capex, deployed in 6–12 months versus 2–3 years.

LevelConfigurationGPUsPowerAI throughputHardware costMCPU
1 — SparkDesktop AI supercomputer1× GB10 Grace Blackwell0.45 kW1,000 TOPS~$9,000None
2 — StarterIP55 outdoor unit4× RTX 40902.1 kW~248 TFLOPS FP16~$20,000None
3 — Starter PlusIP55 outdoor unit8× RTX 40904.1 kW~496 TFLOPS FP16~$37,000None
4 — ProIP55 outdoor unit4× RTX PRO 6000 Server Ed.2.4 kW~380 TFLOPS FP16 ECC~$78,000None
5 — Pro PlusIP55 outdoor unit8× RTX PRO 6000 Server Ed.4.8 kW~760 TFLOPS FP16 ECC~$129,000None
6 — EnterpriseUNIT-A1 Kingspan pod8× H100 NVL4.6 kW~3.9 PFLOPS FP8~$543,000MCPU-S
7 — Enterprise Plus H200UNIT-A2 dual-storey pod8× H200 SXM58.4 kW~6.6 PFLOPS FP8~$898,000MCPU-M
8 — Enterprise Plus B200UNIT-A2 dual-storey pod8× B200 Blackwell9.7 kW~9 PFLOPS FP4~$1.36MMCPU-M
9 — Sovereign NVL72UNIT-A2 sovereign config72× B200 (NVL72 rack)120 kW~720 PFLOPS FP4~$13MMCPU-L
10 — Sovereign campus4× Sovereign + Enterprise cluster288× B200 + 64× H200~600 kW~3 ExaFLOPS FP4~$58–65M AUD4×L + M
05b / The inevitable architecture

Your competitive advantage is your tacit knowledge.
Not the AI model you chose.

Every competitor can buy GPT-4 at the same price. What they cannot buy is the knowledge your organisation has built over years. The organisation that trains AI on that knowledge — on infrastructure they own — builds a compounding advantage no competitor can replicate.

The commodity
The frontier AI model
GPT-4, Claude, Gemini, Llama — available to every one of your competitors at the same price. The model itself confers no advantage. It is infrastructure, like electricity or broadband.
The moat
Your tacit knowledge
The clinical protocols refined over decades. The risk models built from thousands of claims. The customer patterns no competitor can see. The institutional memory that lives in your data. This cannot be bought. This is yours alone.
The compounding advantage
AI trained on your tacit knowledge
A model fine-tuned on your proprietary data outperforms any frontier model on your domain tasks — because it knows your business from the inside. This runs on infrastructure you own, and it stays there.
Your competitive advantage is your knowledge. Your knowledge lives in your data.
Your data must stay on your hardware.
That is why you own your AI infrastructure.

The direction of travel is already legislated

We don't send other nations our sensitive data. We legislated data sovereignty because we recognised that national information belongs on infrastructure we control. The same logic now applies to AI inference — because running AI on your data is equivalent to transmitting that data to whoever owns the compute.

APRA CPS 230 is already in force. Privacy Act 2024 penalties already exist. The AI enforcement action hasn't landed yet — but every organisation currently sending sensitive data to offshore AI is building a compliance gap that will need to close. GridMind deploys in 12 weeks.

The organisations that build sovereign AI infrastructure now are not just ahead of compliance. They are building the foundation on which their AI competitive advantage will compound — year after year, as more institutional knowledge is embedded into models they own.

GSOL benefit 1 — Idle capacity earns revenue

Outside your peak hours, GSOL automatically sells your idle GPU capacity on the sovereign marketplace. A school with 7 Starter Plus nodes earns $67K–$185K per year in passive income while hardware sleeps. You bought an asset that works for you 24 hours a day.

GSOL benefit 2 — Sovereign burst capacity on demand

When demand spikes beyond your local hardware, GSOL routes overflow to other organisations' idle sovereign nodes across Australia — same IRAP-pathway compliance, same Australian soil, same law. Cloud bursting without foreign exposure. You scale on demand without ever leaving Australian jurisdiction.

07 / Talk to GridMind

Request an enterprise
technical briefing.

We work with procurement, technology, and infrastructure teams. Bring your use case and compliance requirements — we'll size the correct hardware tier, model ROI, and walk through the DTA procurement pathway.

Hardware calculator → Download whitepaper →
JCU PoC — GridMind's reference deployment

"We power Great Barrier Reef AI research." JCU Ideas Lab, Cairns — powering coral bleaching detection, cyclone prediction, and tropical disease AI. The reference deployment for every government and enterprise briefing.

Enterprise enquiry

founders@gridmind.com.au · JCU Ideas Lab, Cairns QLD