InferaStack AI Grid
Energy-aware, distributed AI compute for Australia. InferaStack AI Grid connects GPUs, energy systems and networks across qualified sites, and runs each job where capacity, power and location rules allow.
How the grid works
Qualified sites
Each proposed site is assessed for safe operation, suitable infrastructure and agreed data-handling requirements. Capacity becomes available only after site acceptance and workload validation.
Power comes first
The grid is designed around each site's available energy and operating limits. Site safety and the owner's requirements come first. Energy performance must be measured before any savings are claimed.
Data decides the place
The grid is proposed for encrypted, non-identified inference and bounded batch jobs, and residential sites are designed to accept only encrypted, non-identified work. Identified personal, clinical or commercial data stays in your own environment or a certified facility, such as the dedicated GPU environments we design for NEXTDC facilities. Each workload's processing path needs approval. Distributed training is outside the current design.
What it will offer
Planned services include AI inference and reserved GPU capacity, with access through the InferaStack Gateway. Availability, data handling and service terms are agreed for each workload. Register interest, or start with a scoped inference pilot.
Validation status
We are validating the first pilot before expanding. Progress depends on measured results, site acceptance and customer demand.
| Step | What happens | Status |
|---|---|---|
| Today | The grid is in proof of concept and technical validation. The InferaStack Gateway is in development for customers. | Not run on hardware |
| Define and evaluate | Agree the pilot scope, authorised test data and success criteria before evaluation. | Next step, pending |
| Validate the system | Evaluate the proposed NVIDIA systems against the agreed performance, safety and operational requirements. | Proposed; not hardware-validated |
| Accept the first site | Complete site assessment and acceptance before making pilot capacity available. | No site accepted yet |
| Operational readiness | Validate service continuity and support arrangements against each workload's requirements. | Subject to validation |
| Repeat, then decide | Use measured pilot results and contracted demand to decide whether to expand to additional qualified sites. | Later |
NVIDIA systems and energy integration
NVIDIA systems
InferaStack AI Grid is planned to use NVIDIA systems for distributed AI compute. Suitability and performance will be assessed through pilot validation.
Energy integration
ConnectVPP is our energy integration partner. The work focuses on connecting site energy requirements with distributed AI compute, with safety and owner requirements guiding each site. The integration remains in technical validation. Explore energy integration →
Energy integration partnerInferaStack is an NVIDIA Inception member. AI grid is the industry category described in NVIDIA's public material; InferaStack AI Grid is InferaStack's product within that category. The name does not imply NVIDIA approval, validation, co-design or endorsement. NVIDIA's AI grid overview (opens in a new tab)
Host a site or connect energy
We are looking for site owners and energy operators for the first site. Terms are agreed per site after its checks; no site has been accepted yet. Enquiries are expressions of interest during validation.