MICRO / EDGE
Small GPU footprints at commercial or edge sites for inference, computer vision and local processing close to where data is generated.
Peregrine ComputeGrid
Distributed AI infrastructure powered by Australia's energy edge.
ComputeGrid is being developed as a distributed AI infrastructure platform connecting GPU capacity across data centres, commercial sites and renewable-energy-enabled locations. It is being progressively deployed as capacity is commissioned.
The concept
Compute that follows capacity, power and demand.
Most AI infrastructure is concentrated in a small number of large facilities. That model suits tightly coupled training, but much AI work — inference, fine-tuning, batch processing, computer vision — can run well across a network of smaller GPU locations.
ComputeGrid is being developed to connect such locations under a single orchestration layer, so customers see one platform while workloads are placed where capacity, power and data requirements align. Distributed nodes complement, rather than replace, hyperscale data centres and the cloud environments organisations already use.
Architecture
One platform, many locations.
CUSTOMER
PEREGRINE PLATFORM
SECURE ORCHESTRATION
GPU SCHEDULER
DISTRIBUTED COMPUTE LOCATIONS
Orchestration layer manages
- Workload placement
- GPU scheduling
- Availability
- Monitoring
- Metering
- Security
- Power / thermal information
- Capacity allocation
- Billing data
Potential technologies
- Kubernetes
- NVIDIA scheduling
- vLLM
- Triton
- TensorRT
- Containerised workloads
Node types
Potential node types — not fixed SKUs.
Node configurations are determined by site conditions, power availability and the workloads expected to run there.
GPU NODE
Standard GPU servers deployed at data-centre, commercial or renewable-energy-enabled sites for inference, fine-tuning and batch workloads.
HIGH-DENSITY NODE
Higher-power GPU systems placed where power and cooling permit, suited to training and sustained high-utilisation workloads.
Locations
- Data Centres
- Commercial Sites
- Industrial Sites
- Renewable Energy Sites
- Edge Locations
Compute meets energy
AI infrastructure designed for the energy transition.
AI consumes significant power. Australia has substantial distributed solar, battery and renewable resources. Peregrine is developing models to place appropriate compute closer to available energy.
Renewable-energy integration may reduce reliance on grid-only energy depending on site design and operating conditions.
SOLAR / GRID / BATTERY
Energy sources
ENERGY-AWARE SITE
Power and thermal data
GPU COMPUTE NODE
Placed with power
PEREGRINE COMPUTEGRID
Orchestration
AI WORKLOAD
Delivered securely
Interested in ComputeGrid?
Talk to us about distributed inference, fine-tuning and capacity expansion, or register a forecast so future capacity can be planned with your requirements in view.