Serious accelerated compute without a capital programme
On-demand NVIDIA accelerated compute, hosted in New Zealand
GPU capacity is the hardest thing to buy well: prices move weekly, lead times stretch to months, and a bad configuration is a six-figure mistake you live with for three years. GPU as a Service removes that risk entirely.
You consume NVIDIA accelerated compute from our New Zealand data centres on a monthly basis, scale up or down as projects change, and never carry the depreciation. Capacity is delivered from a modern NVIDIA fleet — H100, H200, L40S and RTX PRO 6000 class accelerators — on all-flash VAST Data storage with GPUDirect RDMA and a non-blocking high-speed fabric.
NZ data sovereignty • Monthly billing • Self-service configurator • Local support
Hosted bare metal
Dedicated GPU servers in an ASI data centre with secure remote access — full performance, no hypervisor tax.
What we deliver:
- Full GPU performance without virtualisation overhead
- Secure remote access from your network
- Ideal for training, fine-tuning and research workloads
Virtualised GPU
Virtual machines with attached GPUs on ASI Cloud, for flexible allocation and fractional workloads.
What we deliver:
- Flexible GPU allocation per workload
- Fractional workloads and prototyping
- Self-service through the ASI Cloud configurator
On-site hosted
Hardware placed in your facility, managed end-to-end by ASI.
What we deliver:
- GPU capacity inside your own perimeter
- Fully managed by ASI engineers
- Data never leaves your site
White-label
Partners can deliver ASI GPU capacity under their own brand — talk to us about partner arrangements.
What we deliver:
- Deliver sovereign GPU capacity under your brand
- Partner pricing and support model
- Full NZ data residency for your customers
What You Get
Monthly billing, no lock-in
Pay as you go with bulk pricing available against annual commitments.
Self-service configuration
Configure through the ASI Cloud configurator with transparent, current pricing.
Right-sized advice
We will tell you when an L40S beats an H100 for your workload, and save you the difference.
Full data residency
Workloads, model weights and datasets stay within New Zealand.
Storage that keeps up
All-flash VAST Data with NVMe/TCP and GPUDirect, plus S3 and archive tiers.
Local support
NZ-based engineers, named account team, no offshore escalation queue.
Need GPUs Without the Capital Programme?
Best for organisations running training, fine-tuning, inference or research workloads that need serious accelerated compute — and that cannot send data offshore.
