SXM5 · 141GB
H200
- Memory141 GB HBM3e
- Mem BW4.8 TB/s
- FP81,979 TFLOPS
- FabricNDR 400G IB
$25.44/ node · hr
Sovereign AI compute · AU · SEA · EMEASovereign GPU
Sovereign GPU
clusters, deployed
in your region.
Navigate Network is an Australian neocloud with regions across Australia, South-East Asia and the Middle East. We run dedicated H200, RTX 5090, RTX PRO 6000 and B300 fleets — sized for training runs, inference workloads and everything between.
01 · Compute
The fleet, line by line.
Three Hopper-and-Blackwell-class generations, all wired on NVIDIA Quantum-2 InfiniBand. Reserve months in advance, or burst on-demand by the hour.
Blackwell · 96GB
RTX PRO 6000
- Memory96 GB GDDR7
- Mem BW1.79 TB/s
- FP84,000 TFLOPS
- FabricPCIe 5.0 + RoCE
$16.80/ node · hr
Blackwell · 32GB
RTX 5090
- Memory32 GB GDDR7
- Mem BW1.79 TB/s
- FP83,352 TFLOPS
- FabricPCIe 5.0
$8.40/ node · hr
Blackwell Ultra · 288GB
B300
- Memory288 GB HBM3e
- Mem BW8.0 TB/s
- FP430 PFLOPS
- FabricNVL72 + NDR
Contact/ sales
AU
Australia
- SYD-1SydneyLive
- SYD-2SydneySoon
- MEL-1MelbourneLive
- PER-1Perth2027
SEA
South-East Asia
- SIN-1SingaporeLive
- KUL-1Kuala LumpurLive
- JKT-1JakartaSoon
- BRN-1BruneiSoon
EMEA
Middle East & EMEA
- DXB-1Dubai, UAELive
- RUH-1Riyadh, KSASoon
02 · Regions & sovereigntyPick your region.
Pick your region.
Your data stays there.
Every byte of training data, every checkpoint, every customer artefact is processed and stored in the region you choose — AU, SEA or EMEA. No silent cross-region replication, no telemetry shipped offshore.
- 01Per-region data residency, contractually guaranteedPin your workload to AU, SEA or EMEA. Region attestations are signed and verifiable; data never crosses the boundary without your sign-off.
- 02AU sovereign tier — IRAP-aligned, ISO 27001 in flightDesigned against the ASD Information Security Manual. PROTECTED-suitable deployments available on request out of SYD-1 and MEL-1.
- 03Middle East tier — UAE & KSA jurisdictionDubai- and Riyadh-anchored regions, aligned with UAE Information Assurance and the NCA Essential Cybersecurity Controls. SCC-ready data processor terms.
- 04BYOK with hardware key custody, in every regionBring your own KMS or use our HSM-backed envelopes. Keys never leave the region they were issued in.
03 · Use cases
Built for the full training-to-inference arc.
From early experimentation to production endpoints — one provider, one bill, one set of SLAs.
Foundation model training
Reserved H200 / B300 pods, full bisection bandwidth, Slurm-native scheduling. Multi-week runs without preemption.
Fine-tuning & post-training
Spin up an 8-GPU node in under a minute. Bring your trainer, mount your bucket, walk away.
Production inference
Single-region or multi-AZ deployments behind your VPC. Predictable latency from AU, SEA or EMEA endpoints, no egress surprises.
Sovereign workloads
Federal, state, defence and regulated industries. Air-gapped tenancies, signed runtime, audit-ready evidence.
04 · Performance
Numbers that actually hold up.
Cluster MFU on Llama-3 70B
54.2%
Sustained over 14-day training run, FP8, 512×H200.
Inter-node latency (RDMA)
1.9µs
p50 across 2-rail NDR fabric, same pod.
Provisioning SLA
48h
Reserved capacity, contract signed to first job submission.
Uptime, last 12 months
99.99%
Across SYD-1 and MEL-1, weighted by GPU-hours.
05 · Pricing
Three ways to buy compute.
No egress charges. No surprise multipliers. AUD-billed, GST-inclusive on request.
On-demand
By the hour, by the node.
- From $8.40 / node·hr (8× RTX 5090)
- 1-minute billing granularity
- Slurm or Kubernetes
- Free intra-region ingress / egress
Reserved · Most popular
Locked-in pods, locked-in price.
- 3-, 6-, 12- and 36-month terms
- Dedicated InfiniBand fabric
- Named TAM + Slack channel
- Priority access to B300 capacity
Sovereign
Private cloud, your tenancy.
- Dedicated hardware & fabric
- IRAP-aligned, ISM-mapped
- Cleared personnel on request
- Custom commercial terms
06 · FAQ
Questions, answered plainly.
Hyperscalers give you breadth — every service, everywhere. Navigate Network gives you depth: dedicated GPU pods, no noisy neighbours, per-pod InfiniBand, predictable AUD pricing, and a region you can point to on a map of Australia.
AU: SYD-1 in Sydney's western corridor (Tier III+), MEL-1 in Melbourne, with SYD-2 coming soon and PER-1 (Perth) on the 2027 roadmap. SEA: SIN-1 in Singapore and KUL-1 in Kuala Lumpur, with JKT-1 (Jakarta) coming soon. EMEA: DXB-1 in Dubai and RUH-1 in Riyadh. All facilities are operated by Navigate Network or vetted regional partners under our SLA.
An 8-GPU node, hourly, on a credit card. We are happy to have you start small — but we are built for serious clusters, so call us once you've outgrown that.
Standard SLA is 48 hours from signed contract to first job submission for H200 and RTX PRO 6000 pods. B300 lead times are quoted per deal.
Our Sovereign tier is designed for PROTECTED-suitable deployments, IRAP-aligned and mapped to the ASD ISM. We work with your assessor on certification.
07 · Get a quoteLet's get your
Let's get your
workload onshore.
Tell us what you're training, fine-tuning or serving. A capacity engineer comes back within one business day with a sized proposal — pod composition, region, price, and earliest start date.
- 1 business day response, AU hours
- Sized proposal — not a generic price sheet
- NDA available on request