SYNTH / Pricing

Transparent Pricing. No Surprises.

Clear per-GPU pricing for on-demand instances, straightforward quotes for rack-scale systems. Enterprise discounts for reserved capacity.

01 · GPU Instances

On-Demand GPU Instances

GPU GPU Memory Bandwidth Architecture Price
H100 80GB HBM3 3.35 TB/s Hopper From $2.49/hr
H200 141GB HBM3e 4.8 TB/s Hopper From $3.49/hr

Prices are per GPU per hour. Instances are bare-metal (no hypervisor overhead). Final pricing confirmed at quote.

02 · Rack-Scale Systems

NVL72 Rack-Scale Systems

System GPUs Architecture Price
GB200 NVL72 72 × Blackwell Blackwell Enterprise Quote
GB300 NVL72 Prep 72 × Blackwell Ultra Blackwell Ultra Coming Soon
Rubin Next-Gen 72 × Rubin Vera Rubin Register Interest

Rack-scale systems are delivered as full NVL72 racks or 8-GPU node slices, with pricing based on configuration, commitment term, and availability.

03 · Reserved Capacity

Commitment Discounts

CommitmentBenefit
1-Year Commitment Up to 15% cost savings
3-Year Commitment Up to 30% cost savings
Enterprise Strategic Capacity Guaranteed GPU allocation with custom pricing
04 · Included

What's Included

Bare-metal performance (no hypervisor overhead)
Pre-configured NVIDIA drivers, CUDA, PyTorch, and TensorFlow
99.9% uptime SLA with N+1 redundant power and cooling
24/7 infrastructure monitoring
Local technical support during Thailand business hours
Support for Kubernetes and Slurm orchestration
Network connectivity as specified in configuration
Transparent, predictable billing
05 · FAQ

Frequently Asked

What is the minimum commitment?

On-demand GPU instances (H100/H200): hourly billing with a 1-hour minimum. Rack-scale systems (GB200): monthly or annual terms.

Do you offer proof-of-concept environments?

Yes — we can provision a short-term evaluation environment for qualified opportunities. Contact our team to discuss.

Which frameworks come pre-installed?

NVIDIA drivers, CUDA, PyTorch, and TensorFlow by default. Additional frameworks on request.

Where is the infrastructure located?

Infrastructure is delivered through data-centre facilities in Thailand. Specific details are provided during procurement.

How long does deployment take?

On-demand GPU instances: typically within 48 hours. Rack-scale systems and dedicated clusters: 2–6 weeks depending on configuration.

What is the uptime guarantee?

We provide a 99.9% uptime SLA, backed by N+1 redundant power and cooling and 24/7 monitoring.

Need Reserved Capacity or Strategic Allocation?

Tell us your commitment horizon and workload profile — we'll put together reserved-capacity pricing.

Request Enterprise Pricing