Instant GPUs. Ready for AI
North American Infrastructure. Direct Platform Access. Rent on-demand GPUs in seconds, or contact us for custom solutions and reserved capacity.
Pre-configured Ubuntu VM templates with PyTorch, ComfyUI, and Stable Diffusion Web UI.
Quick Compare
GPU Count
Specs
CPU Model
EPYC 9654
CPU Core Count
24
Total vRAM
32 GB
System RAM
96 GB
Estimated hourly
$0.50/hr
Based on RTX 5090 • 1 GPU • On-demand
Dedicated resources
Ubuntu VM support
Reserved Public IPs
5 Gbps public uplink
One-click rentals
Per-second billing
Live Offers Updated every 2 minutes
Reservation:
North American hostingLow CostSecure CloudUbuntu VM support
Live GPU listings from our platform — click any card to rent instantly through our console.
Dynamic costs (Storage – $0.20/GB/month, Bandwidth – $2.66/TB) are not included in the above pricing.
Custom solutions
Hundreds of GPUs available off-market (5090s, 4090s, RTX PRO 6000s). Contact us to discuss custom solutions for your specific needs.
Reserved nodes and SLAs
8x5090 servers, monthly terms, priority capacity, incident SLA with service credits.
Multi-node clusters and interconnects
Dedicated low-latency networking between VMs for distributed training, with custom VLAN configurations and high-bandwidth interconnects.
Extended storage solutions
High-capacity NVMe arrays (16TB+), network-attached storage for datasets, and custom backup/snapshot configurations.
White-glove onboarding
Custom environment setup, framework pre-installation (PyTorch, TensorFlow, etc.), and dedicated technical support for migration from other platforms.
Why Nova Cloud
North American hosting
Low latency routes across North America for training and inference close to your users.
High efficiency
Average PUE near 1.05, helped by Canada climate.
Off-market capacity
Hundreds of GPUs off market (5090s, 4090s, RTX PRO 6000s).
Per-second billing
Pay only for what you use, down to the second. Transparent pricing with clear reserved discounts.
Direct platform access
Full control with native API and CLI. Dedicated public IP per VM with unrestricted port access.