Edge Compute
Contact Sales Log in Sign up
🍁 Hosted in Canada · Billed in CAD

Canadian GPU cloud,
billed by the minute.

Rent dedicated NVIDIA RTX 5090s on demand. Root SSH and JupyterLab in seconds, an isolated container with your own GPU, and you only pay for the minutes it runs. Your data never leaves Canada.

Get started with $20.00 View available GPUs
0GPUs online
0available right now
$0.79from, per GPU-hour
1 minbilling granularity
Live availability

Real-time GPU inventory

Every card below is a real GPU online in our Canadian network right now. Pick one and you're on it in minutes.

All GPUs are offline for maintenance. Check back shortly or contact us.

Browse all GPUs
How it works

From sign-up to a running GPU in under five minutes

No contracts, no sales calls, no CUDA setup.

1

Add credit

Create an account and top up from $20.00 CAD by card. Unused credit can be refunded.

2

Pick a GPU

Choose an available RTX 5090, paste your SSH public key, and optionally set an auto-stop time as a spending cap.

3

Connect & build

SSH in as root or open JupyterLab in your browser. PyTorch and CUDA are ready. Stop it when you're done.

Get started for $20.00 →
Built for developers

Start. Connect. Ship. Stop.

Every rental is a clean Linux container with one dedicated RTX 5090, ready the moment it starts.

  • ✓
    Root SSHscp, rsync, VS Code Remote, port forwarding — it's your machine.
  • ✓
    JupyterLab over HTTPSA private, token-protected notebook URL for every rental.
  • ✓
    Ready-to-go stackPyTorch 2.7, CUDA 12.8, cuDNN 9, Transformers, Accelerate, Hugging Face Hub.
  • ✓
    Fast internet egresspip, conda, git and Hugging Face downloads work out of the box.
terminal
# connect to your rental
$ ssh -p 20014 root@clearlayer.ca

root@r14:~# nvidia-smi --query-gpu=name,memory.total --format=csv
name, memory.total [MiB]
NVIDIA GeForce RTX 5090, 32607 MiB

root@r14:~# python -c "import torch; print(torch.cuda.get_device_name(0))"
NVIDIA GeForce RTX 5090

root@r14:~# accelerate launch train.py --model Qwen/Qwen2.5-7B
...

# copy results home, then stop from the console
$ rsync -avP -e "ssh -p 20014" root@clearlayer.ca:/workspace/out ./
Products

One network. Pay the way that fits.

Pay by the minute when you need flexibility, run overnight for 25% less, reserve a GPU for a day, week or month — or talk to us about a larger, longer-term setup.

⚡

On-demand GPU Cloud

RTX 5090s from $0.79/hr, billed per minute from prepaid credit. Start and stop any time from the console.

Explore GPU Cloud
🌙

Off-peak, 25% off

Every on-demand minute between midnight–8 AM Pacific costs just $0.59/hr. Queue your training runs overnight and the discount applies automatically.

See off-peak pricing
📅

Reserved GPUs

Lock in a dedicated GPU for a day, a week or a month — up to 30% off. Paid up front, yours the whole time, never interrupted.

Reserve a GPU
🏢

Enterprise

Multiple GPUs, terms longer than a month, Canada-only placement guarantees, an SLA and monthly invoicing. Custom pricing for long-term partners.

Contact sales
Use cases

Built for every AI workload

32 GB of GDDR7 and Blackwell tensor cores handle most single-GPU training and inference jobs.

LLM fine-tuning (LoRA / QLoRA)LLM inference & APIsImage generation Video generationSpeech-to-text transcriptionComputer vision training Batch data processingEmbeddings & RAG pipelinesResearch & experiments CUDA development3D renderingRobotics simulation

32 GB

GDDR7 VRAM per GPU — fits 7–14B models for fine-tuning and ~30B quantised for inference.

Blackwell

5th-gen tensor cores with FP8 / FP4 support. CUDA 12.8+ and the latest drivers.

Dedicated

One whole GPU per rental — never time-sliced or shared with another customer.

Root access

Install anything with apt, pip or conda. Your environment, your rules.

Security & data residency

Your data stays in Canada. Your session stays yours.

Every GPU in our network is hosted in Canada, and every rental is isolated from the host and from other customers.

  • ✓
    Canada-only nodesCompute and storage on Canadian hardware. Billing and contracts in CAD.
  • ✓
    Isolated containersUser-namespace remapping, no privilege escalation, and a firewall that blocks access to the host's network.
  • ✓
    Verified hostsEach host is reviewed before it goes live and re-checked automatically every few hours.
  • ✓
    Wiped after every rentalThe container and its disk are destroyed when you stop — nothing carries over to the next customer.
  • ✓
    Encrypted accessKey-only SSH (no passwords) and HTTPS for JupyterLab.
Pricing

Simple pricing. Bigger savings the longer you run.

No storage, egress or setup fees — you pay for GPU time only.

On-demand
$0.79/hr

Per-minute billing from prepaid credit.

Off-peak · 25% off
$0.59/hr

Automatic, midnight–8 AM Pacific.

Reserve · up to 30% off
$0.55/hr

Prepay a day, a week or a month.

View full pricing
FAQ

Frequently asked questions

Billing, off-peak and reservations: see the pricing FAQ.

Is my data kept after I stop?

No. Stopping deletes the container and its disk so the next customer gets a clean machine. Copy your results out with scp, rsync, git or cloud storage before you stop.

Which software is pre-installed?

It depends on the template. The default PyTorch template has CUDA 12.8, cuDNN 9, PyTorch 2.7, JupyterLab, Transformers, Accelerate and the Hugging Face CLI. You have root, so you can install anything else.

Can I run Docker inside my rental?

Not on GPU Cloud — for security, rentals are unprivileged containers. If you need your own images or a full VM, talk to us.

Where exactly are the GPUs?

All nodes are located in Canada. Each GPU listing shows its region. Canada-only placement is guaranteed for every rental.

Can I use the platform from code?

Yes — see Developers for the CLI, Python SDK and REST API.

Can I host my own GPUs?

Yes, if they're in Canada. Read the host setup guide and apply.

Start with $20.00. Run on Canadian GPUs today.

No contracts. Per-minute billing. Stop whenever you like.

Get started Contact sales