Live GPU capacity · billed by the second

GPU cloud, on demand.

Rent RTX 4090, A100 and H100 servers by the hour. Spin up a container in seconds, connect over SSH or JupyterLab, and pay only for the runtime you use.

$20 in free credit when you sign up · no card required
8 available
NVIDIA RTX 4090
24GB VRAM · 16 vCPU · 64GB RAM
$0.69/hr
4 available
NVIDIA A100 40GB
40GB VRAM · 24 vCPU · 128GB RAM
$1.49/hr
2 available
NVIDIA A100 80GB
80GB VRAM · 32 vCPU · 256GB RAM
$1.99/hr
2 available
NVIDIA H100 80GB
80GB VRAM · 32 vCPU · 256GB RAM
$2.99/hr

Everything you need to ship

A full GPU platform — scheduler, metering, and instant access built in.

Launch in seconds

Our scheduler places your instance on the best available node and boots a ready-to-use container.

SSH & JupyterLab

Every instance ships with root SSH access and a one-click JupyterLab link. Docker runtime included.

Per-second metering

A real billing engine meters runtime continuously and deducts from your wallet — never overpay.

Auto-stop protection

Instances stop automatically when your balance runs out, so you never go negative.

Multi-node scheduling

Add GPU servers that self-register via API. Load is balanced across the whole fleet.

4090 → H100

From cost-efficient RTX 4090s to H100 SXM for the largest training runs.

From zero to GPU in three steps

  • 1
    Add credit to your wallet
    Start with $20 free. Top up any time from the billing dashboard.
  • 2
    Pick a GPU and launch
    Choose 4090, A100 or H100. The scheduler assigns a node and provisions your container.
  • 3
    Connect and train
    Copy your SSH command or open JupyterLab. Billing runs only while the instance is up.
$ ssh root@103.39.67.73 -p 41022
Welcome to your GPU Cloud instance 🚀

root@h100:~# nvidia-smi
+--------------------------------------+
|  NVIDIA H100 80GB   Driver 550.x     |
|  GPU-Util  0%   Mem  0 / 80GB        |
+--------------------------------------+

root@h100:~# python train.py
epoch 1/10  loss 1.84  ✓

Ready to train faster?

Create an account and launch your first GPU instance in under a minute.