raw NVIDIA GPUs

GPU cloud
built for the
command line

Rent a dedicated NVIDIA GPU by the second. Deploy in minutes, connect over SSH or Jupyter, and stop billing with one click. No hardware, no contracts.

deploy a GPUsee GPUs & prices
  • ready in minutes
  • from $1.00/hr
  • billed per second
root@cloudrig — ssh
$deploy: RTX 4090
provisioning instance i-7f3a… [ok]
$ssh root@i-7f3a.cloudrig.space
Welcome to Ubuntu 24.04 LTS · CUDA 12.4
$python train.py — epochs 50
epoch 50/50 · loss 0.014 · 312 img/s
$stop from the dashboard
instance stopped · billing halted · disk kept
$
01

select loadout

// dedicated NVIDIA tiers — pick by VRAM

STARTER

RTX A4000 / RTX 4060 Ti

vram16 GB
vcpus4
ram16 GB
$1.00/hr
deploy →
STANDARD

RTX 4090

vram24 GB
vcpus8
ram32 GB
$1.50/hr
deploy →
PRO

L40S

vram48 GB
vcpus8
ram32 GB
$2.25/hr
deploy →
POWER

A100 80GB

vram80 GB
vcpus16
ram64 GB
$3.00/hr
deploy →

// prices are the rate you are billed — consistent across pricing, products & deploy

02

the loop

// spin up → connect → run → stop

01SPIN UP
$ deploy: RTX 4090

Pick a GPU, hit deploy. A dedicated NVIDIA box boots in a few minutes.

02CONNECT
$ ssh root@i-7f3a.cloudrig.space

Full root over SSH, or Jupyter in the browser. Your stack, your rules — install anything.

03RUN
$ python train.py — epochs 50

Train, render, build. Pay only for the seconds the machine is running.

04STOP
$ stop from the dashboard

One click and the meter stops. Your disk is kept for 24h.

03

workloads

// what people run on cloudrig

[ai]

AI / ML training

PyTorch, CUDA, 24–48GB VRAM. Train models without renting a cluster.

read more →[render]

GPU rendering

Blender Cycles from the command line on a dedicated NVIDIA GPU.

read more →[dev]

Remote dev box

A disposable Linux box with root, reachable over SSH from any device.

read more →
04

cloudrig vs own hardware

// the diff

parameter
cloudrig
own hardware
setup_time
a few minutes
days to weeks
upfront_cost
$0
$1,500 – $5,000+
gpu_upgrade
redeploy on any tier
buy new hardware
maintenance
0 — fully managed
DIY drivers + repairs
scale
spin up N boxes
single machine
access
any device, anywhere
desk-bound
05

capabilities

// what every instance ships with

fast_deploy
minutes to boot, every tier
root_access
full admin, install anything
disk_kept_24h
files kept 24h after stop
per_second_billing
pay for active compute only
ssh_jupyter
connect from any device
preinstalled_cuda
drivers + CUDA out of the box
no_contracts
no commitment, stop any time
one_click_stop
stop billing with one click
06

man cloudrig

// frequently asked

Most instances deploy in a few minutes. Pick a tier, deploy, ssh in.

$ deploy now

Your GPU box is minutes away

No credit card to browse. No contracts. Pure compute on demand — billed by the second, stopped with one click.

$ deploy your workstation
from $1/hr · no commitment