Compute, without the guesswork

Find the GPU for your next breakthrough.

Health-checked GPUs, transparent pricing, and billing by the second. Launch your workloads with confidence.

Explore available GPUs How billing works

Health-checked hardware
SSH access
Transparent billing

H100 from: $0.00; per GPU-hour
GPUs free now: 0; 15 GPU types
Machines ready: 0; updated 02:27 UTC

free GPU rentedone rack per GPU type, one row per machine · tap a rack

I need

1 × H200 NVL$4.05/hr$97.20 a day · cheapest machine

Pick a GPU.

NVLink Free now

52 of 91 machines · 15 GPU types · prices for 1 GPU

- 1 × H200 NVL
  Montreal, Quebec, CA
  1 of 2 free
  $4.05/hr$97.20 a day · $4.05 per GPU-hour
  GPU memory
  140 GB across 1 GPU
  CPU
  128 cores · AMD EPYC 9334 32-Core
  RAM
  126 GB RAM
  Disk
  893 GB disk
  Network
  929 Mbps down · 934 Mbps up
  Interconnect
  PCIe · CUDA up to 13.1
  98.8% reliableUp 9 days
  Sign up to rent

Billing

Prepaid. By the second. Stops at zero.

Prices are per GPU, per hour. A machine with 1 GPU costs 1 × the rate.
The meter runs from the moment the machine is reachable until you stop it; a stopped machine costs nothing.
Credit first; you need at least one hour at the machine's price to launch. Warned at 24 h, 6 h and 1 h left. At zero, the machine stops.
No charge for storage, network or anything else. The price is locked when you launch.

Questions about billing

Estimate

GPUsRuntime

For this run: $97.20
Per hour: $4.05/hr
Per day · per month (730 h): $97.20 · $2,956.50

1 × $4.05/hr per GPU on the cheapest H200 NVL machine right now; the price is locked when you launch.
The rate covers your share of the machine: GPUs, CPU, RAM, disk and network. Nothing else is charged.
Billed by the second against prepaid credit, from the moment the machine is reachable until you stop it. In USD.

1 × H200 NVL, cheapest machine, live

$0.0000at $4.05/hr

Ready0 s

Every machine

What comes with every rental.

Your own container
The GPUs you rented, nothing else.
Your SSH key
Log in the way you already work.
Checked before listing
Every machine passes a test, then health checks.
Watched always
An unhealthy machine is pulled before it costs you.
One GPU or eight
Try on one, train on the whole machine.

Questions

Questions from AI teams

Something else? support@rackloom.in

How do I start?

Create an account, add an SSH key, add credit, and launch a machine from this page. The SSH command is on the machine's page as soon as it is running.

How does billing work?

Prepaid credit, charged by the second at the price shown, from the moment the machine is reachable until you stop it. Nothing else is charged.

Can I overspend?

No. We warn you at about 24 hours, 6 hours and 1 hour of runtime left. At zero the machine stops, so you never owe more than you added.

Which GPUs are there?

What is listed above is what can be launched right now, with the price and the machine's specifications. The list refreshes every minute.

How do I connect?

Over SSH, with a key you add to your account. Ports you expose are listed on the machine's page.

Need something else?

Tell us what you need.

A GPU that is not listed, a bigger cluster, a reserved block for a training run: tell us and we will email you when we can offer it.

Have idle GPUs instead? Earn from every hour they are listed

Rackloom Support

Find the GPU for your next breakthrough.

Pick a GPU.

Prepaid. By the second. Stops at zero.

What comes with every rental.

Your own container

Your SSH key

Checked before listing

Watched always

One GPU or eight

Questions from AI teams

Tell us what you need.

Pay in crypto