GPU Machines
Full VMs with root SSH. One or more GPUs per machine.
GPU MachinesFull GPU machines with root access, from RTX 4090 and RTX 5090 to A100 and A800. Book hours, days or months up front, pay from your balance, and get unused whole hours back when you release early.
No contracts and no surprise charges. You pay for the hours you pick, and get the unused ones back.
How pricing worksChoose a GPU model, GPU count, image and how many hours you need.
You get a quote that holds for two minutes. Paying takes it from your prepaid balance.
The machine is usually ready to SSH in about a minute, with your keys installed.
Done early? Release it and unused whole hours go back to your balance.
Launch, restart and release machines, manage SSH keys and see every charge in one place.
See GPU machinesFull VMs with root SSH. One or more GPUs per machine.
GPU MachinesPrepaid by the hour, day or month. Unused hours refunded on release.
PricingEvery charge and refund as its own line.
Usage LedgerEvery machine is a full VM you SSH into as root. Install what you like, run Docker, keep your work on the data disk.
{ "status": "example", "user": "root", "gpu": "NVIDIA A800 80GB", "image": "Ubuntu 22.04 · CUDA 12.4", "data_disk": "/data", "vram_gb": 80 }
Each rental is its own virtual machine, cleaned before you get it and deleted when you release it.
How machines are isolatedA full virtual machine per rental. Nothing shared with other renters.
Isolation modelFresh SSH host keys, no leftover accounts, logs or shell history.
What we cleanOnly the public keys you add to your account can log in.
SSH keysUsually about a minute from paying to SSH. The machine is prepared before it is handed over: your keys installed, desktop services off, data disk mounted.
Models include RTX 4090, RTX 5090, A40, A100, A800 and V100. Availability changes, so the console shows the live list and price before you pay.
Prepaid by the hour, day or month, taken from your balance when you launch; longer bookings get a lower rate. Release early and unused whole hours of compute and data disk are refunded automatically.
Root SSH on Ubuntu 22.04 with CUDA 12.4 or Ubuntu 24.04 with CUDA 13.0, the NVIDIA driver, Docker, Python, git and common tools. An optional data disk is mounted at /data.
The data disk is kept across restarts of that machine. When you release the machine, the machine and its disk are deleted, unless you chose to keep the disks so you can start it again later.