GPU Cloud
AvailableThe fleet behind our research, by the hour.
Launch GPU capacity in minutes and pay for the hours you use, or reserve it for a month to a year at a lower rate. The same fleet serves our own inference and training.
- Pricing
- From $3.35 / GPU-hr
- Layer
- Compute
- Launch with
- AMD Instinct MI325X, NVIDIA H200, CUDA images, ROCm images, PyTorch, vLLM, SSH keys, Project canvas
Why GPU Cloud
One console for every GPU you run.
Pick an offer by GPU, region, price and availability, launch it from the console or stage it on your project canvas, and follow utilization, memory and power for each GPU while it runs. The same fleet serves our own inference and training.
On demand
MI325X 256 GB and H200 141 GB at $4.79 per GPU-hour.
Reservations
Commit for 1, 6 or 12 months from $3.35 per GPU-hour.
Live telemetry
Utilization, memory and power for every GPU you run.
How it works
AMD MI325X and NVIDIA H200 by the hour or on reservation.
In the console
Everything in one place.
GPU Cloud lives in the LegionEdge cloud console, next to every other service in the workspace, under one account and one identity.
Pricing
From $3.35 / GPU-hr
On-demand capacity is metered by the GPU-hour, with no instance minimums. Reserve capacity for a month to a year and the hours inside the reservation bill at a lower rate.
- Hard budget caps you set yourself: nothing scales past a limit you didn’t write down.
- The launch wizard checks your budget before an instance starts.
- A reservation’s term fee appears on your statement as a plan item.
GPU rates
Regions
Related products