Service — Cloud Hosting
On-demand GPU cloud, billed by the hour
Start with one GPU and add more while your job runs. Instances come with the drivers, CUDA and orchestration tools already installed, and you pay only for the hours you use.
What you are buying
- GPU instances available by the hour from $5.40/GPU/hr
- Pre-installed drivers, CUDA, PyTorch and JAX images
- Kubernetes and Slurm scheduling for batch and container workloads
- Shared NVMe file storage next to the compute
- Console access and hourly usage statements
What is not included
- Guaranteed capacity on on-demand instances — availability is confirmed when the instance starts
- Reserved capacity and custom SLAs, which are separate orders
- Managed application tuning for your own models
- Ownership of the facility: power and cooling are provided by Equinix
Billing: per GPU-hourMinimum commitment: noneStop and restart at any time
Price
From $5.40/GPU/hr
On-demand rates are per GPU-hour and identical across every page of this site.
Quotes are prepared by our team, not generated automatically. We reply within one business day.