Providers
A deployment runs wherever its hardware profile's provider says.
Supported
Wired into amicu apply: this machine (local, local_apple, local_cpu, on_prem) and Modal. Modal is the only one tested end to end.
Preview
An adapter exists and is tested against mocks. Not wired into apply.
- HetznerDedicated GPU servers in German and Finnish data centers.Germany, Finland, plus US and APAC
- OVHcloudPublic Cloud GPU instances from a French sovereign cloud.France, Germany, Poland, plus non-EU regions
- ScalewaySingle-GPU instances in French, Dutch and Polish regions.France, Netherlands, Poland
- RunPodProvider-managed GPU pods across EU, EEA and worldwide data centers.EU and EEA data centers, plus US, Canada and APAC
- Self-hosted over SSHHardware you declare and own, reached over SSH.The location you declare for each host
Planned
No adapter yet. Region and GPU catalog data only.
- NebiusHigh-end GPUs in Finland and France, plus a US region.EU regions, plus a US region
- AWSGPU EC2 instances in five EU regions, plus US and APAC.Ireland, Frankfurt, Stockholm, Milan, Paris, plus US and APAC
- AzureGPU virtual machines across Europe, the UK, the US and APAC.West and North Europe, Germany, Sweden, France, plus UK, US and APAC
- GCPGPU Compute Engine instances in EU regions, plus US and APAC.Belgium, Netherlands, Finland, Frankfurt, Paris, plus US and APAC
Moving a deployment is a change to provider. See Providers and Hardware profile in the docs.