Regions & Availability

machine.dev runs on AWS infrastructure across 6 regions. Pick regions to optimize for data locality, compliance, or cost. If you don’t specify, machine.dev picks the cheapest region with available capacity.

Available regions

GPU availability by region

Not every GPU type is available in every region. CPU runners (X64 and ARM64) are available everywhere.

AI accelerators (Trainium, Inferentia2) are currently US-only. Trainium and Inferentia2 run in us-east-1, us-east-2, and us-west-2. Inferentia2 is also available in ap-southeast-2.

AWS regional capacity changes over time — see machine.dev/runners for the live availability and best-rate-by-region view.

Specifying regions

Pass regions= with a comma-separated list. machine.dev will try each region in priority order and use the first one with available capacity:

runs-on:
  - machine
  - gpu=l4
  - regions=us-east-1,us-east-2

How auto-selection works

When you don’t specify regions=, machine.dev picks based on:

1. Availability — whether your requested GPU type has capacity right now
2. Spot price — if tenancy=spot, the cheapest region wins
3. Historical reliability — regions with fewer recent failures are preferred

Specifying multiple regions almost always finishes faster and cheaper than locking to one. Pin to a single region only when you need data residency or compliance.

Compliance and data residency

Need a region we don’t support?

Email hello@machine.dev to request a region. New regions are added based on customer demand.

Next steps

• Configuration options — every label for runs-on
• GPU Runners — GPU types, configurations, prices
• Cost Optimization — using region selection to save money