Regions & Availability
Machine provides GPU runners in multiple AWS regions around the world, ensuring low-latency access, data sovereignty compliance, and improved availability for your workflows.
Available Regions
Machine beta program currently supports the following AWS regions:
North America
- us-east-1 (N. Virginia)
- us-east-2 (Ohio)
- us-west-2 (Oregon)
- ca-central-1 (Central)
Europe
- eu-south-2 (Spain)
Asia Pacific
- ap-southeast-2 (Sydney)
We plan to expand to many more regions before launching the full platform.
GPU Availability by Region
Not all GPU types are available in every region. The table below shows current availability in our beta program:
| GPU Type | North America | Europe | Asia Pacific |
|---|---|---|---|
| T4G (16GB) | us-east-1, us-west-2 | eu-south-2 | Not available |
| T4 (16GB) | us-east-1, us-east-2, us-west-2, ca-central-1 | eu-south-2 | ap-southeast-2 |
| L4 (24GB) | us-east-1, us-east-2, us-west-2, ca-central-1 | eu-south-2 | ap-southeast-2 |
| A10G (24GB) | us-east-1, us-east-2, us-west-2, ca-central-1 | Not available | ap-southeast-2 |
| L40S (48GB) | us-east-1, us-east-2, us-west-2 | eu-south-2 | Not available |
| TRN1 (Inferentia) | us-east-1, us-east-2, us-west-2 | Not available | Not available |
| INF2 (Inferentia2) | us-east-1, us-east-2, us-west-2 | Not available | ap-southeast-2 |
Note: Availability may change based on AWS capacity and Machine expansion plans
Specifying Regions in Workflows
You can specify one or more regions in your GitHub Actions workflow:
jobs: ml-job: runs-on: - machine - gpu=a10g - regions=us-east-1,us-east-2Benefits of Multiple Regions
Specifying multiple regions provides several advantages:
Improved Availability
If capacity is limited in one region, your job can run in another
Cost Optimization
Machine will select the most cost-effective region with available capacity
Failover Protection
If there’s an outage in one region, your job can still run
Single Region Configuration
If you need to ensure your job runs in a specific region (e.g., for data residency requirements):
jobs: ml-job: runs-on: - machine - gpu=t4 - regions=eu-south-2 # Only run in SpainAutomatic Region Selection
If you don’t specify any regions, Machine will automatically select the optimal region based on:
Current Availability
Current availability of your requested GPU type
Spot Pricing
Spot pricing (if spot tenancy is selected)
Historical Performance
Historical performance of the regions
This automatic selection helps ensure your jobs start promptly and at the lowest cost.
Region-Specific Considerations
Data Transfer
When working with large datasets, consider that data transfer between regions may incur additional costs and latency. Options to optimize:
Use GitHub Actions caching
Cache dependencies and intermediate data
Use region-specific artifact storage
Store large datasets in the same region as your runners
Consider AWS S3 transfer acceleration
For cross-region data movement
Compliance and Data Sovereignty
Different regions are subject to different regulatory frameworks:
United States Regions
Subject to US regulations
Asia Pacific Regions
Various country-specific regulations apply
If your organization has specific compliance requirements, select regions accordingly.
Pricing Variation by Region
GPU pricing can vary by region. Generally:
US and EU regions
Tend to have similar pricing
Asia Pacific regions
May be 5-15% more expensive
The Machine platform automatically factors in these price variations when optimizing for cost.
Best Practices
Specify Multiple Adjacent Regions
For example, us-east-1,us-east-2 for optimal availability
Match Regions to Data Location
Run workflows in regions where your data resides
Consider Time Sensitivity
For urgent jobs, include more regions to improve availability
Consider Compliance Requirements
Choose regions that meet your regulatory needs
Coming Soon
Machine is continuously expanding regional availability. Upcoming additions:
Seoul
ap-northeast-2
Stockholm
eu-north-1
Ireland
eu-west-1
Mumbai
ap-south-1
London
eu-west-2
Singapore
ap-southeast-1