Architecture / Regions
Regions
A region is a facility with at least one cluster. There is one live region and one planned. We list a region when its first cluster passes soak testing, not when the lease is signed.
us-west
Hillsboro, Oregon. One cluster, M1, live since 2026-08-19. The facility is a Tier III colocation site with A and B power feeds and rear door cooling. Our cluster sits on a dedicated 15 kW circuit.
Median round trip from three cities to M1, measured over the last 30 days.
| From | Median round trip |
|---|---|
| San Francisco | 14 ms |
| New York | 68 ms |
| Frankfurt | 146 ms |
Egress leaves through two transit providers. Bandwidth to the internet is 10 Gb/s shared, with a 1 TB per month allowance per team included and 0.02 USD per GB after.
eu-central
Frankfurt. Next. The cluster uses the same bill of materials as M1 and will be listed as M2 once it passes 72 hours of soak testing. Teams that need data to stay in the EU can pre register interest through the request access form by choosing eu-central as the region in the notes.
We do not publish a date for eu-central. We publish the state of the commissioning steps in the monthly capacity report.
Choosing a region
In V1 there is nothing to choose. Every job runs on M1. When eu-central is live, the cluster field in the job request selects the cluster, and each cluster has its own queue and its own reservations. There is no cross region scheduling and no data replication between regions. Shared storage is per region.
What is not planned
We are not planning a dozen regions. Two clusters in two regions, each sold out, is a better business than twenty half empty ones. A third region gets a lease when the second is sold out.
TELEMETRY ON THIS SITE IS LIVE DATA.
X — @melaps_system · GitHub — MelapsSystem
CONTRACT $MELAPS — 0x49b5a9fcd24ca5083fa05f9136e7d04564647621
