How decentralized is Nostr?
Every relay we can see, measured. All numbers computed live from our relay data, 2026-11-09 12:04 UTC.
What the map shows
The world map looks global. The server locations are not. 1234 of 1234 online relays have a known location (100%), and their hosting concentrates in a handful of metros: the same cloud regions everyone else uses. Toronto alone (352 relays) carries more Nostr capacity than most countries. One operator runs 10 relays across 1.
A Gini coefficient of 0 would mean relays spread evenly across every city with relays; 1.00 would mean everything in one place. Nostr's relay network sits at 0.73 across cities and 0.83 across countries: more spread than a single company, far less spread than the map suggests.
There is also a graveyard: 4278 relay addresses have been discovered and never once answered. Anyone can announce a relay. Keeping one running is the hard part.
Top cities by online relays
| City | Online relays | Share | |
|---|---|---|---|
| Toronto | 352 | 29% | |
| Helsinki | 59 | 5% | |
| New York | 57 | 5% | |
| Frankfurt am Main | 55 | 4% | |
| Nuremberg | 39 | 3% | |
| Falkenstein | 30 | 2% | |
| Amsterdam | 28 | 2% | |
| Ashburn | 23 | 2% | |
| Clifton | 21 | 2% | |
| Dublin | 21 | 2% |
How to read this
- Population: relays online at their last check, seen within 24 h, one relay per origin (path duplicates collapsed).
- Locations are city-level GeoIP of the server IP, rounded to about 1 km. They say where the server is hosted, not who runs it.
- Relays without a resolvable location are excluded from city and country numbers (0 online relays have no known location).
- GeoIP shows the hosting city. Anycast networks, CDNs and residential relays blur this further; we do not collect ASN data yet.
- Concentration is not a verdict: it reflects where cheap, reliable hosting exists. It does mean the network has single points of geography.
Raw data: relays.json · network stats · recent changes. Numbers update hourly with the snapshot.