A data center advertising "356 networks" and one advertising "3" are not selling the same thing. The first is a place you connect to others from; the second is a place you host in. This page measures that difference across our whole catalogue, publishes the scale, and lets you check it.
For each band of networks present, we counted the share of sites hosting at least one internet exchange point. The progression is continuous, with no exception:
| Rank | Networks present | Sites | with an exchange point |
|---|---|---|---|
| Interconnection hub | ≥ 100 | 92 | 100 % |
| Major hub | ≥ 30 | 279 | 98 % |
| Regional hub | ≥ 10 | 660 | 83 % |
| Interconnected site | ≥ 3 | 1 520 | 43 % |
| Isolated presence | ≥ 1 | 1 651 | 16 % |
| No network presence | 0 | 2 356 | 4 % |
Measured across the 6 558 data centers in the catalogue, recomputed every time this page is served. Scale version 2026.08-v1.
They are read from the table, not chosen. Above 30 networks an exchange point is near-certain; above 100 it is systematic in every measured case. The low threshold of 3 networks comes from industry practice: a site needs at least three distinct autonomous systems for an exchange point to make sense.
It is not a quality score. A low-rank site can be excellent for hosting a server that needs no interconnection, and it will often cost less. The rank measures interconnection density, nothing else.
Network counts come from the public interconnection registry: a declared total, not a list we can verify line by line. Where the list of autonomous systems exists we show it and the rank reads "backed"; otherwise it reads "declared". A rank inherits the reliability of its source, and says so.