Companies Using Debian

The number is a floor, and that is the interesting part

Most counts on this site approximate a population we can see all of. This one does not. Debian is identified from the web server's own response header, which means we count Debian machines that introduce themselves and miss the ones that do not. The consequence shows up in the stack: 98.66% of these 130,134 domains also run Apache, because Debian's packaged Apache is the thing that prints the distribution name. A Debian box serving through nginx says nothing about its distribution by default. A Debian box behind a proxy that rewrites the response header says nothing either.

So 130,134 is a floor, and I would rather publish a floor and say so than publish a round estimate nobody can check. If you want the mechanics, we wrote up the detection methods behind it. The short version: every number on this page describes a Debian server that speaks, and the population of Debian servers that stay quiet is unknown and certainly not zero.

A European base with nobody selling it

Germany accounts for 17,850 of the 77,750 companies we could locate, 22.96%, and France 11,843, 15.23%. The United States is third at 9,308, 11.97%. Germany and France together hold 29,693 against the American 9,308, and twelve of the top fifteen countries are European, 49,665 companies and 63.88% of the located base.

The United States leads the country table on most profiles we publish, so an inversion this size gets my attention. It is not unique. IONOS and OVH invert it too. But those are companies with German and French sales teams selling German and French hosting. Debian has no sales team, no regional pricing and no channel. Nothing about the product steers it toward Frankfurt or Paris. What this chart actually reads is the shape of the European hosting market, where a lot of small providers built their fleets on Debian and their customers inherited it without ever choosing it.

Debian has no sales motion at all, so its country chart is not a map of where it sold. It is a map of whose servers happened to be built on it.

Small companies, old institutions, and nothing much in the middle

Two distributions on this page point in opposite directions and both are right. On size, 65.25% of the 64,938 companies that report a headcount employ ten people or fewer, 21.3% employ 11 to 50, and only 1,059 companies, 1.63%, employ more than 1,000. On age, the 2010s lead at 33.05%, which is ordinary, but 19.5% of this base predates 1990 and 6.76% predates 1960, which is not.

The named list explains the second half. Fifteen of the 34 organisations on it are educational institutions and five more are government agencies, so 20 of 34 are academic or public sector. Ohio State University's library. Karlsruhe Institute of Technology's intranet. The European Space Agency's historical archives. RWTH Aachen, Université Libre de Bruxelles, Queen's University, and Københavns Universitet, founded in 1479. These are organisations that stand a server up once and keep it running through several fashion cycles, and Debian's slow, predictable release rhythm is built for exactly that temperament.

What is thin is the middle. Very small firms appear because their host chose Debian for them. Large institutions appear because they chose it deliberately and stayed. The mid-market company that revisits its platform every three years is barely here, and I would expect that band to keep drifting toward managed platforms where the operating system is somebody else's problem.

Why this page has no migration table

We publish a switching corridor only when at most 40% of the domains that ever carried both signals still carry both. Debian clears that on nothing. The largest pair, Debian and Ubuntu, sits at 51.5% still-both and runs both ways at once: 530 domains dropped the Ubuntu signal while still showing Debian, and 364 did the reverse, 1.45 to 1. Debian and UNIX is 47.7%. Debian and CentOS is 55.9%.

A near-symmetric flow with half the population still showing both signals is not a migration. It is one organisation running Debian on some machines and something else on others, which is the normal state of any estate with more than a handful of servers. Publishing a corridor out of that would be inventing a story from an org chart.

Half of the domains that ever showed both Debian and Ubuntu still show both. That is not a switch in progress. That is a server room.

The movement we can report honestly is turnover. 11,607 domains stopped announcing Debian since mid-March, 5,651 of them inside the last 30 days. Some of those machines were retired. Some were rebuilt on something else. Some simply stopped printing the token after a configuration change, and from the outside all three look the same.

What the stack says about these sites

Strip out Apache, which is an artefact of how the detection works, and the rest reads like a snapshot of the web around 2014. jQuery is on 76.11% of these domains, jQuery Migrate on 38.58%, jQuery UI on 30.99%. jQuery Migrate exists to stop code written for an older jQuery from breaking, so better than a third of these sites are carrying a compatibility shim rather than current code. WordPress reaches 41.42% and Google Analytics 44.57%.

The row I find most useful is Varnish at 8.32%. Nearly everything else in this stack is what a host installs by default. Varnish is what somebody installs on purpose. Roughly one domain in twelve here has an operator who went and tuned the caching layer, and that one in twelve is a different customer from the other eleven.

What to take from this

Debian is infrastructure that outlived several fashion cycles. It ranks 3rd of the 16 technologies in our Operating Systems category on 15.98% share, behind Ubuntu, which is built on it. The visible base is European, small, old and institutional, and a large share of the sites carrying it run front-end code a decade behind. If you sell into this base, the segment worth separating is the 20 in 34 that are universities and government bodies, because they buy slowly and leave even more slowly. The number to argue with is 130,134, and the honest thing to say about it is that the real figure is higher and we cannot see by how much. TechnologyChecker.io resolves it against 77,750 identified organisations and refreshes it every month.