One company runs the nameservers, website and mail for one corporate domain in five

A domain that is reachable at all relies on three commercial layers. Nameservers answer for the name; if they stop, nothing on the domain resolves. The website sits behind a content-delivery edge or directly on a hosting network. The mail pathtakes the domain’s email. Each is a dependency the owner cannot swap quickly. The question that matters is how often they belong to the same company.

The obvious candidate for one operator in all three layers is Cloudflare, which sells nameservers, the edge and mail routing. That triple exists and is tightly correlated: of corporate domains that route mail through Cloudflare, 84% also use it for nameservers and edge. But it covers 1.2% of corporate domains (1,402,264).

The larger pattern is older and less discussed. For one corporate domain in five (22,975,657), the same registrar or host answers the nameservers, serves the website and takes the mail. Three named dependencies, one company, one control plane.

The hosting layer is concentrated on its own, too. In the Datazag Observatory, 7 networks carry half of all domains that sit on a network, and 227 carry nine in ten (measured 1 October 2026).

A portfolio that models Cloudflare as the correlated dependency, and the domain’s host as an administrative detail, has the scale of the two exposures the wrong way round.

The one-company stack is not spread evenly. It is a European, hosting-led pattern, and in the largest markets it is a dominant way a corporate domain is run.

For a German book, this is the largest correlated exposure in the dependency picture. Two in five corporate domains get nameservers, website and mailbox from one company. A control-plane event there is not three separate risks that happen to coincide; it is one event with three consequences.

The worldwide mailbox platforms barely appear in this pattern. They run a great deal of corporate mail, but almost none of the nameservers or websites behind it. The domains most exposed to a single company are held by the hosts that sell the domain, the site and the mailbox as one product.

17.9% of web domains sit behind a content-delivery edge (49,029,025 of 274,472,709). Among them the market is not concentrated in the ordinary sense. It is, for practical purposes, one vendor.

Cloudflare holds more than 85% of the fronted segment in every market over 50,000 websites except Japan and Korea, where it still leads. What changes by territory is how much of a market is fronted at all: 42% of Indonesian websites against 5.5% of German ones.

Where a domain owner does choose Cloudflare directly, the layers fuse: 83% of websites whose nameservers Cloudflare runs are also served through its edge.

The DNS protocol was built for redundancy, and organizations can buy it from two providers. The population almost never does.

98.86% of operating domains rely on a single nameserver operator. Of the 1.14% that list two, two-thirds come from one website platform that runs its own nameservers across two providers: the owner inherits that redundancy rather than choosing it. Every enterprise that deliberately pairs providers is, at most, 0.38% of the population.

That shapes how a nameserver event belongs in a model. For nearly every domain, the operator answers or the name does not resolve, and with it the website, the mail and every other service on the domain. This is not hypothetical: in October 2016, an attack on the DNS provider Dyn made many major websites unreachable at the same time, across many unrelated companies.

In several national markets, one operator answers for between a third and a half of all operating domains.

A Canadian or Indian portfolio carries around half of a single registrar’s nameserver event, through default settings rather than purchasing decisions. And in Brazil, the leading nameserver operator is the national registry itself, whose DNS answers for a quarter of the market: a sovereign dependency no vendor list would show.

Every figure comes from records the domains themselves publish. Nothing is surveyed, self-reported or inferred from a vendor’s customer list.

A domain’s nameserver records name who answers for it. Its address record shows which network its website sits in, and for the major edge networks that address falls inside ranges the vendors publish themselves. Its mail records name who accepts its mail. Nameserver hosts are attributed to operators through a crosswalk curated in-house, which merges one operator’s nameservers when they span several domain names. Parked domains and domains held for sale are removed first. The corporate cut is the domains that also publish a working mail server and an SPF record: the ones a business runs.

Datazag tracks 408,022,905 domains. 367,238,077 resolved at their latest observation; the other 40,784,828 returned NXDOMAIN, SERVFAIL or a timeout. DNS and edge were observed from 4 May 2026 to 10 September 2026, one observation per domain (the latest).

• Domains, not organizations. One organization may hold hundreds of defensive registrations behind one operator: one claim, not hundreds.
• The edge is read from the website’s own address. That sees the largest networks almost completely, but misses some enterprise deployments that front only a www alias. Shares for the smaller edge vendors are floors.
• Registrar-default nameservers count as a dependency, because they are one.
• A small share of nameservers is not yet named. 13.6% of operating domains touch an operator identified only by its domain name. The largest is 0.7% of the population, so nothing of accumulation size is hiding.
• Country-code domains stand in for territory; .com does not. Confirm territorial figures against a book’s own domicile data.
• No change over time. This is one observation per domain. Movement between operators is not claimed, because the history that would support it is not yet captured at the grain it needs.