Download websites that use Azure

About Azure

A hosting label answers a narrower question than it appears to. What we record is the platform that responded to the request, and Azure is not one thing. A static site served from storage, a virtual machine administered by hand, a managed application service, a container platform and a delivery network sitting in front of an origin somewhere else all land under the same name. Reading a row as proof that a company runs on the platform goes a step further than the evidence supports.

What the label does support is a claim about procurement. The platform is very often chosen by organisations already committed to the rest of the vendor's estate through identity, desktop software, database licensing and agreements negotiated centrally. The web team frequently inherits that decision rather than making it. Finding the platform behind a domain is therefore weak evidence about the site's engineering and reasonably strong evidence about the buying centre behind the site, which is the opposite of what a technology signal usually offers.

As of September 2026 we see 695,857 sites resolving to it, 261,658 of them in United States, with .com the domain ending we record most often. Launched by Microsoft in 2010, the platform arrived after the first wave of cloud adoption and grew by pulling existing corporate customers across rather than by winning hobby projects, and the profile of the domains reflects that history. Comparatively more corporate sites, comparatively fewer personal ones, and a long tail of marketing properties belonging to companies whose real workloads are somewhere else entirely.

Category boundaries deserve a word too. Hosting, in a dataset like this one, is a single column describing something that is really a stack of separate services, and a company can appear under one platform for its origin, another for its delivery network and a third for its mail. Reading any one of those columns in isolation produces a tidy picture that the other columns immediately complicate.

Attribution comes from address ranges and hostname patterns, so the caveats are structural rather than incidental. A third party delivery network in front of an origin masks the platform completely, and multi-cloud estates are common enough that rows have to be treated as non-exclusive, since a marketing site and a product can sit on different providers under one brand. Absence of the label is therefore weak evidence, and anyone computing market share from hosting detection alone will understate every platform that is commonly hidden behind a proxy.

Used properly, the file supports questions that no per-domain lookup reaches. How many companies in a country or an industry segment terminate on the platform, which of them also carry the rest of a Microsoft-adjacent stack, which agencies keep deploying there, and which domains changed platform between one snapshot and the next. Partner channel targeting, competitive displacement and infrastructure research all run on the same shape of data. One row per domain, with the detected stack beside it, filtered to a segment and compared over time.

The comparison is the part worth paying for. A single reading of who hosts what is a curiosity, while the difference between two readings is a list of accounts that just moved, and moving accounts are the only ones that were ever going to take a call.