Two Providers Now Handle 38.6% of Popular-Domain Mail: The Deliverability Risk Nobody Is Mapping

Self-hosted mail servers ran 44.6 percent of the internet's top million domains in 2016. In 2026 that figure is 22.4 percent, according to a new analysis of Tranco-ranked domains built on OpenINTEL's daily DNS snapshots. The domains that left self-hosting did not scatter across dozens of providers. Most of them landed in two mailboxes: Google Workspace now sits behind 21.8 percent of detectable MX records, Microsoft 365 behind 16.8 percent, a combined 38.6 percent of popular-domain mail routed through two companies.

What the Data Shows

The research comes from Artem Berezin at Live Direct Marketing, who tracked MX record ownership across the Tranco top million domains using OpenINTEL's forward-DNS snapshots, a dataset the University of Twente, SURFnet, and SIDN Labs have maintained since 2016. The decline in self-hosted mail was not a sudden jump. It moved down by roughly half a percentage point every 30 days for a decade, the kind of trend that is easy to miss year over year and impossible to miss looked at across ten.

Berezin's own framing of the risk is worth quoting directly: "When more than a third of popular domains depend on two providers to receive mail, an outage, a filtering change, or a policy decision at either one propagates through the whole ecosystem at once." He also notes that independent mail operators now face a harder deliverability climate than they did a decade ago, because spam filtering across the industry has effectively been tuned around the behavior of the two largest providers, leaving smaller and self-hosted senders working against a baseline they had no part in setting.

Why This Is an Outbound Problem, Not Just an Infrastructure Trend

None of this changes how SPF, DKIM, or DMARC work. What it changes is where the leverage sits. A cold outbound program does not choose which provider its prospects use, and B2B contact lists skew toward exactly the two providers this data flags: Google Workspace and Microsoft 365 dominate business email far more than they dominate the general domain population, because both companies built their office suites around exactly the small and mid-size business buyer that most outbound programs target.

That means the real exposure for an outbound team is not "we send through one ESP." It is "a third or more of our active prospect list receives mail at an inbox controlled by one of two companies, and we have never measured which third." A Gmail algorithm update or a Microsoft 365 junk-mail threshold change does not need to target cold outreach specifically to gut a campaign's reply rate. It only needs to move, and the accounts sitting behind that provider move with it.

The Map Most Teams Have Never Built

Building that map is enrichment work, not guesswork. MX records are public DNS data: a prospect list can be segmented by receiving provider the same way it gets segmented by industry or title, and that segment can then be tracked against reply rate and bounce rate independently. A campaign that looks healthy in aggregate can be quietly failing against one provider while performing fine against the other, and aggregate reporting will never surface that split on its own.

Run that segmentation once and the share usually surprises the team that pulls it. This study measured the general domain population; a B2B prospect list skewed toward small and mid-size companies is likely to sit well above the 38.6 percent baseline, since those are exactly the buyers Google Workspace and Microsoft 365 built their office suites to win.

My take: most "deliverability monitoring" setups measure the sending side, domain reputation, warmup status, blacklist checks, and stop there. That is necessary and not sufficient. The recipient-side concentration this data describes means the provider on the other end of the send now matters as much as the domain doing the sending, and a monitoring stack that only watches the outbound half is reading half the picture.

Two Controls Worth Building Before an Incident, Not During One

Segment deliverability metrics by receiving provider. Reply rate, bounce rate, and spam-complaint proxies should be tracked per provider bucket, not only in aggregate. Our deliverability monitoring guide covers the mechanics; the addition here is simple: add "receiving provider" as a segment, not just "sending domain."

Write the response plan before the filtering change, not after. If Google or Microsoft adjusts a threshold, the teams that keep sending unchanged and wait for reply rates to explain themselves lose the most time. A short runbook, who checks which dashboard, what the escalation trigger is, whether to pause sends to the affected provider while the domain's standing is confirmed, turns a scramble into a five-minute decision.

Where This Leaves Outbound Teams

The shift from 44.6 percent self-hosted to 22.4 percent did not happen because self-hosting became harder to configure. It happened because two providers became the default, and defaults compound quietly until a data set like this one puts a number on them. Nothing in this changes tomorrow's send schedule. What it should change is whether a team can answer a simple question before the next filtering headline forces the answer on them: if Google or Microsoft moved a threshold tomorrow, do you know how much of your pipeline sits behind that door, or would you find out from a reply-rate chart three weeks late?

Teams that already treat sending infrastructure as something to own and instrument, not something to rent and hope about, are the ones positioned to answer that question today. That is the same ownership standard our outsourcing lead generation guide applies to domains and mailboxes: if you cannot say who controls a piece of your sending infrastructure, you do not control it, and the same logic now extends to knowing exactly which inbox each part of your list depends on.

Book your free discovery call →