If you asked a marketer in 2016 to name the biggest ESP on the planet, the answer would have been a marketing suite — something with a campaign builder, templates and a pricing page aimed at humans. In 2026, the answer is a cloud primitive with an API and a per-message price. As of the 2026-07-05 snapshot of our daily DNS scan, Amazon SES appears in 6.21% of SPF records across the Tranco top-1M domains — the highest share of any email service provider we track. It stood at 1.51% in 2016 and 2.86% in 2022, meaning SES has grown 2.2x in just four years.
The data behind that claim comes from OpenINTEL's daily DNS measurements: 192 snapshots spanning January 2016 to July 2026, covering 623,370 domains that publish SPF. Because a domain declares its senders in its SPF record, the record doubles as a public census of who actually sends mail on whose behalf. The full breakdown updates every night in our daily email infrastructure report.
Pipes are beating suites
SES is not an outlier; it is the leading edge of a pattern. The providers gaining share are the ones that sell sending as infrastructure. SendGrid grew from 3.67% of SPF domains in 2022 to 4.75% today. Mailgun went from 2.61% to 4.10% over the same window. All three are developer-first products: you get an API, an SMTP endpoint, and a bill. Campaign logic, segmentation, templates and analytics are someone else's problem — usually a product bolted on top.
Meanwhile the archetypal all-in-one suite went the other way. Mailchimp peaked at 5.03% of SPF domains in November 2022 and has since drifted down to 3.69%. The suite model is not dead — but the growth has clearly moved to the layer underneath it. We cover the Mailchimp story in detail in a separate analysis of where its customers went.
Why the unbundling happened
Three forces pushed sending volume down the stack, and none of them is a secret:
- Price per message. SES charges commodity rates for raw delivery. For a sender doing serious volume, the gap between infrastructure pricing and suite pricing is large enough to fund an entire engineering function.
- The app layer commoditised. A generation of campaign tools, CRMs and marketing platforms now treat the ESP as a swappable backend. Many of them literally run on SES or SendGrid behind the scenes, so the "pipe" share grows even when the end user never touches an AWS console.
- Multi-ESP became normal. SPF shares across all providers sum to well over 100% because domains routinely authorise several senders at once — marketing through one vendor, transactional through another, support tickets through a third. A cheap, reliable pipe is the natural choice for at least one of those slots. See our breakdown of the healthy multi-ESP stack.
Who wins from unbundling — and who gets hurt
The unbundled stack is a genuine win for some teams and a quiet trap for others. The dividing line is not company size; it is whether anyone on the team owns deliverability as a discipline.
Unbundling works for you if you have engineering capacity, meaningful volume, and someone who watches bounce rates, complaint rates and blocklists as part of their job. You get commodity pricing, full control over IP strategy, and no vendor lock-in at the content layer.
Unbundling hurts you if you treat the pipe like a suite. A suite ESP polices its shared pools: it will throttle you, force double opt-in, and suspend you before you damage the pool — annoying, but it is a deliverability service you are paying for. A raw pipe assumes you know what you are doing. Nobody at the infrastructure layer will stop you from burning your domain reputation with a cold list; you simply start landing in spam, often without a clear signal about when it started.
Moving from a suite to a pipe transfers reputation management, warmup, list hygiene and placement monitoring from the vendor's ops team to yours. The money you save is, in part, the salary of the deliverability function you are now expected to run yourself.
What the DNS data cannot see
Two honest caveats before you quote these numbers. First, SPF shares measure authorisation, not volume: a domain that includes SES for one internal alert stream counts the same as a domain pushing millions of messages a day through it. Second, senders that flatten their SPF records into raw IP addresses are invisible to include-matching, so every provider's true share — SES included — is somewhat higher than what any DNS survey can attribute. Our pattern dictionary covers 81.5% of SPF includes and grows continuously; the methodology is public in the live report.
Operating in an unbundled world
Whether you run SES directly or through a platform built on top of it, the operational takeaway is the same: the pipe will not watch your placement for you. Practical minimum for any team on infrastructure ESPs:
- Separate marketing and transactional streams onto different subdomains so one cannot poison the other's reputation.
- Keep SPF, DKIM and DMARC aligned per stream — a raw pipe will happily send unaligned mail that a suite would have refused to configure.
- Watch complaint and bounce rates at the provider level, not just in your campaign tool's dashboard.
- Seed-test placement across real mailboxes on a schedule. Suites used to absorb placement drops for you by managing the pool; on a pipe, the first sign of trouble is often your own open-rate graph — weeks too late. A regular inbox placement test against Gmail, Microsoft 365 and the other major providers turns that lagging indicator into a leading one.