Email infrastructure · Economics

The Cost of Email Infrastructure: ESP, Self-Hosted, and Where the Break-Even Really Is

”How much does email infrastructure cost” is one of those questions where the honest answer is “less than you think to start, and more than you think to run.” The headline prices — a few dollars for a server, a few hundred for an ESP — hide the line items that actually decide the economics. This guide breaks down what each sending model really costs, where the crossover between them sits, and the hidden labour that turns a “cheap” self-hosted setup into an expensive one.

What does email infrastructure actually cost?

Before comparing prices it helps to see what you’re buying, because the models bundle very different things. A managed email service provider gives you four things in a single bill: the mail transfer agent software, the IP infrastructure, the deliverability operations, and bounce, complaint, and feedback-loop handling, plus support. Self-hosting unbundles all of that — you assemble the software, servers, IPs, monitoring, and operations yourself, and you own every piece. Managed dedicated sending sits in between, where a provider runs the hardware and maintenance while you keep control of the configuration and reputation.

The single most misunderstood fact about email infrastructure cost is which part dominates. People quote the software — “PowerMTA is a few thousand a year” or “KumoMTA is free” — as if that settles it. But the software licence is the smallest line item in a serious operation. The costs that actually add up are the servers, the IPs, the monitoring tools, and above all the people who run and tune the system. Quote only the licence and you’re quoting the sticker price of a car while ignoring the insurance, fuel, and maintenance.

The three ways to send, and how they’re priced

Each model prices differently, which is exactly why the right choice depends on your volume and capability. The table sets the three against the dimensions that drive the decision.

The ESP end is pay-as-you-go: little or nothing upfront, but a per-email rate that climbs with volume. Amazon SES is the cheapest at scale, around a hundred dollars per million messages, though it’s bare and you build deliverability around it; SendGrid and Mailgun bundle more and cost more, in the hundreds per month at a million emails. Self-hosting inverts the shape — meaningful upfront effort and a near-flat marginal cost afterward — while managed dedicated trades a setup fee for a flat hosting bill with no per-email charge, which is why the crossover between these curves is the whole game.

The real cost components of self-hosting

To cost self-hosting honestly you have to list every line, not just the cheap ones. The software can be free with Postfix or KumoMTA, or a commercial licence in the low thousands per year with PowerMTA. The server is modest for most volumes — a small VPS runs five to twenty dollars a month and handles tens of thousands of emails comfortably, scaling to a few hundred a month for a million, with the one hard requirement being port 25 access, which many cloud providers block. Then come dedicated IPs, sending domains, monitoring and seed-testing tools, and backups.

# Honest monthly cost, self-hosted MTA at ~1M emails/month
VPS / server (4 vCPU, 8 GB) … $   80
Dedicated IPs (a few) … $   40
MTA software (KumoMTA free / PowerMTA) . $ 0-300
Monitoring + seed testing … $   50
Backups … $   10
Labour: 2-4 hrs/mo @ $75-200/hr … $ 150-800   <— the real cost
                                        ----------
All-in, most operators … ~$ 100-500
# Compare: a managed ESP at 1M/mo runs ~$400-1,000+/mo, support included.

The line that dominates is the last one. Maintenance runs one to four hours a month, which at a real hourly rate is a hundred and fifty to eight hundred dollars — and that’s for a stable server, before any incident eats a day. Add the twenty to forty hours of initial learning, the unpredictable hours of blacklist removals and outages, and the fact that a serious operation needs a quarter to a full engineer’s time, and the honest all-in cost for most operators lands between one hundred and five hundred dollars a month. The free software was never the point.

Where is the break-even point?

The crossover follows directly from those two cost shapes: an ESP’s bill rises with volume, while a self-hosted bill is mostly a fixed floor. Below that floor, the ESP is cheaper; above it, self-hosting wins. The practical thresholds are well established. Under roughly twenty-five to fifty thousand emails a month, an ESP almost always wins because the maintenance time exceeds the savings. Up to a couple hundred thousand a month, the ESP is usually still cheaper than the hours you’d spend — unless you already employ a sysadmin who can absorb the work, in which case the marginal time cost approaches zero and self-hosting can pay much earlier.

Above a million emails a month, with dedicated engineering capacity, self-hosting starts to pay back — and above ten million it becomes structurally favourable, with raw infrastructure running eighty to ninety percent below what SendGrid or Mailgun charge before you count labour. The crucial caveat is that this chart moves: the break-even is not a fixed volume but a function of your hourly cost and whether you already have the skills, so a team without an email engineer should mentally shift the crossover well to the right.

The hidden costs nobody quotes

Both models carry costs that don’t appear on the headline price, and they cut in opposite directions. On the self-hosted side, beyond labour, there’s the IP warmup period of two to four weeks of reduced sending before a new IP performs, the work of building your own bounce and feedback-loop processing that an ESP includes, backup infrastructure, and security — an unpatched mail server is an open relay waiting to be abused. None of these show up in the “five-dollar VPS” math, and any one of them can erase the apparent savings.

On the ESP side, the hidden costs are commercial and structural. There are overage and egress charges, a per-email rate that quietly compounds at scale, and the reputation risk of shared IP pools, where a careless neighbour can damage your deliverability through no fault of yours. There’s vendor lock-in, and there’s the operational reality that when something breaks you submit a ticket and wait on someone else’s timeline. ESP prices have also risen after a wave of acquisitions, which widens the gap with self-hosting year over year and is part of why senders eventually outgrow their provider — a moment our when to leave your ESP guide covers in detail.

The middle path: managed dedicated

The framing of “ESP versus self-hosted” hides a third option that fits a lot of senders better than either extreme. Managed dedicated sending — where a provider runs the hardware, handles updates and monitoring, and keeps the lights on, while you get dedicated IPs and full control of your sending configuration — captures most of the self-hosted economics without the on-call burden. In practice this lands around one to two hundred dollars a month all-in for moderate volumes, which is cheaper than most ESP plans above a couple hundred thousand a month and requires little to no technical involvement.

This is the option that resolves the labour problem at the heart of the self-hosting math. You stop paying per email, you own your IP reputation, and you keep the configurability an ESP can’t match — but you don’t spend your evenings reading mail logs or your weekend on a blacklist removal. For senders who want owned-infrastructure economics and control without building an operations team, it’s frequently the right answer, and it’s the model behind our PowerMTA server hosting on dedicated bare-metal infrastructure.

What about hybrid models?

You don’t have to pick one model for everything, and the strongest setups often don’t. The most common hybrid splits by stream: high-volume marketing runs on self-hosted or dedicated infrastructure where the per-email economics matter, while critical transactional mail — password resets, receipts — goes through a premium transactional service like Postmark where reliability matters more than cost. This also enforces the stream separation that protects deliverability, since a marketing complaint spike can’t then taint your transactional reputation.

A second hybrid pattern is reliability insurance: self-hosted infrastructure handles the bulk of normal traffic, with a thin, mostly-unused ESP account kept warm as a fallback for the handful of times a year something goes wrong — a regional outage, an IP listing, a capacity spike. The cost is small and the benefit is self-hosted economics for the ninety-five percent of traffic with managed reliability for the five percent of edge cases. Hybrids cost a little more to design but often deliver the best blend of price, control, and resilience.

Should you self-host to save money?

For most senders, the honest answer is no. Below the break-even — and that’s where the large majority of senders sit — an ESP costs less than the time you’d spend running a server, and “I’ll self-host to save money” turns into twenty hours troubleshooting a blacklist issue an ESP would have handled automatically. The trap is reading the five-dollar VPS price as the cost of self-hosting, when the real cost is everything else. If the only reason to self-host is to save money, and your volume is modest, the math usually says don’t.

Self-hosting earns its place when three things line up: enough volume that the fixed costs amortise, typically a million or more emails a month; the operational capability to run it, whether an in-house sysadmin or an agency managing many client domains; and a reason beyond price, such as full control over reputation, data ownership, or escaping shared-IP risk. When those align, self-hosting is a clear win and saves real money. When any one is missing, an ESP or the managed-dedicated middle path will serve you better and cheaper — once you count the labour honestly.

Matching the cost model to your situation

There is no universally cheapest way to send email; there is only the cheapest way for your volume and your team. Map your monthly volume against your operational capacity and the answer usually falls out: low volume or no engineers points to an ESP, high volume with engineering capacity points to self-hosting, and the broad middle — meaningful volume without wanting to run servers — points to managed dedicated. Layer a hybrid on top to protect transactional mail and add resilience, and you have a sending architecture priced to fit rather than to a slogan.

Whatever you choose, count the full cost, not the sticker price. The software is the cheap part; the people, the monitoring, the warmup, and the incident response are where the money and time actually go. Our email infrastructure guide covers how to design the architecture itself, ESP pricing trends tracks where the managed market is heading, and PowerMTA pricing explained digs into the commercial-MTA line item — so you can put a real number against each path before committing to one.