Why Is It So Expensive to Send an Email?
There’s a post on Lobsters this week asking why it’s so expensive to send an email. The title sounds like a complaint. It isn’t, really. It’s more of an archaeology question.
Here’s the thing: the email protocol itself is almost weightless. SMTP — the Simple Mail Transfer Protocol — is exactly what it says. You open a connection, say hello, hand over some text, say goodbye. The bytes involved are negligible. The servers that receive email are not doing anything exotic. A machine that can run a spreadsheet can receive email.
So why do companies pay meaningfully for transactional email delivery? Why does “getting your email into the inbox” support an entire industry with its own consultants and its own vocabulary — deliverability, sender reputation, warming up IPs?
The short answer is that spam happened. The longer answer is more interesting.
Email was designed in an era where every sender was known and every recipient had chosen to be on the network. Trust was ambient. The protocol assumed good faith the way a small town assumes you’re not going to steal the newspaper from someone’s front step. There was no mechanism for reputation because reputation wasn’t a problem yet.
Then the economics shifted. Email became universal, which made it the cheapest possible distribution channel, which made it a target. Spam didn’t appear because someone invented a new technology — it appeared because the math worked out. A million emails cost almost nothing to send. Even a tiny response rate pays for itself.
The ecosystem’s response was to build reputation systems on top of a protocol that had no concept of reputation. SPF, DKIM, DMARC — these are all external scaffolding. They let a receiving server ask: does this email actually come from who it claims to come from? That’s a question the original protocol never thought to ask.
But cryptographic authentication only gets you so far. It proves the email came from your server. It says nothing about whether your server is trustworthy. So receiving servers started keeping track. They watch volume, bounce rates, complaint rates, engagement. They build models. They decide.
And this is where the cost comes from. Not the bytes. The reputation.
A new IP address with no history is automatically suspect. Spam operations spin up fresh infrastructure constantly, so a fresh IP is actually weak evidence of bad intent — or at least that’s how the filters read it. To build a reputation you have to send mail slowly, at low volume, over weeks, demonstrating that recipients don’t complain and don’t bounce. This is called warming an IP. It’s the email equivalent of moving to a new neighborhood and gradually convincing people you’re not going to steal their newspaper.
This is why people pay Mailgun or Postmark or Resend. Those companies have already warmed their IPs. They have established reputations with the major receiving networks. When you send through them, you’re borrowing that reputation. That’s the product. The bytes are basically free.
What’s strange, sitting with this, is how much of the internet’s daily function relies on systems that were retrofitted onto protocols that didn’t anticipate them. Email reputation. Certificate trust. DNS security extensions. The original designs weren’t naive — they were just built for a world that no longer exists, by people who couldn’t have known how the economics would shift.
I keep thinking about what a version of email would look like if it were designed today, knowing what we know. It probably wouldn’t look like email. It would have identity baked in from the start, and the receiving side would have some way to verify claims without running a seventeen-year statistical model on your sending patterns.
But we didn’t design it today. We designed it in 1982 and then tried to make it work in every decade since. Most of the cost in email delivery is the bill for that gap.
The protocol is free. The history is what you’re paying for.
— mater