How much energy does sending one email actually use?

Every email you’ve ever sent has a real, physical energy cost — small enough per message to be genuinely invisible, and large enough in aggregate that researchers have spent nearly two decades trying to put an honest number on it. The number that keeps circulating is a fraction of a gram of CO2. It sounds trivial. Multiplied by how much email actually moves through the internet in a single day, it stops sounding trivial at all.

The study everyone still cites, and why it’s dated but not wrong

The most widely repeated figure traces back to a 2009 study McAfee commissioned from ICF International, modeling the carbon footprint of spam specifically. Their estimate: about 0.3 grams of CO2 equivalent per spam email, once you account for the energy to transmit it, run it through filters, store it, and have someone eventually view and delete it. Small on its own — but the study estimated roughly 62 trillion spam emails were sent in 2008 alone, putting the aggregate energy cost at about 33 billion kilowatt-hours a year, which the researchers compared to the annual emissions of 3.1 million passenger cars.

That comparison is now sixteen years old, and a lot has changed since. It’s worth holding two things as true at once instead of picking a side.

Two forces pulling the number in opposite directions

Data centers have gotten dramatically more efficient. The industry metric for this is Power Usage Effectiveness (PUE) — how much total energy a data center consumes versus how much actually reaches the computing hardware, with 1.0 being perfect efficiency and no energy lost to cooling or overhead. Google reports its data centers running close to a 1.10 trailing PUE, well under the roughly 1.58 industry average other operators report — meaning a much larger share of the energy going into modern infrastructure is doing useful work rather than just keeping servers cool. Whatever a single email cost to process in 2009, the underlying infrastructure processing it today is meaningfully leaner.

Email volume has grown enormously. The Radicati Group’s current forecast puts worldwide email traffic at roughly 361.6 billion messages a day — more than five times the daily rate implied by that 2008 spam estimate, and that’s total traffic, not just spam. Even a per-message footprint that’s fallen substantially since 2009 is being multiplied by a denominator that’s grown just as substantially, in the other direction.

Mike Berners-Lee, the researcher whose book How Bad Are Bananas? is the other commonly cited source on this topic, revised his own estimates downward between editions — from a flat few grams per email in the original 2010 edition to a modeled range as low as 0.03 grams for the simplest message, up toward 26 grams once you factor in a large attachment, in the 2020 edition. The range itself is the honest answer: a one-line reply costs almost nothing; a photo or a large attachment sent to a big list is a meaningfully bigger draw, because storage and repeated transmission scale with file size, not with word count.

The real waste was never email. It’s email nobody wanted.

Here’s the reframe worth sitting with: none of these researchers were actually arguing that email itself is the problem. McAfee’s own study specifically targeted spam, not email in general, and found that close to 80% of spam’s climate impact came from the human side of the cost — people opening, scanning, and deleting junk they never asked for. The energy waste wasn’t in the sending. It was in mail that delivered zero value to the person who received it, multiplied by billions of recipients doing exactly the same unproductive work.

That framing generalizes far beyond literal spam. An email sent to a disengaged, unverified, or outdated contact — one who never opens it, never reads it, and either deletes it or reports it — carries the same energy cost as a message that lands and gets a reply, for a fraction of the value. A four-step follow-up sequence blasted at an entire list regardless of engagement, rather than stopped the moment someone’s clearly not interested, is the same pattern at cold-email scale: transmission, storage, and inbox processing spent on messages that were functionally never going to work.

Why a precise per-campaign number isn’t the point

None of the studies above were built to answer “what does my specific campaign cost,” and treating any of these per-message figures as a precise calculator for a real send would overstate what the underlying research actually supports — none of it was measured against a specific sender’s real infrastructure, and the estimates already span a wide range even for a single plain-text message. What the research does support, consistently, across every version of it from 2009 to the present: the footprint of a message barely depends on who’s on the list, but the value generated per unit of footprint depends on it entirely. A verified, engaged list and a stale, unverified one cost roughly the same to send to. Only one of them produces something worth that cost.

Why tighter targeting is a genuine two-for-one

This is the part that should change how you think about list hygiene: every practice that improves cold-email deliverability — verifying addresses before sending, pruning contacts who’ve never engaged, stopping a sequence the moment someone replies or unsubscribes — is also, mechanically, sending fewer wasted messages. It’s not a coincidence that the deliverability fix and the lower-footprint choice are the same action; both come from the same root cause, which is sending mail to people who were never going to want it in the first place. A smaller, better-targeted list isn’t a compromise on reach. For the recipients who actually matter, it’s the same or better outcome, generated by meaningfully less total transmission across everyone else.

Norbelys is built around exactly that principle, independent of the environmental angle: email verification catches dead addresses before they’re ever sent to, audience segments let you stop mailing contacts who’ve gone quiet instead of looping them through every future campaign by default, and sequences stop automatically the moment someone replies instead of continuing to fire follow-ups into silence. None of that is marketed as a sustainability feature — it’s just what good deliverability hygiene actually looks like. It happens to also mean fewer of the pointless sends that make up the bulk of email’s real energy cost. Start sending with Norbelys and send fewer emails that were never going to land anyway — better for your reply rate, and less waste for everyone.