What if you emailed every human at once? The maths of the blast

Sending it costs about $600,000. The reply-all that follows would keep the world's mail servers busy for 181 years.

Clerks sorting letters in a large post office sorting room around 1900
Photo: Archives New Zealand from New Zealand, CC BY-SA 2.0, via Wikimedia Commons

You cannot email every human, because only about 6 billion of the world's roughly 8.1 billion people are online (ITU, 2025), but you can come close: sending one 50 KB email to 6 billion addresses through a bulk service like Amazon SES costs about $600,000, moves about 300 TB and uses about 18,000 kWh of network electricity at most.

Calculated in this post from the ITU's 6 billion internet users and Amazon SES's $0.10 per 1,000 emails, with these example inputs: one 50 KB email, the 0.06 kWh per GB intensity from Aslan et al. (2018). The 8.1 billion world population is implied by the ITU's 74% online share, not stated by it.

Key takeaways
  • Only about 6 billion of roughly 8.1 billion people are online (ITU, 2025): that is the ceiling.
  • Sending one 50 KB email to all of them costs about $600,000 on Amazon SES and moves 300 TB.
  • Gmail's 0.30% spam-complaint limit means 18 million complaints would shut the blast down within the hour.
  • A reply-all storm grows with N²: at the NHS's 0.07% rate, about 2.5 × 10¹⁶ emails, or 181 years of world email.
  • Real spam converts at about 1 sale per 12.5 million emails: roughly 480 sales from the whole planet.

The sending is the easy part. If the addresses were visible and just 0.07% of people hit reply-all, the rate seen in the NHS's 2016 incident, the storm would generate about 25 quadrillion emails, which is about 181 years of all the email on Earth.

This is the marketer's fantasy of perfect reach, and it turns out to fail in an instructive way. Physics and money are not the problem. Mail systems are built around three defences:

  • per-message recipient caps;
  • sender reputation rules;
  • storm breakers.

Each one exists because someone, somewhere, already tried a smaller version of this.

6 billionpeople online in 2025 (ITU), about 74% of humanity
376 billionemails sent and received every day in 2025 (Statista)
$0.10Amazon SES price per 1,000 emails, à la carte
500 millionemails in the NHS reply-all storm of November 2016

How many humans can you actually email?

About 6 billion people used the internet in 2025, according to the ITU, and that is the hard ceiling on who could even receive an email. The ITU puts the online share at about 74%, which implies a world population near 8.1 billion, and about 2.2 billion people offline, 96% of them in low- and middle-income countries.

The real reachable number is lower still. Not every internet user has an email address they check, many people have several, and a large share of addresses on any purchased list are dead. So, by my own rough assumption rather than a measured figure, the blast reaches perhaps 4 to 5 billion distinct humans at best. For the maths below I use \( N = 6\times10^{9} \) addresses, the optimistic upper bound, and flag where the answer changes if \( N \) is smaller.

The story

Early on the morning of 14 November 2016, a local administrator in the NHS sent a blank email with the subject "test" to a distribution list called CroydonPractices. She believed it reached only her own organisation. A software configuration error had applied an "All England" rule, and it went to about 850,000 people with NHSmail accounts.

Staff replied to all, about 80 of them asking to be removed, and each reply went to everyone again. Between 08:29 and 09:45 the system carried about 500 million emails, against a normal 3 to 5 million a day. It never crashed, but some senders saw delays of up to three hours at midday.

The official report blamed the configuration error and said volume controls had not yet been built. This post scales that morning up to the whole planet.

The Register: NHS reply-all email fail, half a billion messages (Jan 2017)

What would it cost to send the email?

Sending 6 billion copies of one email through Amazon SES costs about $600,000 at the à la carte price of $0.10 per 1,000 emails, or about $663,000 on its tiered plan. That is the whole bill for the delivery itself.

Here is the working, with the assumptions stated:

  1. Recipients. \( N = 6\times10^{9} \), one copy each (every address in Bcc or a separate envelope).
  2. Message size. \( S = 50 \) KB including headers, a modest HTML email with no attachment (my estimate).
  3. Price. \( p = \$0.10 \) per 1,000 emails, SES à la carte.
  4. Cost. \( C = N p / 1000 = 6\times10^{9} \times 0.10 / 1000 = \$600{,}000 \).

    Calculated in this post from the ITU's 6 billion users and Amazon SES's $0.10 per 1,000.

  5. Data. \( D = N S = 6\times10^{9} \times 5\times10^{4}\ \text{B} = 3\times10^{14}\ \text{B} = 300\ \text{TB} \).
\[ C = \frac{N\,p}{1000} = \frac{6\times10^{9}\times \$0.10}{1000} = \$6\times10^{5}, \qquad D = N S = 3\times10^{14}\ \text{B} \]

On the tiered Essentials plan, according to Amazon's SES pricing page, the first 10 million emails a month cost $0.16 per 1,000, the next 90 million $0.14, and everything above 100 million $0.11, which adds up to about $663,000. Either way, the email to all of humanity costs roughly ₹5 crore at about ₹88 to the dollar.

Calculated in this post from the Amazon SES tier prices: 10 million at $0.16, 90 million at $0.14 and 5.9 billion at $0.11 per 1,000, for 6 billion emails.

How big is one email to everyone? Daily volumes compared
See the data as a table
Emails (billion emails)
All email worldwide, one day (2025)376
One copy to every internet user6
NHS reply-all storm, 20160.5
Gmail Workspace daily cap for one user2e-06

Source: Statista (2025 daily email), ITU Facts and Figures 2025, Digital Health/The Register, Google Workspace help · The reply-all storm modelled below (25 quadrillion emails) would not fit on this axis.

How much energy would the email use?

Very little: moving 300 TB across the internet uses roughly 18,000 kWh at the 2015 network intensity of 0.06 kWh per GB (Aslan et al., 2018, read the abstract), and far less at today's. This is the part that surprises people, because "the carbon cost of email" is a popular worry.

Calculated in this post from the 300 TB computed above and Aslan et al.'s 0.06 kWh per GB.

\[ E = D \times I = 3\times10^{5}\ \text{GB} \times 0.06\ \tfrac{\text{kWh}}{\text{GB}} = 1.8\times10^{4}\ \text{kWh} \]

The intensity figure comes from the most careful review of the question. Its authors found that the electricity per gigabyte of fixed-line transmission fell by half roughly every two years from 2000 to 2015. If that trend held for ten more years, which is my extrapolation and not their claim, 2025's figure would be about \( 0.06/2^{5} \approx 0.0019 \) kWh/GB, and the whole send would use about 560 kWh. Storage and spam filtering at the receiving end add to this, but they do not change the order of magnitude.

Electricity Intensity of Internet Data Transmission: Untangling the EstimatesJoshua Aslan, Kieren Mayers, Jonathan Koomey and Chris France · Journal of Industrial Ecology 22(4) · 2018

Published estimates of the energy cost of moving data online ranged across five orders of magnitude, from 136 kWh/GB (for 2000) down to 0.004 kWh/GB (for 2008). The authors traced the spread to differences in system boundaries, assumptions and data year, not to method, and set out criteria for a sound estimate. Their figure for core and fixed-line access networks in 2015 is 0.06 kWh/GB, with intensity halving about every two years since 2000 in developed countries. The limit: it excludes data centres and end-user devices, and mobile networks are more energy-intensive per GB.

Read the abstract (free)
Cross-section of an optical fibre cable showing its strands
300 TB is a rounding error for modern fibre networks. The bottleneck is not the glass but the mail servers deciding whether to accept you.Photo: Asurnipal, CC BY-SA 4.0, via Wikimedia Commons

Could the internet carry it all at once?

Sending all 6 billion copies in one hour would need about 1.7 million emails a second, which is roughly 38% of the entire world's email traffic at that moment. The network could carry the bytes; the receiving mail servers would refuse them.

Calculated in this post from 6 billion internet users sent over a one-hour premise, and Statista's 376 billion emails a day in 2025.

\[ \dot{n} = \frac{N}{t} = \frac{6\times10^{9}}{3600\ \text{s}} \approx 1.7\times10^{6}\ \text{emails/s}, \qquad \dot{n}_{\text{world}} = \frac{3.76\times10^{11}}{86{,}400\ \text{s}} \approx 4.4\times10^{6}\ \text{emails/s} \]

Gmail's bulk sender rules

Large mailbox providers throttle unknown senders, and Gmail's rules are explicit. Since February 2024 anyone sending more than 5,000 messages a day to Gmail accounts must:

  • authenticate with SPF, DKIM and DMARC;
  • offer one-click unsubscribe;
  • keep the spam-complaint rate below 0.30%, with Google advising under 0.10%.

An unsolicited email to everyone would breach that in minutes: 0.3% of 6 billion is 18 million complaints. The real-world outcome is that the first few million copies get through and the rest bounce or land in spam. I cover the plumbing in how SPF, DKIM and DMARC decide deliverability.

Exterior of a large old post office sorting building
Bulk mail was always a throughput problem. Sorting offices like this one were built for peaks; email servers handle theirs by refusing senders they do not trust.Photo: Tim Sheerman-Chase, CC BY 2.0, via Wikimedia Commons
Portrait of Ray Tomlinson, the engineer who sent the first networked email
Ray Tomlinson, who built the first networked email system in 1971 and chose the @ sign to separate a user from their host. Addressing has been the hard part ever since.Photo: Roll-Stone, CC BY-SA 3.0, via Wikimedia Commons

Why can't you just put everyone in the To field?

Three gates between you and 6 billion inboxesYour email to everyone: 6 × 10⁹ copiesGate 1: recipient caps2,000 per message, 500 external, 10,000 a day(Google Workspace); a 150 GB To line can't be sentGate 2: reputation and spam rateGmail bulk senders: SPF, DKIM, DMARC,complaints under 0.30% (aim for 0.10%)Gate 3: storm breakersExchange Online: 10 reply-alls, 5,000+ recipients,60 minutes, then replies blocked for 4 hours
Each gate exists because someone already tried a smaller version of this. Figures from Google Workspace, Google's sender guidelines and Microsoft's Exchange team.

A single email with every human in the To line is impossible, because the address list alone would be about 150 GB, roughly 6,000 times Gmail's 25 MB sending limit. At an average of 25 bytes per address (my estimate, including the separator), the header is \( 6\times10^{9} \times 25 = 1.5\times10^{11} \) bytes.

Calculated in this post from 6 billion addresses, with these example inputs: 25 bytes per address, and Gmail's 25 MB attachment limit.

Services also cap recipients per message for ordinary users. Google Workspace's sending limits let an account address 2,000 recipients per message, at most 500 of them external, and 10,000 recipients per day. At that rate one account would need about 600,000 days, or 1,600 years, to reach everyone. So the "one email" is necessarily 6 billion separate copies sent by bulk infrastructure, each addressed to one person. That detail, it turns out, is what saves civilisation.

Calculated in this post from Google's 10,000 recipients a day (linked above) and 6 billion addresses.

What happens if everyone hits reply-all?

If the 6 billion addresses were visible and 0.07% of recipients hit reply-all, about 4 million replies would each go to 6 billion people, producing about 2.5 × 10¹⁶ emails, or 181 years of the world's current email traffic. The reply-all rate is not a guess; I took it from the largest reply-all storm on record.

Go back to the NHS morning told above. A test email went by mistake to every NHSmail account, about 850,000 people, on 14 November 2016. Staff replied to all, many asking to be removed, and between 08:29 and 09:45 the system carried about 500 million emails. That is about 590 reply-alls, each reaching 850,000 inboxes.

  1. Reply-all fraction from the NHS case. \( p = 590/850{,}000 \approx 6.9\times10^{-4} \), or 0.07%.
  2. Calculated in this post from The Register's 850,000 recipients and 500 million emails: 500 million / 850,000 is about 590 reply-alls.

  3. Replies in the global case. \( r = pN = 6.9\times10^{-4} \times 6\times10^{9} \approx 4.2\times10^{6} \).
  4. Emails generated. \( E = rN = 4.2\times10^{6} \times 6\times10^{9} \approx 2.5\times10^{16} \).
  5. Compared with world email. \( 2.5\times10^{16} / 3.76\times10^{11} \approx 66{,}000 \) days, about 181 years.
  6. Calculated in this post from the storm size above and Statista's 376 billion emails a day.

  7. Data and energy. At 50 KB each, about 1.25 zettabytes; at 0.06 kWh/GB, about 75 TWh, or about 2.3 TWh at the extrapolated 2025 intensity.
\[ E = p N^{2} = 6.9\times10^{-4} \times (6\times10^{9})^{2} \approx 2.5\times10^{16}\ \text{emails} \]
The N² is the whole story: a list 7,000 times bigger than the NHS's makes a storm about 50 million times bigger.

The \( N^{2} \) is the whole story. Email volume from a reply-all storm grows with the square of the list size, so a list 7,000 times bigger than the NHS's makes a storm about 50 million times bigger.

Why does a reply-all storm explode?

A reply-all storm is a branching process, the same mathematics as an epidemic or a nuclear chain reaction: it dies out only if each message triggers fewer than one new message on average. In a Galton-Watson branching process, if each reply-all causes on average \( R \) further reply-alls, the storm fades when \( R < 1 \) and grows without limit when \( R > 1 \).

\[ R = p\,N \quad\Rightarrow\quad R < 1 \iff p < \frac{1}{N} = \frac{1}{6\times10^{9}} \approx 1.7\times10^{-10} \]
Why reply-all storms explode1 email to N peopleEach reply-all goes to all N people again.Reply-alls per message: R = p × NNHS 2016: p ≈ 0.07%, N ≈ 850,000, so R ≈ 590Global: N = 6 billion, so R ≈ 4 millionThe storm dies out only if R < 1, i.e. p < 1 in 6 billion.
A Galton-Watson branching process. Each generation multiplies by R; with any realistic human reply-all rate, R is far above one.

To keep the global storm from exploding, fewer than one person in six billion may reply-all. Human behaviour is about four million times worse than that.

Each "please remove me" and "stop replying all" is itself a reply-all, so in practice the storm feeds on its own complaints, exactly as NHS staff found. The only reason it would not run for 181 years is that mail systems would collapse or cut it off first: at the world's present rate of about 4.4 million emails a second, the queue could not physically drain.

Engineers have already built the circuit breaker. Microsoft's Reply-All Storm Protection in Exchange Online, rolled out from 2020, detects 10 reply-alls to a thread with more than 5,000 recipients within 60 minutes and blocks further replies for four hours. It is epidemic control applied to an inbox: cut \( R \) below 1 by quarantining the thread.

ScenarioRecipientsEmails generatedData at 50 KBWhat stops it
Bcc blast via bulk service6 billion6 × 10⁹300 TBSpam-rate limits, bounces
One message, everyone in To6 billionCannot be sent150 GB header aloneMessage size and recipient caps
NHSmail storm, 2016~850,000~5 × 10⁸~25 TBEngineers, after ~75 minutes
Global reply-all storm, 0.07%6 billion~2.5 × 10¹⁶~1.25 ZBStorm protection, or collapse

Email everyone: your own assumptions

Would anyone actually buy anything?

Almost no one: the best measured conversion rate for mass unsolicited email is about one sale per 12.5 million messages, which predicts roughly 480 sales from 6 billion emails. That number comes from researchers who took over part of a real spam botnet and counted.

Spamalytics: An Empirical Analysis of Spam Marketing ConversionChris Kanich, Christian Kreibich, Kirill Levchenko and others · ACM CCS 2008

The team infiltrated the Storm botnet and rewrote about 470 million spam emails in its campaigns so the links pointed to sites the researchers controlled: 347 million pharmacy spams, 83 million greeting-card spams and 40 million April Fools spams. The pharmacy campaign produced 28 "purchases", one per 12.5 million emails. The malware campaigns produced 541 infections, about one per 178,000 to 265,000 emails. The limit: the offers were obvious spam, and spam filters removed most messages before anyone saw them, so this is a floor for any email, not a forecast for a good one.

Read the paper (free)

Apply it: \( 6\times10^{9} / 1.25\times10^{7} = 480 \) purchases, for about $600,000 of sending, or roughly $1,250 per customer before anyone counts the cost of the domain being blacklisted. A permission-based list of one lakh people who asked to hear from you will usually beat that. Reach was never the scarce thing; attention and consent are.

That is the same lesson as WhatsApp Business API marketing in India, where Meta enforces quality ratings for the same reason Gmail enforces spam rates.

A large former Royal Mail sorting office building
Physical mail priced every letter, which kept junk mail in check. Email made the marginal message almost free, so the gatekeeping moved to reputation systems.Photo: Tim Sheerman-Chase, CC BY 2.0, via Wikimedia Commons

How sensitive is the answer to the assumptions?

  • Fewer real people. If only 4.5 billion addresses are live, the cost falls to $450,000 and the storm, which scales with \( N^{2} \), shrinks to about 56% of the estimate. It is still more than a century of world email.

    Calculated in this post from the 6 billion cost (SES price) and storm figures above, with a premise of 4.5 billion live addresses (my assumption, not a measured figure).

  • Bigger email. A 500 KB message with images makes 3 PB of data and about 180,000 kWh at 2015 intensity. Still trivial against what the storm would do.
  • Calmer humans. Cut the reply-all rate tenfold to 0.007% and the storm is still about 2.5 × 10¹⁵ emails, 18 years of world traffic. You would need to cut it about four million-fold to stop the chain reaction.
  • Calculated in this post from the 2.5 × 10¹⁶ storm above and Statista's 376 billion emails a day, with a premise of a tenfold lower reply-all rate.

  • Storm protection on. With a circuit breaker like Microsoft's, the thread is frozen after about ten reply-alls, so the storm is about \( 10N = 6\times10^{10} \) emails, roughly a sixth of a normal day's global traffic. Survivable.
Verdict: possible with changes

A single email with all of humanity in the To line is impossible: the address list alone is about 150 GB against a 25 MB limit. Sending 6 billion separate copies is technically possible for about $600,000, 300 TB of data and under 20,000 kWh, but spam-rate rules (0.3% complaints) would block most of it within the hour. If addresses were visible, a reply-all chain reaction with \( R = pN \approx 4\times10^{6} \) would produce about 2.5 × 10¹⁶ emails; only storm-protection circuit breakers make it survivable.

What if next?

Questions people ask

What is the largest reply-all incident ever?

The best documented is the NHSmail incident of 14 November 2016, when a test message reached about 850,000 NHS accounts and reply-alls drove about 500 million emails in about 75 minutes, slowing the system for the health service. Storms like it are why Microsoft built automatic storm protection into Exchange Online.

How many emails can I send a day from Gmail?

A Google Workspace user can send 2,000 messages a day and reach 10,000 recipients a day in total, with up to 2,000 recipients per message, of which at most 500 can be outside the organisation. Bulk senders above 5,000 messages a day to Gmail users must also meet authentication and spam-rate rules.

Does deleting old emails really save energy?

Barely. Transmission energy per GB has been halving roughly every two years, and a stored email is a few tens of kilobytes. The bigger energy costs sit in data centres and devices, and they depend on total usage, not on individual inbox clean-ups.

Is it legal to send marketing email to people who did not sign up?

The rules vary by country. Many require consent or an easy opt-out, and India's Digital Personal Data Protection Act, 2023 treats personal data such as email addresses as needing a lawful basis. Separately, Gmail and other providers will block senders with high complaint rates regardless of the law. Check the rules for each country you send to.

Sources

  1. ITU: Facts and Figures 2025 press release (source for 6 billion users, about 74% online, 2.2 billion offline)
  2. Statista: Daily number of e-mails worldwide (source for 376 billion emails a day in 2025)
  3. Amazon Web Services: Amazon SES pricing (source for $0.10 per 1,000 and tiered rates)
  4. Aslan et al., Journal of Industrial Ecology: Electricity intensity of internet data transmission (2018; source for 0.06 kWh/GB and the halving trend)
  5. Kanich et al., ACM CCS: Spamalytics (2008; source for 1 purchase per 12.5 million spams)
  6. The Register: NHS reply-all email fail, half a billion messages (Jan 2017; source for 850,000 recipients and 500 million emails between 08:29 and 09:45)
  7. Microsoft Exchange Team: Reply-all Storm Protection in Exchange Online (2020; source for the 10 replies, 5,000 recipients, 60 minutes and 4-hour block)
  8. Google: Email sender guidelines (source for the 5,000-a-day bulk threshold and 0.30% spam rate)
  9. Google Workspace: Gmail sending limits (source for 2,000 messages and 10,000 recipients a day)
  10. Wikipedia: Ray Tomlinson (source for the 1971 networked email and the @ sign)
  11. Gmail Help: Attachment size limit (source for the 25 MB sending limit)
email marketingreply-all stormemail deliverabilityAmazon SESGmail bulk sender rulesspambranching process
Sapta

Sapta (Saptarshi Sarkar) is a mechanical engineer turned marketer in Bengaluru and a solo touring rider. He writes the seven Saptaverse blogs. Work with him or read more of Marketblog.

Keep reading in Marketblog

From the other blogs