What is a Digital Carbon Footprint? And Which Numbers Can You Actually Trust?
We first wrote this explainer over a year ago, and we've rewritten it from scratch for an uncomfortable reason: some of the numbers in the original, numbers you'll still find quoted in most articles on this topic, don't stand up anymore.
Digital carbon is a young field, and it has a bad habit: eye-catching but wrong statistics spread much faster than careful ones. As a company that measures website carbon for a living, we'd rather give you smaller, defensible numbers than bigger, broken ones. So this guide does two jobs: it explains what a digital carbon footprint actually is, and it tells you which figures to trust, which to bin, and how to measure your own website honestly.
What a Digital Carbon Footprint Actually Is
Every digital thing you do uses energy across three main operational areas: data centers, networks, and user devices, alongside the embodied carbon required to manufacture the hardware in the first place.
The data centre, where the website or service is hosted and where the computing happens. The network, the cables, exchanges, and mobile masts carry the data to you. And your device, the phone, laptop or TV doing the displaying, which for many activities uses more energy than the other two combined.
Generating that electricity produces carbon emissions, more or less of them depending on how clean the local grid is. On top of the electricity, there's embodied carbon: the emissions from manufacturing the servers, routers, and devices in the first place, which, for hardware like phones, is often the largest slice of its lifetime footprint.
A website's carbon footprint is its share of all that: a small amount of energy per page view, multiplied by every page, every visitor, every day. The per-view number is tiny. The multiplication is what makes it matter.
Why Most of the Numbers You've Read Are Wrong
Here's the cautionary tale that everyone writing about digital carbon should know, and most don't.
For years, the most-quoted statistic in this field was that half an hour of Netflix produced 1.6kg of CO2, roughly the same as driving several miles. It appeared in major outlets worldwide. It was wrong, spectacularly so. When the International Energy Agency fact-checked it, errors compounded throughout the whole calculation: the bitrate was overestimated several times over, data center energy intensity was overstated 35-fold, and network intensity was overstated 50-fold. The IEA's corrected estimate was around 36 grams of CO2 per hour of streaming (based on the global average electricity grid mix of 494g CO2/kWh). However, this figure varies dramatically by region: in France, with its nuclear-heavy grid, streaming produces roughly 2–3 grams per hour; in coal-dependent regions, it can exceed 200 grams per hour. The original Shift Project figure of 1.6 kg per 30 minutes was roughly 90 times the global average estimate, but still dramatically underestimated how much regional grid carbon intensity matters. Even the original authors' own correction remained around 11 times too high.
The same pattern applies to those "every email produces 4g of CO2" claims, The same pattern applies to those "every email produces 4g of CO2" claims, which trace back to Mike Berners-Lee’s 2010 estimate figures based on early network assumptions that modern sustainability researchers no longer consider reliable for today's cloud infrastructure. And it applies to website carbon too. The most widely used calculation method, the Sustainable Web Design Model, released version 4 in 2024, and when the popular Website Carbon calculator adopted it on 14 July 2025, estimates for many pages dropped by up to two-thirds (with typical reductions of 40–66% depending on site structure and caching patterns). This wasn't because websites suddenly got cleaner, it was because better data showed the internet uses less energy per unit of data than the older model assumed: global data transfer roughly doubled between 2019 and 2025, yet total network energy actually fell. Not because websites suddenly got cleaner, but because of a genuine shift in internet infrastructure. Between 2020 and 2025, global IP traffic roughly doubled (from ~4.8 ZB to ~5.29 ZB annually), yet total network energy consumption actually fell thanks to improvements in data center cooling efficiency, widespread adoption of HTTP/2 and compression, and the shift to edge computing and CDNs. The Sustainable Web Design Model v4 incorporates this newer data, which the v3 model, built on 2018–2019 assumptions, had missed.
There are two lessons here worth holding onto.
First, the internet has been getting more efficient and faster, which is genuinely good news.
Second, any carbon figure quoted without its methodology and model version attached should be treated with suspicion, including ours.
When you see a per-page carbon number, the first question to ask is: calculated how, on which model?
What We Actually Know in 2026
So what do the careful numbers say?
The best-measured part of the picture is data centres, and the IEA's current figures are the ones to use. Data centres consumed about 415 terawatt hours of electricity in 2024, roughly 1.5 per cent of global electricity, producing around 180 million tonnes of CO2. That's a real and rising number, but it's smaller than much of the coverage implies.
The rising part matters, though, and the reason is AI. The IEA projects data centre electricity demand will more than double to around 945 terawatt hours by 2030, slightly more than Japan's entire electricity consumption today, with AI as the single biggest driver. Digital emissions are among the fastest-growing sources there are, even as the underlying infrastructure gets more efficient. Add networks and the world's billions of devices, and credible estimates put the full digital footprint in the low single digits of global emissions.The W3C's Web Sustainability Guide
For an individual organisation, the honest summary is this: your website is a small part of a large and growing global system, and it happens to be the part you fully control.
Why It's Worth Caring About Your Website's Share
If the per-page numbers are small, why measure them at all?
Four reasons we see with our customers.
The Multiplication Is Real
A gram or two per page view is nothing. Across a site with a million page views a month, it's tonnes per year, and across a large organisation's whole digital estate it's a number that belongs in the sustainability report alongside travel and buildings. Enterprises are increasingly expected to account for it, and procurement teams increasingly ask suppliers about it.
Carbon Is a Proxy for Waste
A high-carbon page is nearly always a heavy, slow page: oversized images, unnecessary JavaScript, video that autoplays whether anyone watches or not. Cutting page weight cuts emissions, improves load times, and can reduce hosting and bandwidth costs. This is why we talk about the triple win: the same work improves performance, accessibility and carbon together, and it's rarely three separate projects.
The Averages Hide the Problem Pages
In every full-site analysis we run, emissions are unevenly spread. A handful of pages, usually media-heavy ones, produce a wildly disproportionate share of the total. You can't find them by testing your homepage, which is exactly why homepage-only carbon calculators give such a false sense of security.
Credibility Cuts Both Ways
If you're going to make sustainability claims about your website, and more brands are, they need to survive scrutiny. Quoting debunked statistics or unattributed numbers is worse than saying nothing.
A measured, low-carbon website can be a credible claim but only if you meet three conditions:
Publish your methodology. Name the model (SWDM v4), the grid carbon intensity figure you used (global average, or your region), your system boundaries, and your assumptions about repeat visitors and caching. Without this transparency, any claim is indefensible.
Use a reproducible measurement. Third-party tools (Website Carbon, Carbon Badge, Digital Carbon Online) are better than in-house calculations because their methodology is auditable.
Contextualize it honestly. "We use SWDM v4 on the global average grid; our homepage produces 1.2g CO2 per view" is a claim that survives scrutiny. "Our website is low-carbon" alone does not.
How to Measure Yours Honestly
A practical, defensible approach looks like this.
Use a Named Methodology
The Sustainable Web Design Model is the de facto standard for website carbon estimates, and version 4 is current. Whatever tool you use, check which model and version it's on, and say so when you publish figures. If a tool won't tell you, that's your answer. The W3C's Web Sustainability Guidelines (WSG) published its first Draft Note in November 2025, with updates through July 2026. While still in draft status the W3C cautions that this document "may be updated, replaced, or obsoleted at any time" it is the most comprehensive emerging reference for sustainable web design. Expected to reach final NOTE status in late 2026, it's worth bookmarking as the field's leading guidance, with the caveat that any compliance claims should reference the specific dated version you used.
Measure the Whole Site, Not the Homepage
The problem pages are almost never the one page everyone polishes. Full-site analysis is the difference between a vanity score and a to-do list.
Track It Over Time
Every redesign, new campaign and CMS update changes the picture. A single audit tells you where you were; monitoring tells you whether you're improving, and catches the heavy new landing page before it runs for six months.
Fix in Order of Impact
Rank pages by emissions weighted by traffic, and start at the top. The usual culprits, in rough order of payoff: oversized or badly formatted images (modern formats like WebP and AVIF help enormously), autoplaying and unoptimised video, JavaScript nobody needs, and hosting on a provider that can't evidence renewable energy.
This is exactly the job SiteBeacon was built for: it analyses carbon, performance and accessibility across every page of a site, tracks all three over time, and tells you which fixes matter most. One analysis, three wins.
Our Take
Digital carbon has grown up. The era of scary borrowed statistics is ending, partly because fact-checkers caught up and partly because the field now has proper models, real measurement tools and guidelines from the W3C. What's replacing it is more useful: smaller, honest numbers, measured across whole websites, attached to fixes that also make sites faster, cheaper and easier to use.
Our advice is to get ahead of it while it's still a differentiator. Measure your site with a method you can name, fix the pages that matter, and be one of the few businesses whose sustainability claims about their website are actually backed by data. If you'd like to test are no commitment demo, try here or talk to us about monitoring at scale.