Digital, IT & Cloud Carbon Footprint Calculators
Every pageview, prompt, stream and stored file draws power somewhere — and because that power sits in someone else’s data centre, the footprint is easy to ignore and hard to measure. The same website can produce carbon figures an order of magnitude apart depending on whose model and whose grid you use.
Pick what you’re measuring and get a figure built on a named method, with the assumptions that drive it made explicit.
Digital emissions are estimated by multiplying an activity measure — pageviews, AI queries, gigabytes, compute hours or devices — by an energy-intensity factor, then by the grid’s carbon factor. The grid factor usually dominates the result.
Website figures use the Sustainable Web Design model; software uses ISO/IEC 21031 Software Carbon Intensity; data-centre efficiency uses ISO/IEC 30134 and Uptime Institute PUE; crypto uses the Cambridge CBECI. All convert electricity to carbon on grid emission factors, versioned in the MasterBrain data layer.
Pick what you’re measuring
Know what you need? Go straight to the calculator. Not sure which fits, or why two digital tools give different numbers for the same thing? What each calculator covers and the decision table map it out, and why digital footprints are hard to pin down explains the ranges.
Three sub-topic hubs break this down: website & digital content, AI, compute & crypto, and data centre & IT infrastructure.
Website
Per-pageview site footprint, SWD model.
AI & ML
Training and inference compute.
Digital service
Whole digital product, not just the site.
Send, receive and storage volume.
Digital advertising
Ad impressions and delivery.
AI search query
Per-prompt cost of a generative query.
Cloud compute
Allocated compute across instances.
Cloud storage
Stored data by volume and duration.
Data centre
Facility overhead via PUE.
Streaming & video
Streaming and conferencing hours.
Bitcoin
Network or holding electricity, CBECI.
NFT & crypto
Tokens and blockchain transactions.
Devices
Laptops, phones, monitors in use.
IT hardware & e-waste
Purchase-to-disposal lifecycle.
Software (SCI)
Per-function intensity, ISO/IEC 21031.
Provider directory
Need software to run this?
See who does this work. Every listing names the standards it works to, and paid placements are labelled. Including Orki and Terrascope.
Browse 4 carbon software providers →Do this work? A listing is US$390 a year. Get listed →
What each calculator covers
Digital emissions are almost always someone else’s electricity — a server, a network, a device — converted to carbon by the local grid. The activity you can measure differs by workload, so each calculator carries the profile for one type: a website counts pageviews, an AI model counts compute, a data centre counts facility overhead.
Websites & digital services
The website carbon calculator estimates a site’s footprint per pageview on the Sustainable Web Design model — data transfer, energy intensity and grid factor — the quickest way to size a page or a redesign. The digital service calculator goes wider than the page itself, covering the whole service — front end, back end and supporting infrastructure — for teams accounting for a product rather than a marketing site. The email calculator estimates send, receive and storage volume, and the digital advertising calculator covers ad impressions and delivery — a fast-growing line for media and marketing teams.
AI & machine learning
AI splits into two very different questions. The AI compute calculator estimates both training (a large one-off) and inference (small but repeated at scale), so you can see which dominates for your workload — for a widely used model, cumulative inference usually outweighs training. The AI search query calculator answers the consumer-facing question directly: the per-prompt cost of a single generative query. Both depend heavily on the hardware and the grid powering it, which is why headline “per query” figures vary so widely between sources.
Cloud compute, storage & data centres
Cloud emissions are the electricity of infrastructure you rent, not own — accounting-wise usually Scope 3 purchased services rather than your own Scope 2. The cloud compute calculator allocates emissions across instances and utilisation, and the cloud storage calculator estimates stored data by volume and retention. The data-centre PUE calculator handles the facility layer — the power drawn beyond the servers themselves for cooling and overhead, expressed as Power Usage Effectiveness on the ISO/IEC 30134 and Uptime Institute definitions.
Streaming, video & conferencing
The streaming and conferencing calculator estimates the footprint of watched or transmitted video by hours, resolution and device. It is the tool behind the recurring “is streaming greener than downloading” question — the answer turns on how many times a file is watched and at what resolution, and this calculator makes that trade-off explicit rather than asserting a single verdict.
Crypto & NFTs
The bitcoin emissions calculator estimates the electricity behind a share of the network or a holding, using the Cambridge CBECI for network consumption and mining-region grid mixes. The NFT and crypto calculator covers tokens and transactions, where the footprint depends sharply on whether the underlying chain uses proof-of-work or the far lower-energy proof-of-stake — a distinction that has changed some networks’ footprints by orders of magnitude.
End-user devices, e-waste & Software Carbon Intensity
Hardware carries both use-phase and embodied emissions. The end-user devices calculator estimates the in-use electricity of laptops, phones and monitors across a fleet, while the IT asset and e-waste lifecycle calculator covers the fuller picture — manufacturing, use and end-of-life disposal, aligned to R2v3 responsible-recycling practice. For engineering teams, the Software Carbon Intensity calculator applies the ISO/IEC 21031 SCI metric — carbon per unit of software work — the emerging standard for measuring an application rather than a whole data centre.
Which digital calculator do I need?
The workload decides both the calculator and the unit it counts in. This table maps each tool to what it measures, the method it uses, and when to reach for it.
| What you’re measuring | Calculator | Unit | Method / standard | When to use it | Methodology |
|---|---|---|---|---|---|
| A website or web page | Website calculator | Per pageview | Sustainable Web Design model | Sizing a site, page or redesign | Website carbon methodology |
| A whole digital product | Digital service calculator | Per service | SWD + infrastructure allocation | Accounting for a product, not a marketing site | Website carbon methodology |
| AI training & inference | AI compute calculator | Per model / per run | Compute-energy + grid factor | Estimating a model’s build or serving cost | AI compute methodology |
| A single AI query | AI search calculator | Per query | Per-inference energy estimate | The “what does one prompt cost” question | AI query methodology |
| Rented cloud compute | Cloud compute calculator | Per instance-hour | Allocation by utilisation | Cloud infrastructure emissions (Scope 3) | Cloud compute methodology |
| Stored data | Cloud storage calculator | Per GB-month | Storage energy + grid factor | Long-lived data and backups | Cloud storage methodology |
| Data-centre overhead | Data-centre PUE calculator | PUE ratio | ISO/IEC 30134 · Uptime Institute | Facility efficiency and cooling load | Data-centre PUE methodology |
| Streaming & video | Streaming calculator | Per hour | Data-transfer + device energy | Streaming, video and conferencing | Video streaming methodology |
| Crypto | Bitcoin, NFT / crypto | Per network share / transaction | Cambridge CBECI · chain consensus | Holdings, mining or token activity | Crypto emissions methodology |
| Devices & hardware | Devices, IT asset lifecycle | Per device | Use-phase + embodied + end-of-life | A device fleet, from purchase to disposal | Devices & e-waste methodology |
| An application | SCI calculator | Per unit of work | ISO/IEC 21031 SCI | Measuring software, not the whole facility | Software carbon intensity methodology |
Nearly every digital number is electricity converted to carbon, so the grid factor you apply usually moves the result more than any other input. The same AI query, website or stream can differ severalfold between a coal-heavy grid and a low-carbon one. Use a location-based grid factor for a physical-reality view and a market-based one where you procure renewables, and state which you used. A screener figure is a defensible estimate — for CSRD, IFRS S2 or CDP, build the reported number on measured energy data with documented boundaries.
How the numbers are built
Every calculator on this page reads the same emission factors as our detailed, source-level tools. The chain is consistent across workloads: an activity measure (pageviews, queries, compute hours, gigabytes, devices) is multiplied by an energy-intensity factor for that activity, then by the carbon intensity of the electricity powering it. Websites follow the Sustainable Web Design model; software follows ISO/IEC 21031 SCI; data centres use ISO/IEC 30134 KPIs and Uptime Institute PUE; bitcoin uses the Cambridge CBECI; and hardware disposal references R2v3. Greenhouse gases are expressed as CO₂e on IPCC AR6 global warming potentials.
Every digital footprint splits across the same three physical layers — which is what makes them hard to pin down, and what these calculators divide the problem into:
| Layer | What it is | Calculators here |
|---|---|---|
| End-user devices | Making and powering your phone, laptop and screen | End-user devices & e-waste |
| Networks | Moving data across the internet | Folded into website & streaming |
| Data centres | Servers, storage and cooling — the PUE overhead | Cloud compute, cloud storage, data-centre PUE, AI |
Each layer carries both embodied emissions (making the hardware) and operational emissions (electricity to run it), which is why a digital estimate depends heavily on assumed device lifetime and grid intensity.
The electricity conversion draws on our grid electricity emission factors, every value versioned and traceable in the MasterBrain data layer so you can see when and why a factor changed. Each calculator also has a matching methodology page — for example the website pageview (SWD) methodology, the AI compute methodology and the Software Carbon Intensity methodology — setting out the boundaries and assumptions behind each estimate.
A few of the anchor factors these calculators read live from the data layer at MasterBrain v2026.284 — different units, one theme: data movement, data-centre overhead and AI silicon all cost energy:
| Factor | Value | Unit | Source |
|---|---|---|---|
| Website data transfer (SWD blended) | 0.81 | kWh/GB | SWD model v3 |
| Data-centre PUE — hyperscale (best) | 1.1 | ratio | IEA 2025 |
| Data-centre PUE — enterprise (typical) | 1.7 | ratio | IEA 2025 |
| NVIDIA H100 accelerator — power draw | 700 | W | NVIDIA 2025 |
A website “gram per page view” or an “AI query in watt-hours” is built by chaining several assumptions — bytes transferred, device mix, data-centre PUE, grid intensity — each with its own uncertainty. That is why credible digital footprints are reported as ranges, and why two reputable tools can differ severalfold for the same thing. Use these numbers to compare options and find the big levers, not as a precise meter reading.
Why digital footprints are hard to pin down
Digital carbon figures carry more uncertainty than most emission categories, and it is honest to say why. Three things drive the spread. The grid factor: the same workload run in different regions, or costed on a location- versus market-based basis, can differ severalfold. The boundary: a “website” figure may or may not include the device you view it on, the network in between, and the share of always-on infrastructure — different published models draw that line differently, which is why per-pageview and per-AI-query numbers vary by an order of magnitude between sources. And allocation: splitting a shared data centre or a trained model across its many users involves a judgement call, not a measurement.
This does not make the numbers useless — it makes the method matter. A figure built on a named model, with its grid basis and boundary stated, is a defensible estimate you can compare against itself over time and act on. A figure quoted without those is not. That is why every calculator here names its model and links its methodology, and why the honest use of a screener is to size and prioritise, then measure the workloads that turn out to matter.
After your first number
A digital estimate tells you two things: the rough scale of your technology footprint, and which workloads dominate it — often cloud infrastructure and a device fleet more than the website everyone worries about. That is enough to prioritise. When you need figures you can report, digital emissions fold into the wider inventory in specific places: owned servers as Scope 2 electricity, purchased cloud and hardware as Scope 3 Category 1, and a device fleet across both purchase and use. The Scope 1-2-3 inventory aggregator and the ISO 14064-1 inventory calculator pull those lines together with the rest of the footprint under the GHG Protocol Corporate Standard.
If you are sizing a whole organization rather than just its digital side, the organizational & sector footprint calculators give a fast whole-entity baseline that already includes IT and energy as components.
Start with the workload most people underestimate — the website or the cloud bill — or read how the estimate is built before you run it.
It depends on page weight, how efficiently the site is delivered, and the carbon intensity of the electricity powering the servers and the viewer’s device. The website carbon calculator estimates it on the Sustainable Web Design model from data transfer and grid factors. Treat the result as a defensible per-pageview estimate for comparison and improvement, not a precise measurement — the boundary and grid assumptions move it significantly.
A single generative query’s footprint is small but non-zero, and published figures vary widely because they depend on the model size, the hardware, and the grid running it. The AI search calculator estimates a per-query figure on stated assumptions so you can see what drives it. For a model overall, training is a large one-off, but cumulative inference across many users usually outweighs it — the AI compute calculator shows both.
It depends how many times the content is watched. Downloading once and rewatching avoids repeated transfer, so it can be lower for content you return to; streaming is often lower for something watched once. Resolution matters more than either — dropping from 4K to HD cuts data transfer sharply. The streaming calculator makes the trade-off explicit rather than giving a single verdict.
Owned on-premises servers are your Scope 2 electricity; rented cloud is usually Scope 3 purchased services. The cloud compute and storage calculators allocate emissions by utilisation and volume, and the data-centre PUE calculator adds facility overhead. Cloud is not automatically lower-carbon — it depends on the provider’s grid and efficiency — but reporting it in the right scope is the first step to comparing the two fairly.
Software Carbon Intensity, defined in ISO/IEC 21031, measures the carbon emitted per unit of useful work a piece of software performs — per user, per transaction, per API call. It is a rate, not a total, so it lets engineering teams track whether an application is getting more or less carbon-efficient as it scales. The SCI calculator applies the standard’s formula to your workload.
Three reasons: the electricity grid factor (region and location- vs market-based basis), the system boundary (whether a figure includes your device, the network and always-on infrastructure), and how shared resources are allocated across users. Different published models draw those lines differently, which is why a per-pageview or per-query number can differ by an order of magnitude. A figure is only meaningful alongside the model, grid basis and boundary behind it.
It depends entirely on the blockchain’s consensus mechanism. Proof-of-work chains like Bitcoin consume large amounts of electricity, tracked by the Cambridge CBECI. Proof-of-stake chains use a tiny fraction of that, so networks that switched cut their footprint by orders of magnitude. The NFT and crypto calculator asks which chain you are on, because that choice dominates the answer far more than the number of transactions.
Factors are maintained in the MasterBrain data layer and updated as the underlying sources publish — annual grid intensities, updated efficiency and model-energy data, and revisions to the Sustainable Web Design and CBECI models. Every value is versioned, so a calculator always shows which factor vintage it used, and each tool’s methodology page records the basis.
The rest of this topic
9 more pages in this topic that are not linked above.