The Carbon Footprint of Data Centres: Our Guide

Why the carbon footprint of data centres matters (and why it’s growing)

Most of the internet feels weightless. But every email you send, file you store, video you stream, model you train, or dashboard you refresh runs on physical infrastructure: data centres.

Right now, data centres account for a relatively small share of global carbon emissions, but the trend is sharply upwards.

The IEA estimates data centres used ~1.5% of the world’s electricity in 2024 (415 TWh). This highlights the massive scale of data center energy consumption worldwide.
That demand is expected to more than double by 2030 (to ~945 TWh), with AI a major driver.
Using IEA data, Carbon Brief puts data centres at ~0.5% of global CO₂ emissions today. These emissions can vary significantly depending on location, as the carbon emission intensity of electricity differs by region.
Indirect carbon emissions from electricity use make up a significant portion of the total carbon footprint of data centres.

Cloud computing providers such as AWS, Azure, and Google Cloud are key players in the growth of digital infrastructure, driving increased data centre energy demand and associated emissions.

Even if the global percentage looks “small”, the impact can be very local: data centres concentrate huge demand in specific grid regions, which can push up fossil generation at peak times and delay grid decarbonisation if capacity can’t keep up. The carbon footprint of a data centre is heavily influenced by the local electricity mix, as regions with a higher share of renewables will have a lower carbon impact. The electricity mix - whether dominated by renewables or fossil fuels - plays a crucial role in determining the environmental impact of data centre operations.

What is a data centre?

A data centre is a specialised facility built to run servers (compute), storage systems (data), and networking (connectivity) at scale, reliably and securely, 24/7.

Think of it as the industrial engine room of the internet:

Servers run websites, apps, databases, and AI workloads.
Storage systems house everything from business files to streaming libraries, supporting essential digital services and contributing to the energy consumption and carbon emissions of data centre operations.
Networks connect data centres to each other and to end users.
Power + cooling keep it all stable: servers generate heat, and downtime is expensive.

Although you can build your own data centre in-house, most organisations don't - instead they rely on:

Colocation (renting space and power in someone else’s facility), or
Cloud providers (AWS / Google Cloud / Azure), where you rent compute and storage as a service, and cloud software services built on these.

Where do a data centre’s greenhouse gas emissions come from?

A data centre’s carbon footprint has two big components which represents the overall environmental impact of a data centre:

Operational emissions (running it day-to-day: electricity, cooling, backup power): This includes the carbon emissions emitted from daily operations, such as powering servers and maintaining cooling systems. Electricity related emissions are a major part of this, as most data centres rely heavily on grid electricity. Fossil fuel use, such as diesel generators for backup power or fossil-fuel-based grid electricity, also contributes significantly to operational emissions.
Lifecycle / embodied emissions (the carbon in the building materials, servers, batteries, and equipment manufacturing + replacements): These emissions come from the manufacturing processes involved in producing hardware, infrastructure, and equipment, as well as their eventual replacement.

Emissions are typically measured by multiplying energy consumption by emission factors, which convert energy use into equivalent CO2 emissions.

The IEA explicitly includes both operational and embodied emissions when discussing the wider “data centres + data transmission networks” footprint. The total greenhouse gas emissions from both sources make up the overall ghg emissions of a data centre, reflecting the full climate impact.

Scope 1, Scope 2, Scope 3: what these categories mean

Most organisations report emissions using the GHG Protocol, which groups emissions into three scopes:

Scope 1 (direct): emissions from sources you own or control (e.g., on-site fuel use).
Scope 2 (indirect energy): emissions from the generation of purchased electricity/heat/cooling you use.
Scope 3 (value chain): other indirect emissions up and down your value chain (suppliers, hardware manufacturing, waste, etc.).

What Scope 1–3 look like for a data centre

Breaking down the carbon footprint of a data centre into the Scopes is fairly straightforward:

Scope 1 (direct) examples

Diesel or gas burned on-site (often in backup generators during testing or outages)
Refrigerant leakage from cooling systems (refrigerants can have high global warming potential)

Scope 2 (purchased electricity)

The big one for most facilities, this is electricity consumption for:

IT load (servers, storage, networking)
Cooling and air movement
Power conversion losses (UPS, transformers)
Lighting and building systems

Scope 3 (value chain)

‍Often the most underestimated, and, in many cases, one of the largest over the full lifecycle:

Manufacturing of servers, GPUs, batteries, racks (and the mining/refining behind them)
Construction materials and fit-out (steel, concrete, cabling)
Upstream fuel-and-energy impacts
Waste handling and end-of-life treatment (e-waste)
Employees commuting to and from the site

How to measure data centre emissions (and why it’s tricky)

Measuring emissions is straightforward in principle:

Activity data × emissions factor = emissions (tCO₂e)

…but data centres create some real-world complications:

Electricity consumption is shared across thousands of devices and tenants.
Cloud services abstract away the physical layer.
The same electricity can be counted in different places depending on reporting boundaries (your Scope 3 may be your provider’s Scope 1/2).
Water consumption is another important metric, as data centers can vary significantly in water efficiency and overall environmental impact.

How would a data centre measure its carbon emissions?

Start with three data streams:

Electricity (kWh)

Ideally sub-metered into IT load vs “overhead” (cooling, power distribution, lighting)

On-site fuels

generator testing fuel, emergency runtime

Refrigerants

inventory, top-ups, leak rates

Using this, you can calculate two key efficiency metrics:

PUE (Power Usage Effectiveness): Power usage effectiveness (PUE) = total facility energy ÷ IT equipment energy. It’s defined in ISO/IEC 30134-2. A lower PUE means less overhead energy for cooling and power conversion per unit of compute delivered.
Carbon intensity metrics: E.g., kgCO₂e per kWh, per rack, per compute-hour, or per workload, all useful for tracking improvements over time.

Finally, account for your Scope 3 - especially IT equipment, which can form the majority of the carbon footprint of a data center. This might look like:

Using Environmental Product Declarations (EPDs) or Product Carbon Footprints (PCFs) to assess the carbon footprint of IT equipment.
Analysing the weight and type of materials used in construction.
Measuring the weight of e-waste produced and where it ends up.

If you’re a “normal” business using cloud + SaaS

You usually can’t measure data centre greenhouse gas emissions directly - because you don’t own the facility, and your workloads share infrastructure with others.

What you can do instead is use provider-allocated emissions data, based on your usage.

Read our full guide to what cloud providers are doing to help measure and reduce their carbon footprints.

Microsoft 365

Microsoft provides an Emissions Impact Dashboard for Microsoft 365 that estimates greenhouse gas emissions from your organisation’s M365 usage (e.g., Exchange, SharePoint, OneDrive, Teams).

Azure

Microsoft also provides an Emissions Impact Dashboard for Azure, estimating emissions associated with your Azure usage.

AWS

AWS offers the Customer Carbon Footprint Tool, which estimates emissions associated with your AWS services and breaks them down by service and region.

Google Cloud

Google Cloud provides Carbon Footprint, showing location-based and market-based emissions derived from your Google Cloud usage, broken down by project/product/region.

A useful practical note: these tools are typically most helpful for trends and comparisons (what changed, what moved regions, what services are driving impact), rather than pretending you’ve measured the exact molecules of CO₂ linked to your specific workload.

A quick “quality check” for cloud emissions data

When you’re using dashboards (yours or a supplier’s), ask:

Are the numbers location-based, market-based, or both? (They can differ materially.)
Is the methodology transparent and aligned to widely used standards?
Does the estimate cover Scope 2 only, or Scope 2 + relevant Scope 3 (like hardware manufacturing)?
Can you see results by region and service, so you can actually act on them?

Our full guide goes into the detail on this for AWS, GCP and Azure.

Reducing data centre emissions: who can do what?

What the average business can do (realistically)

If you don’t run a data centre, you can’t change its cooling system or power contracts directly.

But you do have a few meaningful levers:

Choose responsible software and cloud partners

Prefer suppliers that publish clear carbon footprint reporting and can share cloud-related greenhouse gas emissions data (or at least credible methodology).
If you’re building on AWS/GCP/Azure, use their carbon footprint tools to inform decisions on architecture and region selection for your data centres.

Be intentional about AI usage

The biggest growth driver is accelerated compute. You don’t need to avoid AI - but it’s worth treating it like any other cost centre: measure, manage, and don’t run heavy workloads by default “just because”.

What data centres (and cloud providers) can do

This is where the heavy lifting sits.

1) Decarbonise electricity supply

Procure renewable energy in ways that genuinely shift the grid, like installing solar directly, or procurement from "gold standard" green energy providers (not just certificates).
The IEA notes that matching annual demand with renewable energy purchases does not necessarily mean a facility is powered by renewables 24/7, because timing and location matter.

2) Improve energy efficiency

Better cooling design (hot/cold aisle containment, liquid cooling where appropriate).
Reduce power conversion losses.
Keep pushing PUE down (and report it consistently).

3) Smarter siting

Build where grids are lower-carbon and where the climate supports efficient cooling (and where water impacts are manageable).
This is why you often hear examples like Iceland: not magic, just a combination of climate and low-carbon power availability.

4) Tackle embodied carbon

Longer server lifetimes where feasible.
Repair/refurbish and circular procurement.
Low-carbon materials in new builds.
Supplier engagement (because a lot of Scope 3 sits upstream).

5) Use waste heat

In some locations, data centres can feed heat into district heating networks or nearby industrial processes (site-dependent, but high potential where it fits).

The takeaway

Data centres are already a meaningful source of carbon emissions: ~0.5% of global CO₂ for data centres alone, and set to grow as electricity demand more than doubles by 2030.

If (like most businesses) you’re not a data centre operator, your best approach is:

Use cloud emissions tooling to get visibility,
Design and run software efficiently, and
Pick suppliers who take responsibility for the footprint they’re creating on your behalf.

At Seedling, if you're using cloud providers (or software built using them), we can help you to get the data you need and feed this into your wider carbon footprint.