Measuring greenhouse gas emissions in data centres: the environmental impact of cloud computing

May 3, 2022
Measuring greenhouse gas emissions in data centres: the environmental impact of cloud computing

This is our first post in a series of four articles dissecting how to measure greenhouse gas emissions from operating data centres and cloud services. Today’s post explores the drivers behind the emissions of data centres and provides a detailed benchmarking, ranking more than 60 locations around the world in terms of their carbon intensity across major data centre providers like AWS, GCP, and Azure. 

Check out subsequent posts in this series to learn about A) the role of the big three, namely AWS, Azure, and GCP, B) how to assess emissions from your specific computing activities, including on-prem data centres and hybrid set-ups, and C) how to lower your cloud computing footprint. Sign up for our free newsletter to stay up to date on new articles.


In public debate, emissions from cloud computing are often overlooked, despite having gained prominence due to reports about the massive CO2 footprint of popular activities like watching Netflix or trading crypto assets.

  • According to the Paris-based nonprofit The Shift Project, emissions generated from watching 30 minutes of Netflix sum up to about 1.6 kg of CO2, equal to driving almost 6 kilometres (roughly 4 miles) in your car. However, more recent research shows that the true output should be substantially lower by now.
  • Bitcoin has a carbon footprint comparable to that of New Zealand, producing 36.95 MTof CO2 annually, according to Digiconomist. As evidenced by Cambridge University, cryptocurrency consumes more electricity than the entire annual energy consumption of the Netherlands. 

When considering all cloud computing emissions, selected comparisons become even more staggering.

Global emissions from cloud computing range from 2.5% to 3.7% of all global greenhouse gas emissions, thereby exceeding emissions from commercial flights (about 2.4%) and other existential activities that fuel our global economy.

Cloud computing providers must take action

Consequently, cloud computing providers are called upon more than ever to take action. However, they often struggle to determine how best to lower emissions, given credible emissions data for their operations are unavailable, inconsistent, or lack transparency. 

This is particularly true for AWS but also other hyperscalers, such as Azure and Google Cloud, that recently began sharing their cloud calculators with some clients, though not all. 

At Climatiq, we see increased interest from forward-thinking organisations striving to better understand emissions from their computing operations. 

To make things clearer, we start by explaining the main underlying drivers of computing emissions. Next, we empower companies to contemplate tactics to lower their emissions.

(Quick reminder: Sign-up for our newsletter to find out more about best practices for lowering emissions in forthcoming articles of our four-part series.)

Let’s start with the fundamentals of the drivers of emissions data.

What causes emissions of data centres?

The carbon footprint of a data centre is affected by three factors:

  1. electricity consumption (to run the servers)
  2. water consumption (to cool the servers) 
  3. and the lifetime of the equipment (which impacts the frequency of replacements).

All three variables have unique features that need to be considered when doing any form of calculations.

Starting with electricity consumption, a key metric to consider when calculating energy usage is the Power Usage Effectiveness (PUE) value: 

  • A PUE value of 1.0 indicates that all energy consumed by a data centre is used to power actual computing devices, a.k.a IT infrastructure. 
  • The best data centres in the world achieve a PUE of 1.2. 

Many calculators out there ignore this variable. Climatiq’s cloud computing emission factors take PUE into account when estimating the emissions of cloud services. 

On the water consumption front, it’s important to consider that data centres are among the top-ten water-consuming industries in the US, according to credible research. However, cloud providers usually do not report water consumption (and corresponding water stress levels), which is frightening, to say the least. 

Climatiq’s Director of Science Maryam Arbabzadeh co-authored a paper that calculates the median water footprint at no less than 0.74 litres per GB of internet use (for networking), which suggests that water consumption impact is relevant and urgently needs to be considered when calculating emissions.

For server lifetime, companies face similar challenges with existing cloud provider reports. Most of them neither share nor consider respective server lifetimes. However, servers usually get replaced every 3 to 5 years (longer for local data centres), leading to spikes in emissions, which urgently need to be incorporated. 

Climatiq follows a more granular approach

By comparison, Climatiq explicitly considers all of these specifics above as best as possible by following methodology developed by CCF, and made available through the Climatiq API.

To be exact, the following factors are taking into account when determining the carbon footprint of cloud infrastructure:

  • The amount of CPU, Networking, Memory, and Storage your services use.
  • The utilisation rate of machines in your data centres.
  • The efficiency at which your data centres are cooled and the usage of electricity, generally specified as Power Usage Effectiveness or PUE.
  • The location of your data centres to determine the electricity grid mix. 

Ranking of data centre types

Generally speaking, hyperscalers and co-located data centres are more efficient (including accounting for water consumption), driven by better energy utilisation, more efficient cooling systems, and increased workloads per server. 

Computing workloads in hyperscale data centres are almost six times more water-efficient than internal data centres. 

Based on the methodology developed by CCF and using Climatiq’s comprehensive emissions database, we have designed a heatmap to rank data centres across more than 60 locations around the world according to their emission intensity. Locations can be filtered by major cloud providers (AWS vs. GCP vs. Azure) and are ranked by CO2 emissions output per CPU-hour. 

We’d love to hear from you if you have any questions regarding our methodology or wish to know more about the data - please get in touch with us here. 

Tags