Companies are just beginning to recognise the impact of data consumption on their carbon footprint. While software giants like Microsoft and Google are ahead of the trend – both say their data centres are carbon-neutral – there will be increased pressure on all kinds of corporations to make data-related energy use a higher priority in CSR efforts as big data continues to skyrocket. As Chief Data Officers navigate their way through new processes, these are some of the key topics to review:
Work to a Recognised Standard
There are many calculators for carbon consumption, with just as many methodologies. Make sure you use one that works to a recognised international standard. The results might not be easy reading, but it’s most important to develop a comparable picture of your data-related carbon footprint.
Resist the urge to tick boxes
Too many of these exercises are designed to show an existing positive picture, which may be misleading. Be prepared for real change.
Know your Scope
Understand what areas of your business constitute Emissions Scope 1, 2 or 3 under the global Greenhouse Gas Protocol. Broadly, Scope 1 emissions are direct emissions from company-owned resources, Scope 2 are indirect emissions from generating energy, and Scope 3 are all other indirect emissions linked to the company’s operations.
Your Scope 3 emissions are still emissions, and they are still part of your business
A common approach is to follow the guidelines and move data operations to a third party in the cloud, so they can be Scope 3 emissions. But they are still emissions caused by your business, without which your data operation wouldn’t, well, operate. Beware the temptation to throw the rubbish over the wall, and hope the neighbours sort it out!
It’s not just carbon (Li, Dy)
Data’s climate impact is greater than just carbon emissions. Lithium is a key component in rechargeable batteries, and several rare earth metals are used for computer manufacture (Dysprosium, for example). These involve considerable mining and processing (for which, read pollution) to isolate in the first place, before they can be used. Work with your computer and data centre suppliers to understand this supply chain and its impact.
It’s not all about being better than the alternatives
A lot of the discussion about digital emissions focuses on what the alternative carbon cost would be to achieve the same task (for instance, how a Zoom meeting reduces the carbon footprint of travel associated with an in-office meeting). But with many businesses moving to hybrid working, this could end up being the worst of all worlds (Zoom meetings held in the office).
Water, water (not) everywhere
One of the most obvious ways data centres impact on the environment is through the use of water cooling systems. These can be very energy intensive, and so have a carbon footprint, but also a water footprint. Data centres are often tucked away in more remote locations, where access to water can be a challenge itself. Where does the water for your data centres come from? Does it deprive other potential users? How is that water reused? Is it a long-term sustainable source?
Where are your servers?
Many data operations have functions in different locations and countries, where human capital is cheaper. However, this can mean a more emission-heavy process, so the net impact versus cost should be considered.
Have you considered all the stages of your AI lifecycle?
You will typically know all the stages of your data supply chain, and where these activities take place, and so consider the emissions for these. For example, training data can often be split into microtasks and either outsourced or crowdsourced. This often results in a distributed network across various countries. These should be considered as Scope 3 emissions.
Recycle, recycle, recycle
Many of the components in computing, particularly the rare earth elements, are finite resources, with some estimates suggesting there is only enough in natural deposits for a few decades’ global supply. Moreover, often computers are deemed too difficult to recycle, and after decommissioning, are thrown away, or sent to specialist landfill, never to be reused. These elements can be hard to recycle, but that doesn’t mean you shouldn’t try. Work with suppliers to find the best recycling options for your decommissioned data assets.