Every action that we take online has physical consequences. While sending an image over text may seem less material than printing it out on a piece of paper and shipping it via a petrol-fuelled car, the digital process is entirely connected through infrastructure.
As society demands more digital, then the physical impact is even greater. For instance, a higher need for connectivity means sending more satellites into space and installing more cables underground.
An increased need for digital technologies such as crypto and AI means more data centres, and, the demand for a flashy new device entails more manufacturing facilities and more supply chain infrastructure.
While the digital world transforms, it has also led to a greater energy demand, a higher clean water demand, and the desire for more land to build data centres – something, Peter Campbell, director of green software at digital transformation consultancy Kainos, does not believe there is enough awareness of.
To address this, he’s written a book Digital Sustainability: The Need for Greener Software which unearths the environmental costs of digitisation.
“A couple of years ago, Kainos started a corporate net zero programme and made very aggressive targets, but we weren’t really dealing with anything to do with sustainability for our customers and we felt there was a bit of a gap there,” Campbell tells TI.
“Our business is building software for our customers, and yet we’ve got no thoughts on sustainability and no voice or views on this subject , and we’re not helping our customers do this.”
So, 18 months ago, Kainos created Campbell’s role of director of green software to tackle a lack of awareness and to write a book on exactly how digital software is impacting the environment, and what IT professionals can do to address this.
“The numbers are kind of staggering whenever you understand the amount of e-waste we generate every year, and this is increasing,” he says.
“The bigger issue is in developed countries where we buy more electronic goods, we throw away more electronic goods, and we change them on a more regular cycle.”
Europe is the highest continent for e-waste, and the issue is only mounting. In our personal lives, drawers fill up with old tech that doesn’t see recycling centres, and even when it does, the amount that can be recycled is very little.
“We’ve got this rising curve of demand, and yet, we’re doing very little to recycle that back or even repair the devices that we can.”
Campbell admits that for businesses calculating the amount of energy and e-waste is complicated. “It’s difficult for businesses because most don’t own their own data centres, they don’t own infrastructure for a lot of their IT.”
Additionally, while businesses can track their in-office e-waste, most are unaware of how the data, cloud, AI, and other digital technologies have an environmental impact – and with the rising popularity of AI, this is only going to get worse.
“For me, both crypto and GenAI are the poster child examples of IT substantially increasing both the carbon emissions and the wider environmental cost, particularly over the last five years,” says Campbell.
“There are a lot of academic studies that have tried to estimate this, but it’s hard at a macro level.”
According to the International Energy Agency, data centres and transmission networks each account for up to 1.5% of global consumption, and, the rise of AI is set to make this even higher.
According to the same study, simply training an AI model is estimated to use more power than 100 households in a year.
“GenAI’s just increasing as far as we can see, and the many billions that are being pumped into that by internet giants is a testament to that,” says Campbell. “Of course, it has so much potential, but the environmental cost is almost entirely, and intentionally, hidden from us.”
Gartner says that by 2025, without sustainable AI practices, AI will consume more energy than the human workforce. And, as Campbell details in his book, academic studies estimate that at the current rate of growth, AI could consume 83-134 TWh of electricity by 2027, the same as a small country.
On top of this, billions of litres of clean water are consumed per year by AI models. This is because the data centres that power LLMs guzzle energy and are running at a constant, meaning they are susceptible to overheating and malfunctioning. Clean water is needed to keep them running at safe temperatures.
But all this comes at a cost: it’s speculated that water withdrawal for AI could reach 4,200 billion litres to 6,600 billion litres in the next three years – half the water needed by the United Kingdom.
This, however, remains as estimations and speculations, and Campbell says that more transparency is needed to get a true understanding of the environmental impact so that suppliers and consumers can choose to use AI with the knowledge of these implications.
“Google, I think without a doubt, is the best path that all the others could follow,” says Campbell.
“It lists all of its data centres, and it says this specific data centre in Ireland uses this much carbon-free energy, versus this data centre in Israel, versus Singapore, versus Oregon, versus whatever,” he explains.
Then, if a software company chooses to deploy a particular AI job, for instance, to a country like Finland because it may involve lower water use, higher in renewable energy, and lower in waste compared to other data centres.
“Therefore, it’s not really a financial cost choice, it’s an environmental cost choice.”
Originally, Campbell explains that the book started with discussing carbon emissions, before realising that digital sustainability is broader than just carbon accounting.
“Carbon accounting gets most of the headlines because it is important to begin with, but it shouldn’t be an endpoint because sustainability is much wider than that,” says Campbell.
“As we all know, sustainability is also about biodiversity, it’s about deforestation, and it’s about the people who live in different areas.”
Carbon accounting relies on data explaining how much money a firm has paid another company for goods or services, and how many litres of fuel or kilograms of material a firm has bought to calculate how many emissions have been made.
What it doesn’t cover is other aspects such as water usage, the environmental impact on the local area, and the effect on biodiversity.
“Digital sustainability needs to speak to the wider sustainability agenda as well,” says Campbell. “So not decarbonisation of software, but getting people to ask their providers in their supply chain: What’s the water use from this? How does it impact the local environment around you and biodiversity? And even going deeper than that into mining questions.”
“There’s so much information there that could be provided that would help us understand the problem,” says Campbell.
Still, AI has potential to save productivity gains, and contribute to better efficiency.
“Whether it’s smart farming, monitoring water leakages from utility companies, or optimising energy usage. There are huge opportunities, but it’s a balance between value and cost, financial and environmental,” says Campbell. “
Ultimately, Campbell hopes the book will inspire people to improve their digital sustainability and encourage providers to become more transparent and allow their users to balance their digital carbon footprint.
For instance: “if everyone in the EU all kept our smartphones for one year extra, that is the equivalent of taking 2 million cars off the EU roads,” says Campbell. “That act in itself is probably the biggest single change we can make to reduce manufacturing of new devices and recycling devices as well.”
Then, it can be as simple as reducing the brightness of our screens, or not having three screens on at once if it’s not needed, reducing our storage and ‘dark data’, which Campbell says can make a significant difference.
“We can look at e-waste, that gets collected with the bins every couple of weeks. But if we start to think about compute ways and data waste and data mountains in the same sense as actual waste, that transforms the choices around computing and devices and our own personal behaviours.”
– Software has a larger environmental footprint than the aviation industry.
– It costs OpenAI $700k a year to run its services such as ChatGPT.
– ChatGPT “drinks” a 500ml bottle of water for a simple conversation of roughly 20-50 questions and answers.
– The IEA estimate that data centres in Ireland will make up for a third of the country’s national grid supply by 2026.
Subscribe to our Editor's weekly newsletter
Newsletter
Share
