蜜芽传媒

As Britain weans itself off fossil fuels and we push towards electric cars on the road, are we about to undo all the good work with our phones and computers?

The growth in the use of artificial intelligence has resulted in vast quantities of data that needs to be processed and stored. This has led to the construction of giant data centres all around the world. Each comes at a significant cost to the planet – they require vast amounts of energy to run and a huge supply of water to cool.

For Rare Earth, Tom Heap and Helen Czerski look at the environmental impact of AI. They take a deep dive into the research around more efficient data centres, designed to exploit renewable energy resources or cool themselves naturally with seawater. And they ask, what are the practical tips we can all take to reduce our own digital carbon footprint?

We’re creating vast amounts of data, explains Tom Jackson, the Professor of Information and Knowledge Management at Loughborough University Business School. By the end of this year, we will have created 175 zettabytes: “the equivalent of one individual watching high-definition Netflix for 6.8 billion years.”

The data centres required to store and route it all are, essentially, “computers in sheds,” says Tom. “Millions of computers, just like the ones we use but formed as servers, in racks in huge barns.”

The scale is vast, he stresses; in some cases, breath-taking. When it comes to the “hyperscalers” – Google, Microsoft, Amazon – these sheds can be as large as multiple football pitches.

2. The big data centres’ energy consumption is estimated to double over the next five years

These data centres require an immense amount of energy. “They are consuming anywhere up to 2% to 4% of the world's electricity supply,” says Tom. And, as AI becomes more complex, this figure is rising at an incredible rate.

The International Energy Agency estimates that the electricity consumption of the big data centres will double between now and 2030. By then, it could equal that of Japan today.

The Digital Decarb Design Group at Loughborough produced the Data Doomsday. “Our forecast suggests global energy supply from renewable energy sources will be unable to meet demand from data centres by the end of this year, 2025,” states Tom. He predicts that by 2033, there may not be enough global electricity produced to power all the data centres in the world. “Something has to give.”

The big tech companies know their need is growing. Microsoft has signed a deal to restart the infamous Three Mile Island Nuclear Generating Station (the location of one of America's worst nuclear accidents). Amazon are investing in a new generation of small, modular nuclear power stations.

4. Data centres generate up to 3.7% of global greenhouse gas emissions

Every time you send an email, you are creating a carbon cost, explains Tom. Bulk it out with photos and files, and the amount of carbon you use increases.

How does the overall carbon cost of our data compare with other large energy consuming areas like aviation or shipping? “2.5 to 3.7 % of global greenhouse gas emissions are generated from data centres.” Aviation is at 2.1 to 2.4%.

5. Every chat GPT query you make uses one pint of water

Data centres also need vast amounts of water to cool their overheating microchips. “Computers are in essence just incredibly efficient electric heaters,” explains Tom. “For every chat GPT query you make, it takes one pint of water to cool the system.”

Many older data centres are incredibly inefficient. For every pound spent powering the computer, they’re spending another £1.20 keeping it cool.

The water consumption involved is a real source of concern for the communities hosting data centres. Aurora Gomez Delgado founded the group Tu Nube Seca Mi Río, which translates as Your Cloud is Drying My River. She’s campaigning against what she sees as the reckless building of data centres in some of the driest places on the planet, including her native Spain where water is in short supply. “We have to explain that Castilla-La Mancha is a very dry region,” she explains. “Even our name means dry in Arabic.”

In 2018, Microsoft ran a trial where they sunk one of their data centres into the cold water of the Orkney Islands. “It's a lot more efficient to cool in water than it is on land,” says Ben Cutler from Microsoft Research. “Whereas on land we might use 15 or 20%, or even more, of the energy for cooling, it's almost zero when we are in the water.”

Could the future be data processing at the site of renewables – underwater? There have been ideas about putting data centres inside wave energy machines for many years, says Neil Kermode of the European Marine Energy Centre. “As a land-based animal we tend to think we'd need to bring electricity on to shore and then do something with it, but maybe we actually do the data processing at sea.”

Other propositions include putting data centres in the deep sea or the Arctic – but these haven’t taken off. “They run into real-world barriers of cost, control, complexity and carbon accounting,” says Tom. Being so remote would make security, maintenance and repair difficult, and could also lead to unwelcome delays: “If you move it further off grid or into the sea, there's a longer time for the data to move from A to B.”

7. Data centres are being used to heat swimming pools

“The amount of money and energy and time we’re spending building data centres is the same amount of money, energy and time we need to electrify our heating systems,” suggests Mark Bjornsgaard, CEO of Deep Green, a company that aims to create more efficient data centres. “All you need to do is build data centres where you need the heat and if you do that, this huge existential challenge becomes a massive opportunity.”

Heating swimming pools is a brilliant way of exemplifying this approach, he states. If you plug a data centre into the heating system of a swimming pool you can do all of the cooling that the data centre needs whilst very efficiently heating the pool.

And then there are large, centralised systems that we can use to heat areas of our towns and cities. “Microsoft and others have plugged data centres into district heating systems already,” says Mark. Deep Green hope to do the same across the country

Currently, big tech is not being held to account over energy use. “It's difficult to find clarity,” says Tom. “It's also difficult to find clarity in terms of the carbon and their sustainability. Very difficult to compare apples with apples between a Google and a Microsoft data centre.”

In the same way that we can look at red, amber and green ratings on food packaging, wouldn't it be great if we had something similar for the tech products we’re using? When choosing social media platforms, for example, we could look at “the embedded carbon that goes into providing that service,” suggests Tom.

9. Deleting “dark data” can reduce our own digital carbon footprint

How can we, on an individual level, help to reduce the environmental cost of data?

Firstly, there’s using AI less – but this isn’t always easy. “I don't see the solution is trying to abstain from this when, frankly, we're just putting ourselves in the way of progress,” argues Mark.

There’s undertaking a digital decluttering. Many of us store photos and documents in the cloud that we’ll never access again. In fact, 55% on average is so called “dark data”, explains Tom. So why not do a digital clear out? There’s also the option of storing your photos on an external hard drive.

When buying equipment, boycott items with unnecessary sensors, that send unseen and unused data into the cloud, says Tom. “It's us, the end user, that want to use social media. We have embedded sensors. So, we’ve got to start pushing back,” he argues.