AI is increasingly finding itself in the crosshairs of investors concerned about its environmental toll. Many are coming to terms with the fact that the technology supposedly leading us into a smarter future might be dragging the planet backwards when it comes to climate commitments.
Big tech’s climate ambitions are hitting a rather inconvenient truth: AI-driven data centres are guzzling energy at an alarming rate.
Despite a pledge to reach net-zero emissions by 2030, Google’s greenhouse gas emissions have ballooned by 48 per cent over the past five years, with a 13 per cent jump in 2023 alone, bringing total emissions to 14.3m metric tons.
“The primary contributors to this surge are the electricity consumption of data centres and supply chain emissions,” said analysts at Shore Capital in a report on AI, released today.
As AI adoption explodes, the energy required to power these advanced models, fuelled by oceans of data and endless hours of training, has surged in tandem.
The International Energy Agency forecasts that data centres’ electricity consumption could double by 2026, with AI potentially gobbling up 4.5 per cent of global energy by 2030.
Currently, data centres account for a whopping 21 per cent of global electricity consumption – up from just five per cent in 2015.
A single response from ChatGPT consumes up to ten times the electricity of a typical Google search by some estimates, with others saying even more.
Google and OpenAI aren’t alone in this. In 2022, water consumption skyrocketed for Microsoft, Google, and Meta, with increases of 34 per cent, 22 per cent, and three per cent, respectively.
These companies are cranking up water usage to cool their ever-growing, ever-thirsty data centres, a trend tied to the booming demand for AI.
And the financial markets have noticed. In April, Meta, the parent company of Facebook and Instagram, saw its shares plunge over 15 per cent after announcing higher than expected AI expenses.
Read more
Is there an AI bubble and will it burst?
Earlier this month, hedge fund Elliott Management slammed AI technology as “overhyped” and warned that tech darling Nvidia is in “bubble land”.
Combine high valuations with rising energy bills and spiralling operational costs, and it becomes easy to see why investors might be starting to wonder if all this AI spending will deliver the returns they’re banking on.
Proponents argue that AI could itself be the quick fix to the energy conundrum, potentially discovering new sources of renewable energy and ways to improve efficiency.
But this vision comes with a steep energy bill, putting tech companies in a tight spot as they try to juggle their green commitments with AI’s ravenous energy appetite.
Solutions are emerging though. In Iceland, a new data centre powered entirely by renewable energy claims to have slashed overall energy usage by 38 per cent and cut consumption by up to 50 per cent during off-peak hours.
The project, led by Vesper Technologies, Sardina Systems, and Borealis Data Centre, offers a “cost-effective solution for enterprises looking to scale their operations responsibly,” according to Kristófer Andri Kristinsson, product manager at Borealis.
In the UK, a company called Deep Green heats swimming pools using excess heat generated by data centres. It plans to expand across more UK sites after a recent £200m investment from Octopus Energy.
Edge computing is also being touted as a possible turning point. Ryan Cox, co-head of AI at Synechron, said that, “by moving computation closer to the user, devices like smartphones and cars can handle more processing locally, reducing reliance on energy-hungry data centres.”