Since the release of ChatGPT in November 2022, the world has seen an incredible increase in investment, development, and use of artificial intelligence (AI) applications. By one estimate, computing power used for AI is doubling roughly every 100 days.
The social and economic impacts of this boom are sparking reactions around the world: European regulators recently pressured Meta to suspend plans to train an AI model with users' Facebook and Instagram data, and the Bank for International Settlements, which coordinates the world's central banks, has warned that the introduction of AI could change how inflation works.
The environmental impact has received little attention so far: a single query to an AI-powered chatbot can consume up to 10 times more energy than a traditional Google search.
Roughly speaking, generative AI systems can use 33 times more energy to complete a task than traditional software requires. This massive energy demand could lead to a surge in carbon emissions and water usage, putting further strain on power grids already strained by climate change.
energy
Most AI applications run on servers in data centers. By 2023, before the AI boom really begins, the International Energy Agency estimates that data centers will already account for 1-1.5% of global electricity use and about 1% of global energy-related CO₂ emissions.
By comparison, the aviation sector accounted for 2% of global energy-related CO2 emissions in 2022, while the steel sector accounted for 7-9%.
How are these numbers changing with the rapid growth in AI usage? Recent environmental reports from Microsoft, Meta, and Google provide some insights.
Microsoft is investing heavily in AI, including in OpenAI, the creator of ChatGPT, and its own Copilot application for Windows. Between 2020 and 2023, Microsoft's reported annual emissions increased by about 40%, from 12.2 million tons of carbon dioxide equivalent to 17.1 million tons.
These figures include direct emissions as well as indirect emissions, such as those caused by generating the electricity used to run the data centers and those caused by the use of the company's products. (These three emissions categories are known as Scope 1, 2, and 3 emissions, respectively.)
Meta is also dedicating significant resources to AI: In 2023, the company announced that its Scope 3 emissions will increase by more than 65% in just two years, from 5 million tons of CO2 equivalent in 2020 to 8.4 million tons in 2022.
Google's emissions are set to increase by about 50% in 2023 compared to 2019. The tech giant's 2024 environmental report noted that planned emissions reductions will be difficult due to “increasing energy demands from enhanced AI computing.”
water
Data centers generate a lot of heat and consume a lot of water to cool the servers: A 2021 study found that U.S. data centers use about 7,100 liters of water for every megawatt-hour of energy they consume.
Google's US data centers alone consumed an estimated 12.7 billion litres of freshwater in 2021.
Data center water use has become a particular concern in regions where water stress is increasing due to climate change, and recent droughts in California, where many tech companies are based, have led companies such as Google, Amazon and Meta to launch “water positive” initiatives.
These leading technology companies have announced a commitment to replenish more water than they consume by 2030. Their plans include projects such as designing ecologically resilient watershed landscapes and improving community water conservation to increase water security.
Climate Risk
If your data center is located in or near a city, it may also compete with people for resources during times of scarcity, such as extreme heat waves.
Globally, the total number of days with temperatures above 50°C has increased every decade since 1980. July 2023 was the hottest month on record.
Extreme heat affects the health of local residents, with a Lancet 2022 study finding that a 1°C increase in temperature can lead to increased mortality and morbidity.
On a hot day, air conditioning can be a lifesaver: Data centers also need to stay cool, so power usage spikes as temperatures rise, increasing the risk of blackouts and grid instability.
What's next?
So what happens next? As we've seen, technology companies are becoming increasingly aware of this problem. How is that reflected in their actions?
A July 2023 survey of Australian sustainability professionals found that only 6% believed data center operators provided detailed sustainability data.
Earlier this year, we spoke to IT managers in Australia and New Zealand about how AI applications are impacting growing energy use and found that 72% are already deploying or piloting AI technologies.
More than two-thirds (68%) said they are concerned about increased energy consumption due to AI needs, but there is also a great deal of uncertainty about the magnitude of that increase.
And regardless of their company's sustainability efforts, many IT managers lack the necessary skills to adequately address these sustainability impacts. There is an urgent need for education and training to help IT managers understand and address the sustainability impacts of AI.
Gordon Noble is Research Director at the Institute for Sustainable Futures at the University of Technology Sydney (UTS) where he works on sustainable finance. Fiona Berry is Research Director at the Institute for Sustainable Futures at the University of Technology Sydney.
This article was originally published on The Conversation.
