Australian companies could slash carbon emissions created by AI tools, a study has found, by using more efficient, cloud technology and powering it with renewable energy.
Amazon released the findings from an Accenture study on Wednesday, which also estimated companies could save as much as 94 per cent of greenhouse gas emissions by using different technology.
The analysis comes amid rising concerns about the impact of artificial intelligence technology on the environment, with one expert telling a federal inquiry that generative AI queries used up to 90 times more energy than web searches.
The Accenture report, which analysed Amazon Web Service (AWS) technology, found local businesses could cut 24 per cent of carbon emissions from intensive computer processes by using more efficient, cloud-based hardware, and could cut 31 per cent of emissions using technology with better power and cooling efficiency.
Another 39 per cent of greenhouse gas emissions created by AI processing could be saved using cloud technology powered by wind and solar energy, it found.
AWS environmental, social, and governance global head Jenna Leiner said she hoped the study would give clarity to the rising number of businesses figuring out how to deploy AI tools without creating an environmental mess.
"In Australia, organisations are scaling the use of AI, and more and more we're seeing a conscious outlook on the environmental impact of doing so," she told AAP.
"Because it is early days, it is hard to navigate."
Despite significant potential savings outlined in the research, Ms Leiner said each business would make different carbon emission cuts by moving AI tools and data from their own computers to the cloud, and would need customised plans to achieve the greatest savings.
US biotechnology firm Illumina cut emissions by 89 per cent after moving its processes to the cloud, she said, while NAB reduced its emissions by 40 per cent.
"It's going to be a range depending on how savvy they are with their (computing) architecture," she said.
"We could build the most sustainable washing machine but unless we're telling our customers not to wash one T-shirt at a time, we haven't finished our job."
The study follows widespread criticism about the environmental impact of generative AI technology and its high energy consumption.
Google last month revealed its greenhouse gas emissions had risen by 13 per cent in 2023, and its target of reaching net-zero by 2030 would be challenged by "uncertainty around the future environmental impact of AI".
RMIT University senior lecturer Ascelin Gordon recently told Australia's federal inquiry into adopting AI that the technology had the potential to raise carbon emissions unless it was handled carefully.
"A single query on ChatGPT is something between 10 to 90 times more energy intensive than just a Google search, and if we generate an image that's probably 20 times more energy intensive again," Dr Gordon said.
"When we think there's now millions or possibly billions of queries on these AI platforms every day, the energy use is quite staggering."
