Artificial intelligence might be the most irresistible technology ever offered to humanity, but Google has admitted to a 13 per cent increase in its greenhouse gas emissions and increased by 48 per cent over the past five years, fuelled by current consumer reliance on AI in all aspects of life.
The global giant estimated in its recently released annual environmental report for 2023, that its data centres consumed more than 24 terawatts an hour, equivalent to 7 to 10 per cent of all electricity consumed by data centres worldwide.
Despite much of the report discussing how the use of AI had helped the rest of the world with decarbonisation, footprint reduction and other aspects of sustainability, the report admits that responsibly managing the resource consumption of AI had been a “challenge”.
The report says that the electricity consumption of its data centre had grown by 17 per cent, but it is still maintaining its commitment to a 100 per cent global renewable energy match. The intensive electricity usage is also expected to continue “as Google’s infrastructure continues to power the digital transition.”
“We see our growing infrastructure as an opportunity to drive the innovations and investments needed to power a low-carbon economy,” the report said.
Chief sustainability officer Kate Brandt told GreenBiz that the rising energy use was “because AI is becoming so deeply integrated into our products” and that “the distinction between what is an AI workload and what is not an AI workload is kind of not meaningful.”
The company had committed to running “24/7 on carbon free energy by 2030” in 2021, but it reports that growing demands of its data centres and offices have kept emission free energy at the global average of 64 per cent since the last year.
It also reported a total greenhouse gas emission of 14.3 million tonnes of carbon dioxide equivalent in 2023, a 13 per cent year over year increase, and a 48 per cent increase compared to 2019 “due to increases in data centre energy consumption and supply chain emissions”.
Proposed mitigation strategies
While Brandt and co-author Benedict Gomes, the senior vice president of learning and sustainability, writes there were “uncertainties ahead AI” regarding the company’s ability to mitigate its footprint, they “remain optimistic about AI’s potential to drive positive change”.
“We’re also clear-eyed about its potential environmental impact and the collaborative effort required to navigate this evolving landscape.”
In order to manage the environmental impacts caused by its AI, the organisation has flagged three reduction strategies:
- model optimisation: through research and deploying a new model that can train an AI up to 100 times faster, reducing emissions by up to 1000 times. The company is also releasing its Go Green software guide to help developers using AI to reduce their digital footprints
- efficient infrastructure: improving the energy efficiency of both its data centre and its hardware chip products offered to customers as well as improving water usage and diversion of its operational waste towards circular economy
- emission reduction: the company has recently signed a contract for 25 terawatts an hour to push towards its goal of net zero by 2030. It is currently negotiating with the supply chain to improve decarbonisation road maps in high impact areas. Among these are also the company’s ambitions to out-innovate the rising demand through new energy efficient technology
“My message is, this is hard,” said Brandt. “This is the decisive decade; we need to stick with this.”