Research shows data centres running artificial intelligence increase carbon emissions | 

Sarah Palmer, Staff Writer | 

When hearing the phrase ‘artificial intelligence,’ mental screenshots of ChatGPT’s landing page probably come to mind. From waking up to an alarm powered by Amazon’s Alexa to trusting Grammarly to carve the perfect email, it’s becoming harder to imagine a world without it.

Just as the internet and the advent of user-friendly computers helped expand the World Wide Web beyond scientists in the early 2000s, recent developments have made AI technology more accessible—sparking a rise in energy consumption and indirectly raising greenhouse gas (GHG) emissions.

In 1999, Jordan Kidney, a computer science professor at MRU started his undergraduate degree. At that time, the internet was something most people accessed at their local library, and AI was an unrealistic concept known only after watching sci-fi movies.

Hitting fast-forward on the VCR, Kidney worked as a research assistant for a professor who did AI work during the second year of his undergrad at the University of Calgary. Inspired by the opportunity, he completed his master’s degree in AI, machine learning and multi-agent systems in 2006.

Teaching everything from advanced computer programming to introductory-level science courses under the general education category, Kidney has been an instructor at MRU for the past 15 years. He says that as technology evolves, so does the way he approaches teaching. “Technology is something that helps us to manipulate the world around us,” said Kidney. “We’re used to physical information, but we’re in the world of digital that’s expanding more and more.”

But while digital media advances provide socioeconomic benefits, research shows that the rise in AI software development comes at the expense of a hefty carbon footprint—despite offering some clean energy solutions, according to the World Economic Forum.

Tech booms causing carbon doom

The prominence of wireless technology and feather-light personal devices has made it easy to forget where computing power comes from. Namely, Kidney finds that the ‘cloud’ has added to the impression that digital information is stored far, far away in the depths of the atmosphere.

“Anything that’s in the cloud is not in a physical cloud,” said Kidney. “It’s in a machine somewhere that’s storing information and those machines take power for transmitting.”

What he’s referring to are data centres which, in simplistic terms, are physical warehouses filled with an assortment of specialty computers. Powering everything from social media to search engines, data centres require high-speed internet and are where the ‘cloud’ calls home.

Although data centres are only directly responsible for one per cent of energy-specific GHG emissions, they’re powered by electricity grids that are supported by fossil fuels—amounting to a total contribution of 40 per cent of emissions, according to the International Energy Agency (IEA).

The Government of Canada reported a 24 per cent increase in global GHG emissions between 2005 and 2021. Since 2010, the global sum of internet users has doubled. It’s understandable that a number of nations pledged emission-reduction strategies in both the Kyoto Protocol and Paris Agreement around that time.

After signing the treaties, the expectation was that each nation would enforce some form of regulation aimed at limiting CO2 emissions. Hence, Canada launched the Canadian Net-Zero Emissions Accountability Act in 2022 which has since taken shape as financing efforts for promoting renewable and efficient energy.

That same year, the World Economic Forum released research saying AI promotes green energy by making automotive procedures more productive through helping “boost the accuracy of emission measurement”—an action proven to have reduced emissions by 40 per cent in some corporations.

In fact, the AI software market is projected to reach the grand sum of $126 billion by the end of this year. Contributing to this growth is the increase of companies implementing AI software into their business operations—a pattern that’s swollen by 270 per cent since 2021, according to a market snapshot published last year in the Journal of Hybrid Advances.

Aside from the economic advantages of replacing human labour with AI, it seems that a number of companies have began to use it to assist in complying with the carbon neutral regulations placed on their industry.

The future’s looking brighter, right? Well, not entirely. Despite there being superficial success in the event of a corporation following carbon-conscious guidelines by introducing AI technology, the software demands significant power be generated from data centres that use and emit carbon.

Cutting costs with AI: How the environment pays the price

Hitting a milestone of 200 million weekly users in August 2024, ChatGPT’s mainstream popularity has sounded environmental alarms.

Asking AI questions takes around 33 times more energy than a traditional search engine according to a publication by Cornell University.

This is because ChatGPT falls under the large language learning model (LLM) category of AI and needs to be trained on information in order to be able to supply responses—demanding significant computing energy from the data centres powering them.

To put it in perspective, the amount of electricity produced by data centres globally in 2022 was equivalent to roughly 71 per cent of all of the electricity generated in Canada that same year. The IEA reported AI as being a major factor and predict that the global consumption of data centre energy will double by the end of 2026.

As of December 2024, the Government of Alberta announced their intention of using the province’s natural resources, cold climate and “low tax regime” as leverage for enticing the installation of AI data centres—adding to the already 239 data centres in Canada if successful.

Sarah Palmer is a Staff Writer for The Reflector 2024-2025.