.png)
The corporate sector’s mad dash for AI may be leaving sustainability goals behind.
Driving the news: Data centres are driving up electricity demand, which is expected to double by 2026. While some of this can be attributed to more internet use and electrification in countries like China, a big culprit is AI, which requires massive amounts of data processing.
- Some U.S. states with many data centres are putting off closing coal and natural gas power plants to meet rising demand.
- A lot of variables make AI’s carbon footprint hard to measure, but one analysis suggested training OpenAI’s GPT-3 model required 1,287 megawatt hours of electricity, enough to power over 100 average Canadian homes for a year.
Zoom out: Programmatic advertising also puts data centres to work as ad tech companies constantly process data to target ads. According to tracking data from Scope3, an average web display ad generates 270 grams of CO2 for every 1,000 impressions it earns in Canada.
- For perspective: If just one ad was seen every time someone visited the Toronto Star website, it would generate over 7 metric tonnes of CO2 every month — comparable to two round trips between Montréal and Vancouver in a Cadillac Escalade.
- Emissions are almost three times higher for automated video ads, a rapidly growing channel as ad-supported streaming and connected TV get more popular.
Why it matters: Besides the obvious wrench this throws into global plans to cut emissions, it’s also a problem for companies’ ESG commitments. With corporations leading the frenzy for AI adoption while also moving more ad spending to automated systems, their environmental track records might not stand up to scrutiny if investors start including data centres in their analyses.
What’s next: Measuring and reducing AI-related emissions is part of the EU’s Artificial Intelligence Act, as well as proposed AI legislation in the U.S. It is not part of AI rules in Canada’s Bill C-27, but as it is still in front of a committee, that could change.