Canada’s newly-announced AI strategy includes plans for “large scale” AI data centres, a move that comes amid growing pushback from many communities across the country and in the U.S. concerned about the impacts these facilities could have.
Data centres require large amounts of energy and resources to operate, on top of their initial construction impacts and knock-on effects to local communities, like potentially higher electricity rates.
“Data centres require huge, huge amounts of energy, and Canada has both carbon-based energy and now more and more renewable energy. So we do seem like a natural home,” says computer science professor Arvind Gupta at the University of Toronto.
“If we want public acceptance of data centres, we’re going to have to think about sustainability.”
A recent Angus Reid study released on June 1 polled about 1,800 Canadians, with 68 per cent saying they would oppose a large AI data centre near their home. The majority cited energy and environmental sustainability and the impacts to their local neighbourhoods and surrounding communities.
One such example was in Saskatchewan, where an AI data centre was approved after being protested by the local community.
At the same time, nearly half (46 per cent) of survey respondents said they support the idea that Canada needs domestic AI infrastructure, like data centres, to maintain sovereignty over these technologies.
The AI strategy, written by Artificial Intelligence Minister Evan Solomon, describes how the government’s plans will require doubling Canada’s electricity grid capacity by 2050, which echoes Prime Minister Mark Carney’s announcement on May 14.
Gupta said it’s key that community consultations are done to get these projects up and running.
“If you don’t consult the community, it becomes really tricky, and the local politicians will then make it really hard for that company to operate,” he says.
“So it’s actually in the company’s interest to make sure they’re doing proper consultation and really becoming part of that community.”
Artificial intelligence, generally, requires a massive amount of energy to function because of the advanced physical technology that powers the programs.
Get breaking National news
Get breaking Canada news delivered to your inbox as it happens so you won’t miss a trending story.
Similar to a standard computer, this includes microchips, processors, racks and connective components, as well as cooling systems — but on a massive scale.
These data centres not only require their own dedicated spaces to function, but often entire compounds comparable to a small town. This means they can have a big impact on their surrounding communities not only because of their size, but what they require to function.
“Data centres are energy hungry,” says Mike Welland, an engineering physics professor at McMaster University, whose background also includes nuclear engineering.
“So because the Ontario power grid is largely decarbonized, due in large part because of nuclear power, it’s a good fit to put data centres as a sovereign resource inside Ontario.”
In a separate interview with Global News from May 14, Welland explained that the energy required by some AI software like ChatGPT can be compared with the time used on a consumer microwave oven.
“A simple factual question is about one to 1.5 microwave seconds — on par with a Google search. If you’re having a conversational query, where the AI has to reread the script every time, you’re looking at two to five microwave seconds. If you give it a short document and you’re asking it to summarize, it’s about 10 seconds,” Welland said at the time.
“And then short video generation becomes massive. That becomes five to nine minutes, even for video that is a few seconds long.”
In the AI strategy, it says these data centres will scale to at least 100 megawatts, and Welland says a typical nuclear reactor in Canada may put out between 700 and 800 megawatts.
If these data centres are going to be plugging into the same electricity grid as local residents, Gupta says that has the potential to drive up electricity rates.
“Are we subsidizing electricity for a data centre and driving up the cost for the local residents?” he says.
“These are very energy-intensive operations, and so it’s just demand and supply. If you get a new operator coming in that has huge demand for energy, there won’t be as much energy, and of course, energy providers, the hydro companies will jack up rates.”
Data centres don’t just require a large and reliable source of energy to power their machines. They also need a way to keep those devices at a stable temperature while they produce heat.
This, Welland says, is why Canada is a good spot to build out data centres, especially because of the vast amount of fresh water.
“A computer is a remarkably effective space heater … that heat needs to go somewhere. So it needs to be either put into the air or into the atmosphere,” Welland says.
“One of the benefits of potentially putting one of these data centres near a large body of water — for example, Lake Ontario — is because it’s a freshwater system.”
Welland explains that, like nuclear power plants, these data centres can pull water straight from the bottom of a lake to flush out the heat from the data centre and return the water back to the lake.
He adds: “Unfortunately, there is an aspect of heat pollution involved in that, which is a very valid concern, but it doesn’t consume drinking water.”
Heat, or thermal pollution, is defined by the United Nations as “the discharge of heated effluents from industrial processes such as electric power generation, atomic power stations and other factories at temperatures that can affect the life process of aquatic organisms.”
The AI strategy does not address mitigating thermal pollution impacts, but says Canada’s approach will include “robust environmental standards, and tangible benefits for local communities.”
The strategy also highlights Canada’s “physical advantages,” including the cooler northern climate as being a way to reduce the cost and energy intensity from these data centres.
On the same day as the federal announcement, Manitoba Premier Wab Kinew announced the province will reject a hyperscale AI data centre, citing the size of the project, the energy it will consume and its impact on the community outweigh the limited gains.
“There’s a big threat to the environment and not much benefit to the economy,” he told reporters at the Manitoba legislature.
Prairie communities are among those who have been pushing back, with similar opposition to a project in New Brunswick from locals.
“There’s no reason that these companies can’t have a proper dialogue with the community to make a plan, and make it a net positive for everybody,” Gupta says.
“So I see this as a partnership between these companies and the communities they go in, and if they approach it this way, do proper consultation, really think about the needs of the community, then I don’t see why communities would be against it.”