Developers Talk Of Big Data Centres While Reality Calls For Smaller Projects
Canada’s data center landscape could undergo a significant shift brought about by a trio of proposed one‑gigawatt projects in Alberta and Saskatchewan, analysts say, developments that would be up to 10 times larger than anything operating in the country.
But industry executives contend that the reality on the ground is far more constrained, with the largest data centers under construction topping out at roughly 35 to 50 megawatts. A megawatt is about 1,000 kilowatts, or enough energy to power up to 900 average homes.
That gap between long‑term ambition and near‑term delivery surfaced repeatedly during a recent data center panel at the Montreal Real Estate Summit, where industry players described a market still limited by power availability, infrastructure and execution timelines.
David Arsenault, vice president at Montreal‑based hardware maker Hypertec, said artificial intelligence is changing what data centers require to be most efficient in the AI world.
“AI is driving much higher density, with far greater power requirements than before,” Arsenault said.
James Beer, CEO of QU Data Centres, said that despite those pressures, the largest data centers currently under construction in Canada remain relatively modest in scale — in the 35‑ to 50‑megawatt range — even as developers outline
proposals calling for vastly larger facilities.
While companies are beginning to measure future sites in hundreds of megawatts or more, Beer said a 50‑megawatt facility already marks a step change for Canada in terms of capital cost, construction complexity and power consumption.
The contrast is even sharper in Quebec, where power constraints have effectively capped individual data center developments at about five megawatts. Panelists said that has resulted in Quebec being dominated by smaller, high‑density facilities, rather than the expansive campus‑style developments being contemplated in Western Canada. Those constraints, speakers said, are beginning to influence how the industry organizes itself.
Moving toward split model
Ambroise Lalloz, co‑founder of First Block, said Canadian data center development is increasingly being planned based on a bifurcated setup rather than a single type of site.
“We are moving toward a distributed model,” Lalloz said, with large, remote facilities intended to handle AI training workloads and smaller sites positioned closer to users for inference and real‑time processing.
Panelists said that structure reflects how AI is actually deployed as it moves from research environments into everyday commercial use. Arsenault said inference workloads — centred on AI models making predictions on new, unseen data — that support tools used directly by businesses and consumers need to sit closer to population centres.
“The smaller sites are getting more compressed,” he said. “You are doing more in less space, closer to the user.”
Kal Benedict, vice president of business development at Shared Fibre, said access to fibre has become just as critical as access to electricity. “If the fibre is not there, the location does not work, no matter how much electricity you have,” Benedict said.
Several proposed megaprojects in Western Canada, including two one‑gigawatt sites in Alberta and Bell’s Regina‑area build, were part of the backdrop. Panelists said those projects are long‑term bets that still face significant obstacles and would not eliminate the need for smaller urban facilities.
Several speakers also pointed to data sovereignty as reinforcing a more decentralized approach, arguing that a national network of interconnected facilities could help keep Canadian data and AI workloads within the country.
David Cervantes, senior director at CBRE and moderator of the panel, said data centers can no longer be viewed as isolated real estate projects.
“It is not city‑centric anymore, but it is not remote‑only either,” Cervantes said, describing a connected network in which facilities of very different sizes are designed to complement one another.
