There is a national consultation underway to develop Canada’s AI strategy. A few weeks ago, a summary of the submissions was released.
Canada’s AI capabilities are having a moment, but not what we might necessarily brag about. For years, we’ve celebrated an AI research pedigree while quietly ignoring the infrastructure gap beneath it. A recent Policy Options article by Joe Rowsell frames the issue clearly: Canada is talent‑rich but compute‑poor.
That’s more than just an inconvenient detail for the tech sector; we might ask if it could be a national strategic vulnerability.
Rowsell describes a reality that many in the telecom and digital infrastructure space have seen coming: Canadian researchers routinely ship their workloads — and often their data — to American‑based hyperscale facilities because domestic compute capacity isn’t there.
In other words, we may have built a world‑class AI research ecosystem, but it is riding on top of others’ servers.
More than a question of performance, it raises questions about sovereignty. When the computational infrastructure is outside our borders, so is operational control. And, when AI begins to underpin the operation of critical infrastructure — energy, communications, transportation, health — control matters.
The paper highlights another tension: AI’s global energy appetite is exploding. Servers optimized for AI workloads can consume ten times the electricity of conventional machines, and global demand could rival Japan’s power consumption by 2030.
That’s a major electric grid planning problem.
Canada is in a better position than most countries to solve it. More than 80% of Canada’s electricity is considered non-emitting, powered primarily by hydro and nuclear, and supplemented by solar and wind. Canada’s climate naturally reduces cooling loads. These are significant structural advantages over the US and Europe, if we build on them.
Rowsell proposes a framework built on three pillars:
- Sustainable-by-design: Build data centres that start efficient rather than retrofit later.
- Sovereign-by-design: Keep Canadian data, models, and operational control inside Canadian infrastructure.
- Responsible-by-design: Bake governance, safety, and Indigenous data sovereignty into the architecture itself.
A compelling argument in the article is for AI to be a grid asset, not simply an “energy hog”. With carbon aware scheduling, and waste‑heat capture, data centres can actually strengthen the grid rather than strain it.
From a telecom perspective, this conversation is overdue. AI workloads will reshape network demand patterns, data‑centre siting, spectrum planning, and cloud‑edge architectures. If Canada wants sovereign AI, it will need sovereign transport, sovereign interconnect, and sovereign cloud.
Compute without connectivity is useless. Connectivity without compute is a missed opportunity.
These lines stuck with me: “AI will not run out of code or chips, but it will run out of clean electrons. Electrons are the new currency of intelligence, and unlike most of the world, Canada’s supply is clean.”
Rowsell ends with a warning: Canada’s clean‑energy advantage won’t last forever. Other countries are racing to decarbonize their grids and scale their compute. Canada has those clean electrons — now. If we don’t move quickly, we could lose this significant structural advantage.
Earlier this week, an editorial in the Globe and Mail warned us to “Pay attention to what’s behind the AI curtain”.
The loudest proselytizers for artificial intelligence like to present the technology as inevitable. That may be so. However, we all need to be sensible enough not to fall for tech-bro bravado. Exaggerating what AI has actually achieved cannot be a way to convince people that this future has already arrived.
When the government released its summary of submissions, Michael Geist wrote a review, “What the Government Isn’t Saying About the Results of its AI Consultation“. He found the Government’s official summary softened some of the expert panel’s key messages. “The experts are trying to sound the alarm on the risks to Canada if it fails to act but that isn’t the message the government seemingly wants to communicate.” The summary of submissions may be less of “What We Heard”, and more like “What We Wanted to Hear”.
Canada’s AI research capabilities and talent are well known. Will we move quickly enough to build the necessary infrastructure to retain that talent at home.