The great AI debate: industry pros turn down the temperature on data centres
During a Canadian Science Policy Centre panel, industry leaders discuss what underpins sustainable, sovereign AI, highlighting the impact of policy, economics, and social unity.
DATA CENTRE POLITICS
Building up the backbone of Canada’s AI ambitions
In the 1800’s, many North Americans rued the money being invested in railways. The politics were fraught. The projects moved fast. They cost far too much. And some didn’t understand just how much those rails would improve lives and the economy.
Fast forward a couple hundred years, and we now know perfectly well just how instrumental those arteries were to the West’s progress, our trade, and human connection. Despite early tensions, all that steel and board formed the backbone of North America (4,700 km in Canada and 3,070 km in the U.S.).
Today, we’re seeing loud, albeit toothless, protests against data centers popping up across North America; it seems AI has a bit of a communications problem. (This has emerged since the government’s fresh AI strategy was dropped.)
While data center debates are healthy, any significant delays could hold the country back to the point where it becomes beholden to more powerful nations.
“From our perspective, sovereignty is about two things: control and choice,” says Phil Harris, CEO of Cerio. Establishing control and choice would allow the country to select its strategic AI partnerships and set sustainable AI policy, explains Seychelle Cushing, SFU Executive Director (Partnerships Hub).
The country also wants to avoid having more powerful nations capture the value created on Canadian soil, which is not uncommon.
“The risk is not that Canada, Canadians, stop inventing,” explains Craig Tavares of Buzz HPC; the risk is that “others commercialize and control what Canadians invent.”
“Canada’s AI future will not necessarily be secured by models alone. It will be secured by the infrastructure that trains them. The servers that serve them, and then protects them.” — Craig Tavares
AI data centres as critical infrastructure
Proponents like Tavares present sovereign AI compute as a “strategic national asset” on a par with railroads, energy grids, roads, and hospitals.
From disease research to drug discovery, AI is being primed to help the nation with other pressing issues. Indeed, Joy Johnson, SFU President and Vice-Chancellor, considers AI and its infrastructure as the next critical step in addressing those foundational challenges.
“We’re not going to get out of this problem by just continuing to do what we’re doing. We need something to… move us forward in a different way.”
A hybrid model: getting the energy mix just right
AI infrastructure may only be as good as our energy infrastructure. But to remain globally competitive, Cushing suggests we should approach AI infrastructure as a network of integrated systems and partnerships spanning infrastructure, energy, business, and innovation.
Tavares and Diego Mandelbaum, CEO of Corix, echo this as well. They argue that a hybrid approach to resources like nuclear, solar, wind, natural gas, electricity, and heat will be key. However, Mandelbaum also notes that “it’s really hard to do that from a pure private sector perspective.”
“I think we’ve got to be pragmatic and get away from the sort of ideological sustainability lens…” — Mandelbaum
According to Mandelbaum, greater connectivity between provinces could help expand both Canada’s AI infrastructure and energy grid.
There’s no one solution, says Tavares, a common belief among pragmatic sustainability proponents: “whether it’s nuclear plus gas plus solar and wind, we’ve found ways to… combine, so that we can provide capacity…”
“If we look at what we have today, we have some of the cheapest, greenest, most reliable power in the world… nuclear needs to play a big role going forward…” — Mandelbaum
“There is a real misunderstanding around how [energy] grids work,” says Jason Reichl, COO of Bell AI Fabric. He believes that many don’t understand the energy mix required for major infrastructure. Particularly, the limitations of intermittent renewables like solar and wind.
“You started to see that in the grid issues that Spain is having, and California is having, and Germany is having,” Reichl says, outlining how a reduction in nuclear reactors contributed to grid instability and to the temporary need to burn coal in Europe.
He believes that our AI power requirements could help push our nuclear ambitions forward more quickly:
“I think the only real prospect of fixing [broken grids] is nuclear. And that’s definitely one of the things that I think these [AI infrastructure] projects can help with… justification for people to move there faster.” — Jason Reichl
Mandelbaum argues that Canada might benefit from following in the footsteps of parts of Scandinavia or the Netherlands in its heat reuse requirements: “... the government said [to Meta] if you want to go build a 200‑megawatt data center, you’ve got to be able to take that waste product of heat and reuse it.”
With so many power options and cooler weather, much of Canada presents an almost ideal option for data center infrastructure. While concerns about data centre water usage are proliferating, Canada holds a unique advantage: the country’s notoriously cold weather provides a “natural cooling effect,” explains Tavares.
“I believe that no amount of environmental assessment is going to have the people with the pitchforks not protesting. So I think there’s a big, big job we have to do in communications and education. But I don’t think that the answer is more environmental assessment or more red tape.” — Mandelbaum
Communication breakdown: why people are panicking
Despite the opportunities AI presents, people remain concerned about water use, power sources, and jobs, as evidenced by anti-data-centre protests popping up in Vancouver and Hamilton.
However, Canada’s cold weather does give it a distinct sustainability advantage. Mandebaum and Nancy Ross, Queen’s University Research VP, agree that people still don’t understand how AI will benefit them, or the innovation it enables.
“...folks don’t necessarily see themselves benefiting from AI directly, immediately, and they get worried…” — Nancy Ross
Part of public concern appears to stem from economic insecurities and other social priorities. Despite this, AI-led transformation could also provide some solutions, depending on how it’s implemented.
“Some of the biggest critical issues that we’re facing in Canada are… the state of our health care system, the state of our infrastructure, etc. We’re not going to get out of this problem by just continuing to do what we’re doing…” — Joy Johnson
Reichl suggests that there is a growing gap between Canadian aspirations and the economic reality if we get left behind and don’t deploy AI and infrastructure effectively: “People don’t care about the environment when they don’t have jobs….”
Is economic growth the key to unity?
Social cohesion is being tested on a number of fronts, as well documented, and Phil Harris stresses the importance of addressing it economically. Harris ties AI infrastructure, well-managed technology, and economic growth to the avoidance of social breakdown.
“Go back to the Treaty of Rome after the Second World War. That was because they recognized that, as Europe came back together, if we didn’t have that social cohesion through economic growth, our social cohesion would be fractured without that economic growth. That we would lead ourselves back into another world war.” — Phil Harris
Mind the bureaucracy gap
It’s increasingly clear that governments have an immense role to play in establishing sovereign AI, including the underlying energy infrastructure. Of course, any bureaucratic system is lengthy and often does not keep pace with technological change.
“We have too many guardrails,” explained Mandelbaum, “there is more red tape doing business in Canada than anywhere else I’ve done business.” He acknowledges that the industry will have to simultaneously lean on the government to support the country’s AI transition, while doing so efficiently.
Canada’s AI building blocks are all there.
Its ability to train AI talent and generate cutting-edge research is well known and commendable. And its ambition is clear, says Tavares.
“The barrier is that AI infrastructure requires everything to move together: it’s power, capital, fibre, procurement policy and talent.” Tavares adds that “infrastructure often moves in years… that’s the gap that Canada has to close.”
“For every five Canadian startups, three go to the US in some way or another. One survives, maybe, and the others don’t… And that’s because of the lack of VC bedrock funding in this country, which we need.” — Phil Harris
One way around bureaucracy could be to stimulate stronger private-sector and VC capital investment, taking a page out of the Palo Alto playbook, “where you can’t trip over 10 feet without falling into a venture capital company that is funding a lot of infrastructure development right now,” describes Phil Harris.
“We don’t really have that same venture capital mindset in Canada…. that will come from tax incentives. That will come from other forms of government willing to allow venture capital to come in and fund a lot of this work. Because we need more Canadian companies.” — Phil Harris
Indeed, at the time of writing, the province of Alberta and Meta announced a $13 billion AI data centre project in Canada, the largest of its kind outside of the U.S. The project is slated to generate thousands of jobs and includes an adjacent natural gas facility to power the data centre. The project is slated to use a closed-loop or dry-cooling setup.