Canada requires an integrated national clean electricity grid, more supercomputing power and greater venture capital investment to achieve sustainable sovereign AI, say industry and academic leaders.

To ensure the country has control and choice in AI and quantum, Canada also needs to take a different approach to the two technologies than U.S.-based hyperscalers and Big Tech, they said during a Canadian Science Policy Centre webinar.

Canadian AI and quantum should be developed as part of public infrastructure, to produce societal benefits like an affordable public health care system and improve the wellbeing of Canadians, they said.

“Sovereignty is about two things: control and choice,” said Phil Harris (photo at right), president and CEO at Ottawa-based Cerio, whose mission is to build the most effective and efficient AI data centres.

“Part of the control is where do we do things and how do we do things under our terms? And the choice is how do we do them in a way that's most responsible and most applicable to the things that matter to those [Canadian] values?” he said.

Canada can leverage AI and make it available to more people by not creating large hyperscale-type designs, but creating a distributed infrastructure model –  “not a scale-up model but a scale-out model,” he added.

Harris noted that massive U.S. government investment in the 1980s and 1990s led to the Internet and created the tech companies that flourished and controlled that technology.

So the Canadian government’s and the private sector’s supportive policies and adequate investments will underpin Canada’s ability to use its natural and human resources to build and deploy domestic AI and quantum technologies, he said.

However, “We don't have a policy that really instills confidence or incentive in venture capital funding,” so tax incentives are needed to encourage more venture capital investment, Harris said.

The country needs more Canadian companies, he said. “For every five Canadian startups, three go to the U.S. in some way or another. One survives maybe a month. The others don't. And that's because of the lack of VC bedrock funding in this country.”

There needs to be a coordinated federal, provincial and municipal framework “that combines one path to success,” said Craig Tavares (photo at right), president and chief operating officer at Vancouver-based BUZZ High Performance Computing, which owns and operates one of the largest sovereign AI clouds in Canada.

That framework “is really a national sovereign compute policy. If I'm building AI infrastructure, it's a national priority,” he said.

“That means fast-tracking, permitting, interconnections for those qualified sovereign AI infrastructure projects, [and providing] long-term government and industrial off-date contracts for Canadian control and compute,” Tavares said.

Canada is a global leader in AI research and talent, he noted. But having that talent without high-performance computers “is kind of like pilots without airports.”

Within the next three to five years, Canada has to focus on converting its abundant clean energy, land and talent produced by universities “into large-scale, Canadian-controlled AI infrastructure,” Tavares said.

That means the country needs to make sovereign compute a national priority and strategic asset, and invest in it the way it invests in defence, ports, railways, telecommunications and energy, he said.

“The barrier is not that Canada lacks ambition. The barrier is that AI infrastructure requires everything to move together: power, capital, fiber, procurement, policy and talent,” he added. “So infrastructure often moves in years, unfortunately, and that's the gap that Canada has to close.”

Create public-private partnerships and invest in supercomputing power

Seychelle Cushing (photo at right), executive director of partnerships at Simon Fraser University (SFU), said Canada’s strongest opportunity is creating public-private partnerships that will drive sustainable, national AI infrastructure, “both for scale and for sovereignty.”

She pointed to Burnaby, B.C.-based fuel cell company Ballard Power Systems as an example of a successful public-private partnership. The company has used SFU’s 4D Labs, an advanced materials facility, and Cedar Supercomputing Centre.

“That’s been able to drive development of our new fuel cell technology that is powering buses that are in communities,” Cushing said.

“As we close that loop on benefit to community, how do we ensure that this AI infrastructure helps also strengthen our clean energy transition in Canada as well?” she added.

Canada can tap its “stack of partnerships” – including government, industry, academia and SMEs – to help support Canadian researchers and companies develop new, sustainable technologies and get them to market faster using national AI infrastructure, Cushing said.

Dr. Nancy Ross (photo at right), vice-principal of research at SFU, said Canada needs to make a sizeable investment in supercomputing power, which “will bring people together around an all-purpose coping tool for our problems, both in the university sector and in industry.”

“We have a GPU (graphics processing units) shortage we have to solve. We have to solve it sustainably,” she said. “We do need supercomputing capability and capacity in order to get the R&D, the digital R&D done in Canada.”

But sustainability also means retaining Canadian talent, including university graduates, to build AI models and run the infrastructure, so Canada is less reliant on other countries, Ross said.

Harris noted that he has seen a deficit, just from some of Canada’s research institutes, of 200,000 required GPUs within the next 12 to 18 months. “You can’t do AI research without [sufficient] GPUs.”

The federal government is expected to announce a significant investment in a new, world-class, Canadian-built supercomputer this fall.

When it comes to powering data centres and other AI infrastructure, Canada’s clean hydroelectric electricity grids are “effectively tapped out. For the time being, the hydro resources that we have in Canada are effectively off the table for large-scale data centre building,” said Jaethan Reichel (photo at right), chief operating officer at Bell AI Fabric, a Canadian-owned cloud and AI platform created by Bell Canada.

Canada also needs to resolve the way energy is priced, he said, adding: “We allow basically a solar developer to get the same energy [carrier] price for what they produce, even though that energy has a cost to the rest of the grid as far as resiliency and performance.”

“That’s something that we'll need to solve going forward, because we do have a lot of grids in Canada that are just sort of broken. And I think the only real prospect of fixing that is nuclear,” Reichel said.

Given Canada’s aging demographic and infrastructure and the industry sector transitions that need to happen, the country must take advantage of the current transformational possibilities, he said. “There is just this sort of inevitability of the widening gulf between our national aspirations and our economic reality.”

Historically, Canada has built infrastructure without completely understanding what all the use cases are going to be, Reichel noted. “We've lost our ability to do that at this point. And I think that's holding us back from being able to make some of the changes that we should as quickly as we need to.”

Natural gas and regulatory reform are required to build AI infrastructure

Diego Mandelbaum (photo at right), chief development officer at Vancouver-based Corix, which develops, invests in and owns district energy systems, said using natural gas has to play a role in powering Canada’s AI infrastructure.

Some people “have villainized natural gas as part of this transition,” but Canada needs to get away from this sort of “ideological sustainability lens,” he said.

“I believe that no amount of environmental assessment is going to have the ‘people with the pitchforks’ not protesting,” Mandelbaum said. “So I think there's a 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.”
“Solar and wind [energy] can be good supplements, but there is no way that that is going to be driving our grid by themselves,” he added.

The most crucial thing Canada can do is move forward with connecting more east-to-west electricity transmission, rather than just south-to-north transmission to sell power to the U.S., Mandelbaum said.

Also, Canada’s biggest issue is that the country is risk-adverse, has too many guardrails and government gets too involved in the private sector’s business, he added.

“As private sector trying to put money to work, there is more red tape doing business in Canada than anywhere else,” Mandelbaum said. “And so how do we both lean on government to help facilitate this big transition we have to make while at the same time making it efficient and not having too much red tape?”

Ontario is expected to have several small modular nuclear reactors within 15 years, but Canada needs additional power-generating capacity now to keep pace with AI development, he said.

Canada also requires some education about what productivity actually means, Mandelbaum said.

“As Canadians, we love our social goods. We love public health care, but we can't afford [it]. We're pretty close to being broke and we've had no productivity gains in about a decade.”

“So I think it's becoming a fundamental need to become more productive because just opening up the immigration floodgates for GDP growth has come to an end,” he added.

Tavares said success for Canada means the country “is not just an AI talent exporter or a cloud consumer. It means Canada becomes a trusted global platform for clean, sovereign and production-grade AI infrastructure.”

Ross said if Canada doesn’t invest in AI and quantum infrastructure, the risk is “I think we give up Nobel prizes. We give up our place in the global order of discovery and science and life science.”

However, many Canadians don’t necessarily see themselves directly benefitting from AI, and they worry about the environmental impacts from data centres, she said.

But Canada “is not going to compete in the global race of hyperscalers. We're going to do things differently here,” Ross said. “We are going to be building very carefully. We are going to focus on innovation and discovery here for a lot of our public infrastructure.”

Ross said she hopes that in five years, “Canada has a global reputation, not only for AI research and AI talent, but also some impressive public infrastructure, and it has created alongside that investment and public infrastructure an innovation ecosystem that's really impressive and is globally competitive.”

R$