The western edge of Thompson Rivers University (TRU) in Kamloops, B.C. is alive with activity, as a small army of engineers, project managers and skilled tradespeople engage in the careful work of building, wiring and cooling the first of two 13-megawatt Bell Canada data centres set to open at the school in the next 18 months. From the local lumber being sourced to erect the mass-timber structures, to the Canada-made closed-loop thermal system that will keep it cool, to the Secwépemc ways of knowing underpinning the entire development, many stakeholders are invested in making sure the project is both built well and built right.

For the institution hosting the data centres, and whose students and faculty will use the resultant compute power, the current hustle and bustle is a harbinger of a more secure AI future. “The risk of not building is not theoretical,” reasons Andrea Li, a tech veteran who now serves as a Special Advisor on AI to TRU’s leadership. “Every year without sovereign compute is a year Canada risks losing its data, its talent and its ability to shape AI in its own image.” 

Those accustomed to thinking of AI in terms of algorithms and LLMs, the physicality of the TRU build, and the dozens of projects like it in development across the country, might come as a surprise. But it’s the nuts-and-bolts reality as Canada races to turn AI sovereignty from a federal talking point to material infrastructure capable of safely meeting the compute needs of the country’s most sensitive and critical AI applications. If Canadian businesses, institutions and governments want more dependable domestic AI capabilities and less reliance on global giants, experts say the country needs to build: Construct more data centres, install more electrical and thermal management systems and lay more fibre-optic cables. 

“Sovereign AI is a critical asset for Canada to advance our economic growth, productivity, innovation and competitive position on the global scale,” says Namir Anani, President and CEO of the Information and Communications Technology Council (ICTC). “And it is very physical.” If the near-daily announcements about new Canadian AI infrastructure projects roll out in thoughtful and responsible ways, he adds, the economic impact will extend far beyond the digital economy: “If we play this moment well, it could generate a domestic industrial boom.”

Early evidence suggests how fulsome that upsurge might be. Consider the ripple effects of Bell Canada’s current multi-billion AI Fabric infrastructure build-out. “We’re building a national AI infrastructure backbone for Canada,” sums up John Watson, the company’s Group President for Business Markets, Bell Cyber, AI Fabric and Ateko. And that is transforming the contractors, suppliers, skilled workers and communities working to make sovereign AI a reality. 

Building a Canadian backbone

The global AI infrastructure landscape is highly concentrated, with so-called hyperscalers (such as AWS, Google and Microsoft) now controlling nearly half of worldwide data centre capacity and spending trillions to grow their share. Around the world, leaders are coming to grips with the risks associated with running important AI applications (such as those related to health, defense and government work) through foreign servers, data centres and cables: 71 per cent of global investors, executives and government officials surveyed by McKinsey & Co. late last year classified sovereign AI as an existential concern or strategic imperative to organizational goals. As Rahul Krishnan, Canada CIFAR AI Chair at The Vector Institute, wrote in a May op-ed: “Countries that fail to build the energy, compute and institutions required for AI will become dependent on those that do.”

In Canada, this sentiment is translating into a building rush. While the country has long lagged G7 counterparts in domestic AI compute capacity, evidence suggests the tide is turning. “Canada’s domestic stack for infrastructure is developing faster than the prevailing narrative suggests,” observed researchers Jaxson Khan and Sabreena Shukul in a June report from RBC Thought Leadership. Indeed, The Logic’s recent investigation found more than 300 operational data centres in Canada, and a York University analysis published in March tallied upwards of 150 additional sites announced or under construction, with an average power capacity more than 10 times greater than the centres already running. 

“There’s strong momentum in the space,” confirms the ICTC’s Anani. The $2 billion earmarked for capacity-building in the federal government’s 2024 Canadian Sovereign AI Compute Strategy triggered new activity that is now materialising into substantial private, public and pension fund investments, he explains. “All indicators suggest that AI infrastructure is now seen as a critical component for innovation, productivity and economic growth for any economy, and that creates a strong signal for Canada to put the investment necessary to get us where we need to be.”

That’s evident in the ambition of Bell Canada’s AI Fabric strategy, which involves the rapid development of a network of data centres, high‑performance compute and secure environments with a combined capacity of 800 megawatts. AI Fabric includes two AI data centres that have gone online in the past year (in Kamloops and Merritt, B.C.) and four more in development (including the pair at TRU and a massive 300-megawatt facility in the Rural Municipality of Sherwood, Sask. whose first data hall is slated to open next year). 

“This is nation-building infrastructure, and it’s happening on a massive scale,” explains Dan Rink, Bell Canada’s President of AI Infrastructure & Strategy. “We see a growing demand in enterprise and public sector organizations for trusted AI. Our ultimate goal is to give them access to the AI infrastructure they need in a safe and sovereign way, with as much Canadian inputs as possible.”

Strengthening supply chains

Made-in-Canada data centres carry clear benefits for the organizations that own them and use their services. But that’s only part of the value chain being forged in the current press to develop sovereign AI infrastructure. Every new data centre in Canada needs to be built to exacting technical specifications, equipped with sophisticated HVAC and electrical systems and hundreds of highly specialized components, and, once up and running, operated and maintained by engineers and data scientists who know what they are doing. That translates into a lot of economic activity.

“These data data centres are very large in scale, and they’re very specialized,” explains Teri McKibbon, President and CEO of Mississauga, Ont.-based Bird Construction, which was recently named construction partner for Bell Canada’s Sherwood complex. “They are multi-year builds that require a considerable level of technical complexity and coordination.” Data-centre construction is akin to that required for major industrial, hospital, or energy projects, he adds, which bodes well for the highly skilled project leaders and tradespeople on his team. “The scale of this build-out represents years and years of work, and real career opportunities,” he says. 

Bell Canada’s Sherwood project is expected to support at least 800 skilled trades and engineering jobs during construction, 80 permanent roles once operational, and as many as 750 ancillary jobs. For its part, Bird is supplementing its own skilled talent pool by hiring from local and Indigenous communities. “It adds another dimension to the economic impact,” McKibbon says. “We want to ensure the benefits are shared across the communities where these projects are located.”

This approach is shared by Gustavo Paredes, Senior Vice-President, Americas, at Smardt, a Montreal-based specialized HVAC manufacturer that provides chiller equipment for several Bell AI Fabric sites. Smardt has seen “significantly larger demand” for data centre work in recent years, he says; the company is developing new products to address the sector’s growth, and is building a second manufacturing site in suburban Montreal to help meet the need. “The more volume we get, the more engagement we have in the ecosystem,” he says, pointing to downstream companies Smardt engages with for design, installation and service. “We see the ripple effects: Our partners are growing their staff, and the knowledge base in trade labour is growing. The job creation that’s happening is quite significant.”

These are exactly the kinds of spinoffs Bell Canada was hoping to generate when it began its data centre builds, and is hoping to accelerate as more sites get closer to completion. “We made a deliberate choice to invest in Canada and back a Team Canada model for AI,” says Watson. “There’s no shortage of talent or innovation here, but there has been a gap in the large-scale, trusted infrastructure needed to bring it all together and create meaningful economic opportunity.” Accordingly, nearly all of the labour used in the AI Fabric build is Canadian, while the vast majority of the components used during construction are sourced from domestic suppliers, from steel to concrete to sophisticated industrial chillers.

Forging new futures

Bell Canada’s focus on local sourcing and hiring is not altruism: Given the concerns many Canadians have about data centres, community support for large-scale AI infrastructure builds is far from guaranteed. “We know we need the social license to build these facilities. We need to be a good community partner, and we need to build projects in a way that aligns with Canadian values,” notes Rink. “So, we’re choosing to do things the right way.” In practical terms, that means striking partnerships with a broad range of stakeholders, including local First Nations, governments, community organizations and universities, to ensure projects serve the economic and social interests of all involved.

The centres under construction at TRU offer an instructive example. Bell Canada’s agreement with the school allocates funding for scholarships and bursaries, and also gives the TRU researchers and trades students access to high-performance compute capacity needed to pursue solutions for everything from wildfire risk to rural health access to microplastics detection to  salmon monitoring (the latter in partnership with local First Nations). “Our students are entering a world where AI literacy matters in virtually every field,” says Dr. Airini, TRU’s President and Vice-Chancellor. “Our job is to give them the ability to understand, question and use these technologies wisely, and this partnership makes that tangible.” 

In the longer term, the TRU team aims for Kamloops to become what Li describes as a “significant node” in Canada’s sovereign AI infrastructure, drawing talent, anchoring investment and signalling that top-tier AI capabilities can exist outside of major urban hubs. “For a mid-sized interior city, that’s a different kind of future,” Li says, “One where our graduates have reason to stay, build careers and contribute to their community.” 

In Rink’s view, any discussions about manifesting “AI for All” (the core theme of the federal government’s recent AI strategy) should centre examples like these. “This is a new frontier, it’s a whole new industry, and it’s not just a software story,” he says. “Our success will depend on physical infrastructure on a national scale, built in the right way, with the right regulations and in a way that ensures value for all Canadians.”

This content was paid for and directed by Bell and was produced independently of The Logic’s newsroom in consultation with the advertiser. You can read our policies on advertising, sponsorships and partnerships here.