AI sovereignty depends on scale and consistency. That requires a new national approach to compute.
George Ross, CEO, Digital Research Alliance of Canada
Feridun Hamdullahpur, Board Chair, Digital Research Alliance of Canada
Artificial intelligence sovereignty is at the forefront of Canada’s nation-building agenda. To that end, Ottawa has committed $705 million through the AI Sovereign Compute Infrastructure Program to establish new sovereign domestic compute capacity. This is excellent. But when it comes to AI compute today, the strategy matters even more than the spend.
To understand why, we first need a clear definition of AI sovereignty. Here’s what it should mean: Canada’s public compute capacity must be located on Canadian soil; it must reduce dependence on foreign-owned commercial cloud platforms; it must operate at a scale and level of sophistication comparable to peer nations; and it must remain publicly owned and deployed in service of the national interest.
Achieving this sovereignty requires a new approach to compute — a national approach.
A national approach requires breaking old patterns. Historically, funding agencies have divided funds among many systems at different sites across the country. Most of these systems are customized to meet their own local users’ needs. This approach built a strong foundation, but also created a fragmented landscape that isn’t suited to deliver two things that researchers tell us they need: scale and consistency. Both needs have profound implications for Canada’s AI sovereignty going forward.
Scale is vital to working with today’s largest datasets and most complex models. If Canada can’t provide its researchers and other compute users with that scale, then they may feel compelled to turn to systems that can: namely, those supplied by major multinational hyperscalers. Without a Canadian alternative, we risk seeing our researchers increasingly rely on foreign-owned commercial systems and their private digital ecosystems.
Inconsistency poses a similar risk. While Canada’s public compute systems generally use the same authentication, security, administration and software tools, they have different hardware configurations that lead to different levels of performance. In addition, each system comes with a steep learning curve, which creates a substantial barrier for new users. These systems, distributed across the country, can’t be combined and users can’t easily port work or data from one system to another without incurring significant delays or costs. These limitations also risk spurring some users to seek more flexible commercial systems to meet their needs.
With the AI Sovereign Compute Infrastructure Program, we have the opportunity to adopt a plan that can address these key challenges to Canada’s AI self-reliance.
The solution is to take a new national approach.
This means investing in one large-scale supercomputer, hosted at a single site, administered by a single organization, and made accessible to users across Canada. It also involves incorporating Canadian technology wherever possible. This approach would minimize fragmentation and maximize scale to support the most demanding AI workloads. It would deliver a system that could scale with future investment, while providing consistency and a single learning curve for researchers.
This would be Canada’s new AI backbone and the cornerstone of our sovereign AI future.
There’s an additional nuance. The national approach will see the most success if the organization leading it is pan-Canadian and independent — one positioned to allocate public compute resources in the national interest, free from institutional or regional demands.
Entrusting this national approach to an independent public organization with a clear national mandate will ensure that compute resources are allocated responsibly and equitably throughout the country, and that the system will remain insulated from competing priorities or commercial pressures. An independent public organization could also represent Canada internationally, allowing us to engage as equals with other nations that have built similar public models for their sovereign compute.
This combination – a national approach led by an independent public organization – is how Canada can optimize for success and play to win. It’s how we can provide the scale and solutions that maximize Canada’s self-reliance. And it’s how we de-risk the AI Sovereign Compute Infrastructure Program to ensure that this nation-building project has staying power, growing Canada’s capacity today and well into the future.
The stakes are high. This is our chance to lead with the ambition and transformative change the nation needs.
By taking a national approach, administered by an independent public organization, we can build out a new Canadian AI backbone that enhances our sovereignty, grows with future investment and serves the national interest in perpetuity.
The Digital Research Alliance of Canada (the Alliance) is the national organization responsible for integrating and advancing Canada’s digital research infrastructure (DRI) ecosystem. Working with government, academic and industry partners, the Alliance leads national DRI strategy, funds world-class infrastructure, and services and equips researchers with the technology needed to drive discovery, innovation and economic growth. With a membership of more than 100 universities, colleges, research hospitals and institutes, the Alliance delivers coordinated, accountable and sustainable digital infrastructure that strengthens Canada’s research and innovation capacity.
Learn more at ai.alliancecan.ca.
