Canada’s 2017 Pan-Canadian AI Strategy was a research strategy — and it worked. The next one will be industrial policy. Expect sovereign compute as national infrastructure, harder pressure for Canadian AI champions, governance through standards instead of sweeping regulation, an agent-and-MCP focus rather than chatbots, and Canada positioning itself as the “Trusted AI” country. The question has shifted from how to become relevant in AI to how to hold onto sovereignty, economic value, and public trust as AI reshapes national infrastructure.
Back in 2017, the world was obsessed with crypto, fidget spinners, and whether Netflix was going to kill cable.
“Despacito” was everywhere. The Nintendo Switch had just launched. And somewhere in Canada, a small group of researchers helped set off the modern AI revolution.
That same year, Canada launched the world’s first national AI strategy.
At the time it felt almost academic. AI mattered, but it still mostly lived in research labs, university papers, and conference stages. ChatGPT didn’t exist. Claude didn’t exist. Most businesses still thought “AI” meant chatbots that barely worked.
Canada saw something early.
The 2017 Pan-Canadian Artificial Intelligence Strategy, led by CIFAR and backed by $125 million in federal funding, focused on a few core ideas: keep elite AI researchers in Canada, build globally recognized AI institutes, train the next generation of talent, push collaboration between academia and industry, and develop leadership in ethical and responsible AI.
It helped strengthen three major hubs: Mila in Montreal, the Vector Institute in Toronto, and Amii in Edmonton.
It also created the Canada CIFAR AI Chairs program to attract and keep world-class researchers. That mattered, because Silicon Valley was aggressively recruiting Canadian AI talent at the time.
Looking back, the strategy worked remarkably well.
Canada became globally recognized for AI research. Geoffrey Hinton, Yoshua Bengio, and Richard Sutton became foundational figures in the deep learning era. Toronto, Montreal, and Edmonton turned into respected AI ecosystems.
But there was a catch.
Canada became very good at inventing AI.
The United States became very good at making money from it. (I dug into this earlier in Industrializing AI: Canada’s Next Challenge After Research Leadership.)
That tension now seems to be driving the next phase of Canadian AI policy.
The New Strategy Is Already Emerging
The federal government hasn’t formally launched its next AI strategy, but the signals are everywhere.
Over the past year, Canada has run consultations, stood up AI task forces, opened sovereign compute discussions, launched AI safety institutes, and announced new funding programs tied to infrastructure and commercialization.
This no longer looks like a pure research strategy.
It looks like industrial policy.
The original strategy focused on talent, research, and academic leadership. The next one will probably focus on compute, commercialization, sovereignty, infrastructure, and adoption.
Put simply: Canada no longer just wants to invent the future. It wants to host it.
Prediction #1: Sovereign Compute Becomes National Infrastructure
This might become the defining issue of the next strategy.
In 2017, compute was mostly invisible. Researchers rented GPUs from cloud providers and moved on.
In 2026, compute is geopolitical infrastructure. (I mapped what I actually found when trying to build on Canadian AI compute a few weeks back — it’s thinner than the headlines suggest.)
Training frontier models takes enormous GPU clusters, power, cooling, networking, and billions in capital. Canada increasingly looks worried about dependence on foreign cloud providers, data residency, domestic inference capacity, and long-term resilience.
Expect the next strategy to lean hard into Canadian-hosted compute, regional data centre growth, public-private compute partnerships, and AI infrastructure tied directly to energy policy.
Here’s the part most people miss. This could reshape Canadian tech policy well beyond AI. Electricity grids, nuclear development, broadband, cooling, and industrial energy strategy may suddenly become AI policy issues.
Prediction #2: Canada Pushes Harder for “Canadian AI Champions”
The first strategy created research excellence. The second will try to create economic winners.
One criticism of Canada’s AI ecosystem is familiar by now. Canadian researchers invented breakthrough technologies, but the large-scale commercial value usually flowed somewhere else.
This time, expect more pressure around scaling domestic firms, procurement preferences, commercialization funding, and keeping intellectual property in Canada.
The language around “sovereign AI” is becoming more economic than technical. The government wants Canadian companies, Canadian infrastructure, Canadian data, and ideally Canadian ownership layers around strategic AI systems.
Whether Canada can realistically compete with U.S. hyperscalers is a separate question. But the policy direction is getting clear.
Prediction #3: Governance Shifts From Regulation to Standards
Canada’s proposed AIDA legislation struggled politically and operationally. Meanwhile, AI keeps evolving faster than governments can write laws.
The next strategy will probably move toward standards, certification, procurement requirements, auditability, and governance frameworks.
That approach is more practical and more flexible. It also fits Canada’s traditional role as a standards-and-trust country rather than a hyperscale platform giant.
In practice, organizations may face AI governance requirements through procurement, insurance, cybersecurity standards, funding eligibility, and industry accreditation. Not necessarily through sweeping AI-specific laws.
Prediction #4: Agents Matter More Than Chatbots
The 2023 to 2024 cycle was dominated by chat interfaces. The next cycle is about agents.
AI systems are moving from answering questions to taking actions, coordinating systems, using tools, automating workflows, and operating semi-autonomously.
That changes the infrastructure conversation completely. The next strategy will likely have to deal with agent interoperability, identity and trust systems, data permissions, secure APIs, MCP and A2A-style ecosystems, and machine-to-machine governance.
It sounds futuristic. It isn’t. Plenty of organizations are already running experiments like this today.
Prediction #5: Canada Positions Itself as the “Trusted AI” Country
Canada probably can’t outspend the United States or outscale China.
So it may try to win on a different axis: trusted AI, transparent governance, responsible deployment, public-interest AI, and democratic infrastructure.
That positioning already shows up across consultation documents and AI safety initiatives.
The opportunity for Canada may not be building the largest models. It may be becoming the safest, most trusted place to deploy them.
The Big Question
The next strategy probably won’t look like the 2017 version. Because AI itself no longer looks like 2017 AI.
Back then the question was: how do we become relevant in AI?
Now it’s: how do we hold onto sovereignty, economic value, and public trust in a world increasingly shaped by AI?
That’s a much bigger question. And this time the stakes are national infrastructure serious.
Frequently Asked Questions
What was Canada’s first national AI strategy?
The 2017 Pan-Canadian Artificial Intelligence Strategy was the world’s first national AI strategy, led by CIFAR and backed by $125 million in federal funding. It focused on retaining elite researchers, building three AI institutes (Mila in Montreal, the Vector Institute in Toronto, Amii in Edmonton), training new talent, and establishing leadership in responsible AI.
Did the 2017 Canadian AI strategy work?
For its stated goals, yes. Canada became globally recognized for AI research, kept foundational researchers like Geoffrey Hinton, Yoshua Bengio, and Richard Sutton in the country, and built durable AI ecosystems in Toronto, Montreal, and Edmonton. The gap is on the commercial side — Canada invented a lot of the underlying AI; the large-scale economic value mostly accrued elsewhere.
What will Canada’s next AI strategy focus on?
Based on current signals — consultations, AI task forces, sovereign compute discussions, AI safety institutes, and infrastructure-tied funding — the next strategy will likely move from research toward industrial policy. Expect emphasis on sovereign compute, Canadian AI champions, standards-based governance, agent infrastructure, and “Trusted AI” positioning.
What is sovereign AI compute and why does it matter for Canada?
Sovereign AI compute means GPU clusters, data centres, and training infrastructure hosted domestically and under domestic control. It matters because training and serving modern AI systems is capital-intensive, energy-intensive, and increasingly tied to national security, data residency, and economic resilience. For Canada it also pulls electricity, nuclear, broadband, and industrial energy policy directly into the AI conversation.
What happened to AIDA, Canada’s AI law?
The Artificial Intelligence and Data Act (AIDA) struggled politically and operationally and did not pass in the form originally proposed. The likely shift now is away from sweeping AI-specific legislation and toward standards, certification, procurement requirements, auditability, and sector-by-sector accountability frameworks — closer to how Canada has historically led on trust and standards in other industries.
Why is the agent era more important than the chatbot era for national strategy?
Chatbots answer questions; agents take actions. That shift forces the policy conversation onto identity, trust, data permissions, secure APIs, machine-to-machine governance, and emerging interoperability standards like MCP and A2A. National strategies that only optimize for the chatbot era will be out of date before they ship.
Can Canada actually compete with the US and China in AI?
Not on raw scale or capital. Canada can’t outspend the US or outbuild China. The realistic strategy is to compete on a different axis: trusted AI, transparent governance, responsible deployment, and being the safest, most credible jurisdiction to host and operate strategic AI systems. That’s a smaller market position than “biggest models” — but it’s a defensible one.
