Canada's National AI Strategy: The Right Vision, the Hard Questions, and Who's Watching
Canada launched AI for All today. The vision is right. The gaps are real. And from Kampala to Nairobi, others are watching to see if implementation matches the ambition.
This week, Prime Minister Mark Carney stood in Toronto and launched AI for All — Canada’s new national artificial intelligence strategy. Minister of Artificial Intelligence Evan Solomon was beside him.
The AI community has been waiting for this. After months of consultations, 11,000 public submissions, a 28-member expert Task Force, and draft leaks that raised expectations high, the strategy is finally here.
I work at Amii — the Alberta Machine Intelligence Institute, one of the three federally funded institutes at the heart of Canada’s AI ecosystem. This is not academic for me. This strategy reshapes the context I operate in, and it carries implications far beyond Canada’s borders.
Here is my honest read.
The baseline: understand what this strategy is correcting
Before the ambition, the reality.
Carney said it plainly: Canada ranks near the bottom of countries in AI training, literacy and trust. A KPMG–University of Melbourne global trust study ranks Canada 44th of 47 countries on AI training and literacy. Ipsos data released this week shows Canadian enthusiasm for AI has actually dropped since last year. Only 12% of Canadian businesses were using AI in 2024–25. By mid-2026, that crept to 14.5%.
That is the baseline. A country that invented the modern AI era — where Geoffrey Hinton, Yoshua Bengio, and Richard Sutton did foundational work — is barely using what it helped build.
The strategy targets $200 billion in additional economic growth and 250,000 new jobs over five years. It aims to push business AI adoption from 12% to 60% by 2034.
Those are the numbers you will hear everywhere. Hold them carefully — we will come back to them.
The three principles
The strategy organises everything around three anchors:
Build trust. Modernise privacy laws. Criminalise deepfakes and surveillance pricing. Expand the Canadian AI Safety Institute. Introduce online safety rules for chatbots and social media.
Create opportunity. Launch a National AI Literacy Initiative reaching 1 million post-secondary students. Help SMEs adopt AI. Launch an AI Missions Program starting with health diagnostics.
Reinforce sovereignty. Build Canadian-owned compute, cloud, and data infrastructure. Construct a world-class public AI supercomputer by 2031. Reduce dependence on foreign tech giants.
That last one is the centrepiece. And the most contested.
The sovereignty question — and what Western Canada is already living
A report released this week by the Canadian Anti-Monopoly Project found that three American companies — Amazon, Microsoft, and Google — control 85% of Canada’s publicly available cloud infrastructure. Well above the global average. For a country that has watched trade tensions expose the cost of economic dependence, this is not just a tech policy problem. It is a national security concern wearing a technology shirt.
The strategy’s answer is to build. Large-scale, Canadian-owned data centres. A sovereign AI supercomputer. Less reliance on foreign compute.
But here is what the press conference did not say: sovereign compute is already happening in Western Canada, and it is already generating friction.
Alberta had 11.9 gigawatts of data centre connection requests in Q1 2025 alone — more than the entire province added in the previous decade. Telus, working with the Government of Canada and developer Westbank, is building AI-focused data centres across three Vancouver sites, eventually scaling to over 60,000 GPUs and 150 megawatts of computing capacity by 2032.
This is the compute build the strategy depends on. It is real. It is creating jobs. It is also creating conflict.
Canada’s National Observer identified 42 data centre projects from Alberta’s electricity regulator’s March 2026 project list. Three-quarters of the planned sites sit in high water stress areas. In Rocky View County, a proposed 450-hectare technology park and 900-megawatt power plant was voted down 6-1 by councillors in September 2025 after community submissions citing water usage, grid impact, and agricultural land concerns. Alberta’s technology minister has acknowledged that natural gas is “really the only option for the next three to five years” to power these facilities, with carbon capture as a possible but expensive mitigation.
A March 2026 Abacus Data survey found 34% of Canadians would oppose a data centre in their community, against only 16% in favour — though the largest group, 39%, said it depends on specifics. Alberta showed the highest conditional openness at 45%.
The sovereignty instinct is correct. Renting 85% of your AI infrastructure from three foreign corporations is a vulnerability. Building at home is right. But building at home in Alberta means negotiating a genuine tension between national AI ambition and local water tables, grid stability, and community trust — the same trust the strategy says it is trying to build with Canadians about AI itself.
British Columbia is pointing toward a model: its Look West strategy explicitly ties data centre development to Indigenous consultation, local government engagement, and clean electricity prioritisation. That framework needs to go national. The strategy needs a social licence architecture for sovereign compute as urgently as it needs the compute itself. One without the other stalls.
The $200 billion question: what do the projections actually tell us?
Now back to those numbers.
History offers a calibration. The UK’s 2018 AI Sector Deal — the closest peer-nation comparator — projected AI could add 10% to UK GDP by 2030. That projection was built on assumptions of rapid, widespread adoption that largely did not materialise at the pace promised. The UK strategy was not a failure, but its headline numbers became an embarrassment to the officials who championed them.
Canada’s 250,000 jobs figure and $200 billion growth projection carry the same methodological uncertainty. Neither the strategy document nor today’s press conference specified the modelling basis — gross or net job count, what baseline, over which sectors, under what adoption curve assumptions.
Here is the underlying problem the numbers obscure: Canada’s 12% adoption rate is not a mystery. Research from Toronto Metropolitan University’s Diversity Institute and an OECD analysis produced for Canada’s own G7 presidency are specific about the structural barriers. 41% of medium-sized Canadian firms cite lack of internal expertise as a key difficulty. 58% of small firms point to affordability as their primary barrier. The consultation summary from Canada’s own pre-strategy engagement process named the same constraints directly: outdated IT systems, siloed data, risk-averse purchasing behaviour, and regulatory uncertainty.
The BDC’s $500 million fund addresses capital access. That is one constraint. The literacy initiative addresses awareness. Also one constraint. But data readiness — arguably the most fundamental precondition for AI adoption — receives no dedicated programme in the strategy. Neither does the concentration of AI talent in three cities, which means businesses in Lethbridge, Prince George, and Moncton face a different adoption reality than businesses in Toronto.
Going from 12% to 60% is not just about money and awareness. It requires a cultural and operational shift inside businesses that have spent decades building processes that do not talk to each other. The strategy’s adoption targets are achievable in theory. In practice, they require solving problems the strategy only partially addresses.
What changes for Amii, Mila, and Vector
If you thought of these institutes as university research labs, update that model today.
The strategy commits $130 million specifically for commercialisation programs across Amii, Mila, and Vector. The mandate is shifting from publishing papers to building economic engines. These institutes now formally serve as the bridge between Canada’s research depth and the businesses trying to put AI to work.
Just weeks before the launch, the government and CIFAR announced a $24 million investment to appoint and renew 42 Canada CIFAR AI Chairs. 32 of those 42 are anchored at Amii in Edmonton — confirmed independently by both CIFAR and the University of Alberta. That is a significant concentration of research leadership in Alberta and a federal signal that the Prairies are more than an energy story.
These researchers train the graduate students, post-docs, and specialists who fill the roles the strategy is projecting. The institutes are the human capital pipeline. The $130 million commercialisation mandate is a recognition that building pipeline is not enough — you have to connect it to the economy.
The trust contradiction no amount of strategy language resolves
The strategy promises to ban surveillance pricing, criminalise non-consensual deepfake imagery, and introduce age restrictions for AI chatbots and under-16s. These are the right directions.
But the actual legislation is deferred to future bills. And one bill already moving — Bill C-22 — sits as a direct contradiction inside the strategy’s own trust framework.
Michael Geist — Canada Research Chair in Internet and E-commerce Law at the University of Ottawa — called this “digital self-sabotage” in the Globe and Mail days before today’s launch. His argument: the government is promising new privacy rights with one hand while pressuring a parliamentary committee to pass Bill C-22’s mandatory metadata retention regime with the other — a surveillance mechanism that comparable countries have already struck down as a violation of fundamental privacy rights. Apple has stated the bill forces companies to insert backdoors into their own products.
His critique is structurally correct and politically inconvenient. The government has launched a forward-looking sovereignty strategy while moving a backward-looking surveillance bill through committee simultaneously. These are not separate files. They share the same trust foundation. The same Canadians being asked to embrace AI adoption are being asked to accept mandatory metadata retention.
Where I part from Geist is on prognosis. He frames the contradiction as fatal. I think it is serious but not determinative — if the government resolves it through the legislative calendar rather than messaging its way around it. The strategy has the right instincts. Parliament is where the credibility test happens.
Who else is watching — and why it matters
Here is the dimension the strategy’s domestic coverage has almost entirely ignored: Canada is not writing this document for Canadians alone.
Across Africa, 13 national AI strategies have been adopted since 2018. Uganda, where I spent two decades deploying technology at national scale, is currently developing its National AI Strategy for 2026–2031. Kenya published its National AI Strategy in 2025. Rwanda — widely regarded as Africa’s most deliberate AI governance actor — published its AI policy in 2022. The African Union endorsed its Continental AI Strategy in July 2024, with Phase I (2025–2026) focused explicitly on governance frameworks, resource mobilisation, and national strategy development.
These countries are not waiting for guidance from the Global North. But they are watching. When Canada builds something that works — a commercialisation model that connects research institutes to SMEs, a talent pipeline that moves from PhD to practitioner, a sovereign compute approach that navigates community consent — those are blueprints that developing economies study, adapt, and reference.
The problem is when Canada’s work does not move beyond strategy documents and conference proceedings. Research published in the absence of demonstrated adoption does not travel well. It gathers dust. Canada’s global AI leadership depends not just on producing world-class research and policy frameworks, but on demonstrating that they can be implemented at scale in a complex, federated economy. That demonstration is what makes Canadian experience genuinely useful to a government in Kampala or Nairobi designing their own approach.
The inverse is also true. African AI strategies are grappling with problems Canada will eventually face in different form: how to build AI capacity without the infrastructure preconditions the West takes for granted, how to govern AI in contexts where data is scarce and trust in institutions is fragile, how to ensure AI adoption serves communities that have historically been excluded from technology’s benefits rather than harmed by its deployment. Canada’s strategy has nothing to say about any of this — and its silence is a missed opportunity for the kind of reciprocal learning that genuine international co-operation requires.
Canada’s G7 leadership on AI standards this year matters because it shapes the frameworks that travel globally. But frameworks only travel when they are grounded in demonstrated practice. AI for All can be that demonstration — or it can become another well-intentioned document that other countries politely cite while building their own path around it.
The bottom line
AI for All is the right strategy at the right moment. Canada needed to pivot from being the country that invented AI to being the economy that actually uses it. This strategy makes that pivot official.
The Amii mandate expanding to commercialisation is good. The CIFAR Chair concentration in Alberta is good. The BDC fund for SMEs is concrete and useful. The sovereignty instinct — building rather than perpetually renting — is correct.
But the gaps are real. The adoption projections need honest methodology, not headline ambition. The structural barriers to SME adoption — data readiness, talent geography, risk culture — are only partially addressed. The sovereign compute build requires social licence that Western Canadian communities are already demanding. And the Bill C-22 contradiction sits like a splinter in the strategy’s credibility until the legislative calendar resolves it.
And for the countries watching from Kampala to Nairobi to Accra: Canada’s job is not just to get the strategy right for itself. Canada’s job is to implement it well enough that its story remains a valid reference and benchmark — not a cautionary tale about a country that wrote brilliant policy and then struggled to execute.
Canada invented this field. It is past time we built the economy around it — and showed the world how it is done.
The question now is whether implementation matches the ambition.
I’ll be watching.
Brian Ssennoga is an AI Policy & Governance Practitioner and ML Project Manager at the Alberta Machine Intelligence Institute (Amii). He writes at the intersection of AI strategy, governance, and real-world adoption across Canada and the African continent. He spent over two decades leading technology deployments across East and Central Africa before relocating to Edmonton in 2024.
← My Written Thoughts