Artificial intelligence (AI) is increasingly being used in our workplaces, institutions and daily lives.
Reflecting this, the federal government is encouraging further adoption of AI systems. Ottawa has its first-ever minister of artificial intelligence and digital innovation, Evan Solomon. Prime Minister Mark Carney’s single mandate letter to all his cabinet ministers speaks of deploying AI “at scale” while last year’s budget pledged $2 billion in investments for domestic AI companies.
Yet, despite all these measures, Canada may actually be behind the curve. Many countries are rushing to develop AI industries, seemingly at any cost. The United States has even framed the development of large, general AI models as a “race” it is determined to win.
This poses the question: Is this AI race one Canadians have a chance of winning? Do we even want to be in the running?
In our view, Canada simply can’t outcompete American companies with vastly greater sums to invest. Instead, the federal government should steer our domestic AI industry toward a green and ethical niche. Creating a different Canadian AI future is an opportunity to offer something to the world that is usefully different from the American approach.
The American advantage
Recent executive orders signed by President Donald Trump promise to fast-track power plants and data centres to service AI despite their negative environmental impacts and to restrict the export of unbundled AI components needed by companies looking to develop their own products – all to give U.S. businesses an advantage.
American AI companies, already among the largest in the world, are spending so much on data centres in 2025 that this investment is now contributing more to U.S. economic growth than consumer spending. Meanwhile, anti-discrimination regulations are not being written for fear they would slow down industry momentum.
Realistically, a made-in-Canada approach must look different.
Canadian AI companies are not as established or as wealthy as their American competitors.
Moreover, our treaty and confederation systems do not allow us to simply develop projects on Indigenous and provincial lands without proper impact assessments and meaningful consultation (although Bill C-5, the One Canadian Economy Act, enables the federal government to fast-track projects deemed to be in the national interest, which could include data centres).
Canada’s next great project must be to build an AI nation
Canada needs a national strategy for generative AI in higher education
Trust in government a key factor in acceptance of greater federal use of AI
Generative AI and deeper thinking: What’s in our heads still matters*
Regulations and consultations are ultimately a good thing. They provide an opportunity for Canadians to ask what our AI goals are and if they are aligned with public rather than private good.
As a group of academics specializing in the study of AI’s social and environmental justice dimensions, we argue that the Carney government has an opportunity to shape a new direction for Canadian AI, building a system where Canadians can offer more sustainable and accountable AI systems and infrastructures to support them.
Through clear regulations and creative policy, Canada could distinguish itself on the global stage and avoid an impossible-to-win competition.
The unwinnable race
The current bigger-is-better model assumes the AI race will be won by the companies that create the most powerful general-purpose tools. The hope is that with enough (often stolen!) data, these systems will eventually develop to the point they can be sold for use in a wide range of contexts, earning monopoly profits.
However, this is an expensive gambit for businesses and dangerous for the global climate. Currently, to achieve even a modest increase in AI system performance, companies need to invest in exponentially greater amounts of computational power. This approach clearly rewards players who already have a lot of money or the most access to (consumer) data.
It also privileges those who can secure energy contracts to power new data centres. These energy costs have non-trivial climate and health impacts, especially when the rush to bring more data centres online is prolonging the use of coal and gas.
This is why so many tech companies have retreated on their climate targets and why there’s a growing link between data-centre expansions and spikes in consumer power bills.
This AI arms race is ultimately a gamble. Today’s general purpose AI models have continually failed to deliver on their promise of automating skilled service work or eliminating enduring biases.
Additionally, efforts to use AI to find efficiencies in government services have often failed. Elon Musk’s DOGE disaster is a vivid example. Canada also has experienced canceled projects and failed attempts to make the government more efficient through automation. Adding fuel to this fire is the recent report from MIT indicating that 95 per cent of generative AI companies are getting zero return on their investments.
While some hope that time and money will eventually solve these problems, the current AI hype could prove to be a bubble – or even another subprime crisis – with potentially devastating consequences.
Going small to get ahead
An alternative approach is to create small AI models, built with high-quality data to do specific tasks in specific contexts. These systems are better at dealing with sensitive data and potential bias. They also use less energy. This means there are more opportunities to power these systems with low-carbon electricity, thus avoiding compromising national climate targets or raising domestic consumer prices.
This scale of AI development is also less dependent on cutting-edge American chip manufacturers or cloud providers, thus offering a better fit for Canadian sovereignty concerns. This could allow a domestic AI industry to “buy Canadian” across more of its supply chain and to avoid running its systems on American data centres.
Given that American laws now allow the U.S. government access to data stored on American-owned systems even if they are located in Canada, this is an urgent problem. Small and fit-for-purpose AI would also keep Canadian investment dollars within Canada.
Globally, there could be a growing market for smaller contextual systems, especially those that are not linked to the American state and military. Drawing on Canada’s history in peacekeeping, emergency response and environmentalism, there is an opportunity to find a green and ethical niche in international markets.
How to support small and sovereign AI
To better shape the prospects for made-in-Canada AI, the federal government should explore a range of regulations and policy measures. Some of these include common sense, unambiguous, bright-line rules to prevent bad actors and outcomes.
For example, AI should never be used to facilitate discrimination in housing or job markets, nor to take advantage of workers and consumers through predictive pricing algorithms. Models trained on stolen data shouldn’t be able to turn a profit.
While Canada has shown international leadership on regulating automation inside the government, we lack policies for private-sector AI. Ensuring that AI regulations are in place won’t slow the industry. Instead, they set boundaries and goal posts that point businesses in a direction where they can find success and promote social benefits over the longer term.
Additionally, as the Carney government explores potential new federal infrastructure megaprojects, there is an opportunity to set up data-centre parks provisioned with clean energy and governed with local and Indigenous communities.
Long-term, participatory planning can ensure that power prices and carbon emissions don’t spike, while benefits stay in the communities that host the data centres that make up the cloud. This might also involve creating a Crown corporation to provide affordable, public and sovereign computing resources to both domestic companies and public actors.
When the world’s powers square off in opposing directions, Canada has historically found a third way. With AI increasingly looking to be the next such contest, we have a lot to gain from walking our own path and sticking to Canadian commitments to human rights, environmentalism, deliberative governance and Indigenous reconciliation.
Sarah-Louise Ruder and Olivia Doggett are contributing co-authors on this article. We also thank the authors of the Challenges and Opportunities for a Made-In-Canada Approach to AI whitepaper, whose arguments are reflected in this piece.
