Canada has long been a hotbed of AI research, training and talent. Nobel Laureate Geoffrey Hinton’s pioneering work on machine learning and neural networks at the University of Toronto helped lay the groundwork for today’s AI revolution. Institutions like the Vector Institute continue to push that frontier. Yoshua Bengio, Professor at the Universite de Montreal, won The Turing Award for his pioneering work in the field. Hinton protégé Ilya Sustkever went on to co-found OpenAI, now worth $150 billion. You get the idea.
We’re great at research and our universities graduate some of the best AI talent, but as a country we have struggled to commercialize these discoveries. Yes, Canada is home to a burgeoning ecosystem of AI startups, including companies like Cohere, Arteria AI, Wabai, ADA and Ideogram, but could we be doing more?
Some of our challenges are structural. Our venture capital industry is just five percent the size of the U.S. and some of our best grads move to the U.S. and elsewhere where pay is higher. However, some can be fixed through sensible policy: we invest only 80 cents on the dollar in capital spending per capita compared to our neighbors to the south, so we should encourage investments in Canadian businesses through lower taxes and incentivize those grads to stick around.
Another challenge is the sheer scale of investment many assume is needed to compete in AI. Microsoft, Google, Meta, and even some venture-backed startups in the U.S. are spending billions acquiring GPUs and AI talent, training models, and controlling the AI supply chain.
How can Canadian companies compete when Tech Giants and their better funded global counterparts are making nation-state sized investments in hardware to train the latest and greatest AI models?
The answer: work smarter, not harder. That’s the lesson we can learn from the success of China’s DeepSeek, which built an advanced AI model for a fraction of what Big Tech spend, sending shockwaves through the market, wiping out nearly $600 billion in Nvidia’s market cap in a day. Of course, there’s more to that story and we should take any news out of China with a grain of salt.
However, it is clear now that the AI game isn’t just for trillion-dollar companies.
Indeed, Canada's Cohere just launched a cutting-edge AI model, proving that innovation and efficiency—not just brute-force compute—can win the day. As Cohere’s co-founder Nick Frosst put it, "People have been chasing the wrong rabbit in LLM development, thinking more compute is going to always lead to breakthroughs. But now folks are coming around to our point of view: innovation and efficiency, not excessive compute, is the key."
In releasing this new model, Cohere claimed that "
Command A is on par or better than GPT-4o and DeepSeek-V3 across agentic enterprise tasks, with significantly greater efficiency." Independent analysis confirms that the model performs on par with some of the systems from the best-known technology firms. BetaKit called it
“Canada’s DeepSeek Moment.”
That’s a lofty statement but perhaps we truly are on the brink of a Cambrian explosion in Canada’s AI scene. Big Tech may even be standing on a burning platform of their own making. By investing so heavily in AI, they have lowered the barrier for others to innovate, setting themselves up to be disrupted. “Big tech seems to be in charge because compute, particularly GPUs needed to train and scale AI models, is both scarce and expensive” says Tom Serres, Co-founder of Nautilus Asset Management. But that seeming lead is illusory, as access to computing power becomes easier.
Tech giants were also once plucky startups themselves, reminding us things can change quickly. Companies like Google, Netflix and Amazon harnessed the internet to take on media companies, brick and mortar retailers, and others. Emerging companies – in Canada and elsewhere - can do the same with AI.
The democratization of AI may happen faster than most expected. We have seen this play out before: as a technology scales, it becomes cheaper and much more powerful, eventually becoming ubiquitous. At one time, only corporations, governments and universities cold afford computers. To operate one, you needed to be a trained programmer. The PC and the internet made computing and access to information universal. Smartphones today are more powerful that the most expensive supercomputers of only a generation ago.
This should be a boon to countries like Canada, where we have deep pools of AI talent but shallower puddles of traditional capital, compared to the U.S. As capital and technology barriers fall, we and others can build global leaders. We should do everything in our power to encourage these innovators to do it here.
March 20, 2025
