DeepSeek, a relatively unknown Chinese AI startup, has upended expectations with a new model that rivals the best in the world – yet which it says cost only US$5.6-million to train, about a one-10th of what was previously thought necessary.Dado Ruvic/Reuters
Joël Blit is a professor of economics at the University of Waterloo, the chair of the Council on Innovation Policy, and a senior fellow at the Centre for International Governance Innovation.
AI’s revolution won’t arrive with a bang. It will seep in slowly, reshaping industries over decades rather than years. The steam engine, electricity and the internet all followed the same pattern: initial excitement, sluggish adoption, then an explosion of value once businesses figured out how to use the technology in transformative ways.
But the timeline for AI’s economic impact may have just been shortened dramatically. DeepSeek, a relatively unknown Chinese AI startup, has upended expectations with a new model that rivals the best in the world – yet which it says cost only US$5.6-million to train, about a one-10th of what was previously thought necessary. More importantly, the company made it open-source, effectively handing the technology to anyone who wants it.
This changes everything. The AI race is no longer confined to a handful of deep-pocketed players such as OpenAI, Google GOOGL-Q and Anthropic. The barriers to entry are crumbling, and control over AI models is slipping away from the few companies that hoped to monopolize them. If you assumed AI’s value would be captured primarily by model developers, DeepSeek’s success suggests otherwise.
Opinion: DeepSeek and the trillon-dollar history lesson
History teaches that the biggest winners from a disruptive technology are rarely those who create it. The internet’s early infrastructure giants, such as Cisco and Nortel, made some money – but the real wealth went to companies that used the internet in revolutionary ways: Amazon AMZN-Q, Google, Meta META-Q and Netflix NFLX-Q. These firms reimagined entire industries and, in the process, became some of the world’s most valuable companies.
AI will follow the same path. Open-source models such as DeepSeek-V3 and Meta’s Llama are making it clear that the real economic impact of AI will come from how businesses apply it, not from who builds the models. Companies no longer need billion-dollar R&D budgets to leverage state-of-the-art AI. Now, any startup or enterprise can download a powerful AI model, fine-tune it for their purposes, and deploy it at a fraction of the previous cost.
DeepSeek’s model is revolutionary not just in its training cost and accessibility but also in its operating efficiency. Estimates suggest its operating costs are an order of magnitude lower than those of comparable models. This makes AI applications that were previously too expensive to be viable suddenly profitable. As AI becomes cheaper and more powerful, businesses will find new and unexpected ways to integrate it, accelerating adoption.
Opinion: How DeepSeek, an unlikely Chinese app, disrupted western AI and sank the markets
These recent developments make it apparent that Canada remains stuck in the wrong AI paradigm. Ottawa’s latest AI investment package focused almost entirely on compute infrastructure, as if the biggest challenge was a lack of servers. Meanwhile, little to no funding was allocated to fostering AI adoption or literacy. That’s like building libraries while neglecting to teach Canadians how to read.
The real opportunity isn’t in spending billions on infrastructure – it’s in mobilizing the creativity and ingenuity of all Canadians. Adoption begins with widespread AI education, from high-school classrooms to corporate boardrooms.
With models such as DeepSeek-V3, the cost of developing AI-powered applications has never been lower. AI is now cheap, plentiful and fundamentally democratic. We must empower all Canadians to leverage AI so that it can be deployed in every sector, from health care to agriculture. We have a historic opportunity to ensure that some of the Amazons and Netflixes of the AI age are Canadian – and in the process to change our economic trajectory.
Canada must urgently refocus its AI strategy on adoption. The AI revolution won’t happen overnight, but if we keep waiting, the real winners will be elsewhere.