By Qi Wu, senior advisor on strategy and innovation
Of the patents produced at Toronto’s Vector Institute and Montreal’s Mila — two of Canada’s crown-jewel AI institutes — only about 7 percent are held in the Canadian private sector. Roughly three-quarters have left the country. The largest foreign holders are Uber, Disney, and Nvidia.
Notice what didn’t happen. The labs didn’t leave. The work is still done in Montreal — much of it inside the Google, Microsoft, and Samsung research operations that set up there. The activity stayed. The ownership left. Value drained in place.
Canada’s innovation debate keeps missing this. Montreal is a genuine cluster. But it’s a research cluster, not an industrial one — and that difference is the whole game. A research cluster concentrates talent, papers, and ideas. An industrial cluster concentrates anchor firms, supplier depth, owned IP, and the decision rights that compound into the next product and the next company. Canada builds the first kind well. It rarely builds the second. Its firms sit as nodes on value chains that others define and capture, instead of anchoring ecosystems of their own.
So why don’t strong inputs become owned value? The usual answers name a missing piece — not enough VC, weak commercialization, too few large firms. Each is true. None is the mechanism. The mechanism: innovation inputs are complementary. They don’t add. They multiply. Talent × Capital × IP × Demand → Ownership. In a multiplication, the term closest to zero decides the result. It doesn’t matter how strong the other three are. And in Canada, the term closest to zero is demand.
That’s the part most discussions skip. Demand isn’t the most important input. It’s the missing one — and because the inputs multiply, the missing one sets the ceiling. Without a visible, committed market, capital can’t see a return path. So it won’t wait. Acquisition becomes the rational exit. The market is what gives capital a reason to be patient; Canada keeps funding the inputs and leaving the coupling to chance.
Before blaming Canadian character, look next door. Australia is the near-twin: small market, resource economy, a giant English-speaking neighbour, comparable institutions. Same syndrome — strong research, weak ownership. A near-identical structure producing near-identical results. That rules out the lazy explanation. This isn’t a national flaw. It’s a structural one — which means it can be changed.
It isn’t even new. Britain has watched Cambridge’s “Silicon Fen” produce three genuinely global firms — ARM, Autonomy, Cambridge Silicon Radio — and sell all three abroad (ARM alone went to SoftBank for £24 billion). Economists have a name for the broader pattern: the European Paradox. One honest caveat travels with it — some argue the research in these places isn’t quite as frontier-leading as the boosters claim. Fair. The point isn’t that the front end is flawless. It’s that even a strong front end leaks without the back.
If the constraint is coupling, the fix isn’t more force on the front end. It’s to supply the missing coupler. Three moves, and each one couples something:
Demand couples capital. Committed buyers — including the government as the first buyer — give patient capital a reason to wait. The honest catch: public procurement is deeply risk-averse, and getting it to buy unproven domestic innovation is the hardest link in the whole chain. No point pretending otherwise.
Recycling couples exits. Israel sells constantly and still compounds, because exit money and people flow back into the next company. Don’t copy the form — its military-industrial networks won’t transplant. Build the function: a reason for value to circulate rather than drain in one direction.
Allied procurement couples markets. Canada’s home market can’t carry a champion alone. So borrow allied demand. A coordinated Western commitment to source critical minerals midstream — the refining, cathode, and magnet steps now concentrated in China — outside China is, in effect, the launch market a single small economy can’t supply for itself. Tech sovereignty is the window that makes that demand real for the first time.
One discipline holds it together: back the position, not the company. Naming “allied critical-minerals midstream” as a position to hold is an industrial strategy. Anointing a national champion is winner-picking — the failure mode that discredits the whole enterprise. The hard case is real: in a narrow field, sometimes only one or two firms can fill the position. Keep it a position anyway by attaching support to performance, not identity — open to new entrants, milestone-triggered, time-limited. The moment support tracks a company’s name instead of a position’s results, “strategy” has quietly become picking winners.
Return to that 7 percent. It isn’t a story about weak science or stolen ideas. It’s a country that funds every input and leaves the coupling to chance — then watches ownership settle elsewhere while the work stays put. The next stage of industrial policy needs a different unit of measure. Not how much research goes into it. Not even how many firms come out. Whether the inputs ever lock together around something Canada owns. We’ve spent a decade measuring what goes in. When do we start measuring what stays — and what finally multiplies?
Qi Wu, is a Vancouver-based executive and board director with over 20 years of experience in Chinese and international industry. He writes on Canada’s industrial and innovation strategy from a cross-border perspective.
Photo: Copilot
