Foreign-controlled firms hold 14% of the corporate assets in Canada. They perform nearly 40% of the in-house business R&D. (Statistics Canada, 2023 data, released Oct 2025)

The foreigners we're worried about do more research in this country than we do. Hold that number. Everything below runs into it.

Canada has written more sovereign technology policy in the last eight months than in the previous twenty years.

Buy Canadian took effect in December. The first Defence Industrial Strategy landed in February, alongside a Sovereign Technology Alliance launched with Germany. April brought an $890 million competition to build a Canadian-owned AI supercomputer, the largest piece of a $2 billion compute strategy. In June, Mark Carney launched AI for All, with sovereignty as one of three organizing principles.

That deserves credit. For twenty years Ottawa talked about innovation, which is a word about activity. It now talks about control, which is a word about power. The diagnosis got much better.

The execution has two problems, and they compound.

The first is focus. We've attached the word "sovereign" to AI, quantum, space, defence, critical minerals, biomanufacturing, telecommunications, energy, advanced materials and health data. Taiwan picked one layer of one industry and made itself impossible to route around. We named nine sectors and a supercomputer. Naming everything is the same as naming nothing, only more expensive.

The second is demand. Four in five Canadian businesses still don't use AI in producing goods or delivering services (StatCan, Q2 2026). Among the firms with no plans to adopt, 78% say it isn't relevant to what they make or sell (StatCan, Q3 2025). That's not a dead market. It's an unsold one.

So we're building the supply side of a sovereignty strategy in a country that doesn't buy technology. And the instrument we've picked to fix that is a domestic price preference in federal procurement, which rewards being Canadian rather than being good.

They fail differently, which is why the word is doing so little work:

A server in Mississauga is not sovereign if the control plane, the licences and the keys sit in Seattle. That distinction decides whether $890 million buys infrastructure or a very expensive lease.

The pattern is boring and consistent. Each created a buyer with a hard problem, let companies fight over who solved it, then made them export. Not one started by funding a building.

Credit where it's due. Build-Partner-Buy commits us to building at home first, allies second, off-the-shelf last, and it carries a checkable number: 70% of defence acquisitions to Canadian firms within a decade. A real target, publicly stated, with a date on it. More of that.

Canadian businesses spend 1.06% of GDP on R&D. Israeli businesses spend about 5.9%.

That is not a gap. That is a different sport.

It puts us second-lowest in the G7, ahead of only Italy, and roughly $29 billion a year below the OECD average (OECD, March 2026). Canada's total R&D intensity was 1.86% of GDP in 2000 and 1.79% in 2024. Flat, and we're the only G7 country that didn't rise. The UK went from 1.6% to 2.8% in the same window.

A generation of innovation policy, and the line went sideways. Which brings us back to the number at the top. The foreign-controlled subsidiary in Mississauga out-researches the Canadian company down the road, and has for years.

So let's be honest about whose problem this is. Ottawa didn't decide that Canadian firms would rather buy a competitor than build a product line. The federal government can fix demand, compute and procurement. It cannot make Canadian companies want to invent things.

That one is on us, and no supercomputer touches it.

A policy that pays extra for nationality will get you nationality. It will not get you innovation. Give a company a 10% price advantage and a captive federal buyer and you've handed it a business model whose only moat is a procurement rule. It doesn't need to beat the American, European or Asian product. It needs to be Canadian and adequate.

We've run this experiment. It was called the branch-plant economy: tariff walls produced Canadian operations making smaller, costlier versions of American products for a captive market. Real companies, real payroll. When the walls came down, most didn't survive, because being domestic was never a capability.

A weaker Canadian copy of a foreign product is not sovereignty. It's a dependency with a maple leaf on it.

1. Make export the condition of the preference. The domestic advantage should sunset per company after three to five years, and renewal should require a paying customer outside Canada. Israel's export discipline came from a market too small to hide in. Ours is big enough to hide in, so the policy has to force it.

2. Weight Canadian value-added toward IP and R&D, hard. The policy already counts both. Use it. A bidder whose Canadian content is "we have a Toronto office" should lose to one whose Canadian content is "we own the patents." Right now those can score alike. That's the whole ballgame.

3. Cut the list to five missions. Name them, fund those, and move everything else to "buy from allies." A hierarchy that says what we won't build is worth more than one that says we'll build it all.

This becomes real when a Canadian department buys something a Canadian company built, runs it without the vendor in the room, and that company then sells it to a foreign government. Not when the supercomputer opens.

I want to track that second number: federal customer, then foreign customer. If you've done both, or you're close, reply and tell me. I'll name names in the follow-up, and I'd rather the list be long.

If you've got the federal-customer-then-foreign-customer story I asked about above, come tell me in person and connect with fellow founders at our next Together Toronto. Monday, August 10. Members get in for free.

We are hiring an Operations Coordinator. Learn more below:

Have a job you want to share? Currently looking to find your next gig? Check out the full list of jobs we curated just for you on our job board.

Powered by GuruLink

​Book your all-access SAAS NORTH AI pass and pay nearly 40% less than the online price. Same conference, same two days, a lot more in your pocket.

If you’ve been showing up to TechTO events for a while or loving our content, membership is the next layer - ~$20 dollars a month. TechTO members receive complimentary access to members only digital community, monthly hosted Together events, quarterly industry vertical events like Health, Commerce, Sales and networking socials.

Annual: $249/year | Founding: $1,199 One-time payment, lifetime access

Want to showcase your company, events, and opportunities to thousands of tech leaders, professionals, and investors across the country?

Contact us about potential partnership and advertising opportunities