AI may be fuelling U.S. business creation, but few signs of a similar trend in Canada

U.S. business formation signals are mixed: applications to start companies are up 21% since ChatGPT’s release, but U.S. business births are down about 8% to 323,000 in Q3 2025 from 351,000 in Q3 2022, while Canada’s entrant count is essentially flat at 15,730 per month versus 15,687 pre-ChatGPT. The article argues AI may be lowering startup costs, but evidence remains inconclusive and may also be reflecting paperwork, remote work and pandemic-era effects. Canada appears to be lagging the U.S. in AI-driven startup activity amid weaker venture funding and founder outflows.

Analysis

The key market signal is not that AI is creating a broad startup boom; it is that AI is selectively lowering the cost of prototyping, but not yet the cost of scaling, distribution, or survival. That matters because the investable winner set is likely to be infrastructure and workflow enablers rather than the average seed-stage software company. In other words, AI may be increasing the number of “attempts,” but not yet the number of durable companies, which keeps the economic payoff concentrated in the picks-and-shovels layer.

For APOS, the implication is second-order but important: if capital formation shifts toward fewer, larger, better-networked AI-native winners, private-market returns should become more barbelled. That favors platforms with access to late-stage AI growth, data-center adjacency, and operating leverage from a small number of breakout assets, while hurting broad venture exposure that relies on a high exit-count environment. The current environment also supports higher diligence premia and lower tolerance for generic “AI wrapper” businesses with weak moats.

RY is a mild beneficiary only if the thesis proves to be sustained at the ecosystem level. More formation plus more payroll complexity eventually increases demand for business banking, treasury, FX, and credit, but the lag is long: if startup births do not convert into employers over the next 6-18 months, the earnings impact is negligible. The more immediate risk is competitive leakage to U.S. institutions and the U.S. market structure, which could cap Canadian fee pools even if nominal formation data improves.

The contrarian view is that the market may be underestimating how much AI can centralize value creation rather than democratize it. If AI lowers founder headcount requirements but raises compute, distribution, and compliance intensity, then the net effect is stronger incumbent capture and a weaker domestic startup base in smaller markets. The reversal catalyst would be a visible step-up in Canadian AI adoption and financing, not just more incorporation paperwork.

Key Decisions for Investors

• Stay underweight broad Canadian venture exposure for the next 6-12 months; prefer selective late-stage AI infrastructure names over seed-stage software, as the data still favors concentration of winners rather than a true formation boom.
• For APOS, bias toward exposure to AI-adjacent growth equity/late-stage private assets and reduce capital committed to early-stage software managers with low follow-on reserve quality; expect a barbell outcome over the next 12-24 months.
• Long RY only as a lagged beneficiary: use weakness to accumulate on a 3-6 month horizon if Canadian business formation converts into payroll growth; otherwise treat it as a low-beta hold, not a near-term catalyst trade.
• Pair trade idea: long U.S.-listed AI infrastructure beneficiaries vs. short a basket of Canadian VC-sensitive financials/exchanges on a 6-12 month view, on the thesis that value capture migrates south even if Canadian incorporation activity stabilizes.
• Option structure: buy 12-month calls on leading AI infra beneficiaries and fund partly by selling upside in generic software names with weak moats; the risk/reward favors scarce compute and distribution over commoditized AI application layers.