Canadian resources should mean Canadian access
If Canada supplies the land, power, gas, water access, grid connection capacity, permits, and public tolerance for gigawatt-scale AI campuses, Canada should get guaranteed access to serious compute — not just construction jobs and a thank-you card.
If someone wants to generate fake pictures, spam content, junk videos, or another chatbot answer about nothing, that can run anywhere. That is not the national-interest argument for using Canadian land, power, gas, water access, grid connection capacity, permits, and public patience.
Canadian researchers, universities, hospitals, startups and public agencies will need serious AI compute. The useful question is whether they will be able to get it when they need it.
Every foreign-controlled gigawatt-scale AI campus should be required to sign a Canadian Compute Contract before the project application is treated as complete. This is a public-interest framework, not a final statute, but it is the right starting point.
A starting target around 5% of the project’s usable AI compute value, delivered as practical compute access, credits, funding for equivalent Canadian sovereign compute, or another auditable structure. It should not require direct access to a hyperscaler’s internal systems.
Priority access for Canadian universities, hospitals, medical research, science, agriculture, energy, climate, public agencies, Indigenous research priorities, startups, and small businesses.
This dividend does not replace property tax, local jobs, procurement, grid-cost payments, water mitigation, cleanup security, or normal regulatory obligations.
Before a foreign-controlled AI campus receives land-use support, grid connection arrangements, power contracts, water licensing, public incentives, or expedited approvals, the proponent should sign a simple public pledge.
This is not about stopping serious projects. It is about making sure Canada gets more than construction jobs and a thank-you card.
| Term | What it should require |
|---|---|
| Minimum allocation | A starting target around 5% of usable AI compute value, delivered through usable access, a value-equivalent compute-credit pool, or funding for equivalent Canadian sovereign compute reserved for public-interest and innovation users. |
| Foreign-controlled projects | Applies to foreign-owned or foreign-controlled hyperscale AI campuses, especially projects at or above 1 GW or projects seeking major public infrastructure support. |
| Equivalent capacity option | If direct access to the operator’s internal platform is impractical, the proponent must fund equivalent Canadian sovereign compute through an approved Canadian provider or public compute trust. |
| Canadian Compute Trust | The allocation should be administered independently, with transparent eligibility rules and representation from research, medicine, public-sector, startup, Indigenous, and provincial/federal interests. |
| Useful compute only | Credits must be usable for serious workloads, not leftover scraps, obsolete hardware, artificial waitlists, marketing trials, or capacity that cannot realistically be scheduled. |
| No data grab | Canadian users keep control of their data, intellectual property, research outputs, and confidential information unless they explicitly agree otherwise. |
| Annual audit | Publish total available capacity, Canadian allocation, used credits, unused credits, waitlists, sector distribution, and delivery shortfalls. |
| Use-it-or-fund-it | If the project cannot deliver the promised allocation directly, it must fund an equivalent pool of Canadian compute capacity elsewhere. |
| Approval condition | The pledge should be filed before the application is accepted as complete, then converted into enforceable conditions in development agreements, power arrangements, benefit agreements, or licence approvals where possible. |
Once the power contracts are signed, the land is rezoned, the water licence is issued, and the campus is under construction, governments lose leverage. The compute dividend belongs at the front of the process.
Foreign-owned hyperscale AI campuses can still be welcome in Canada. But welcome does not mean free. If the project needs Canadian resources, the Canadian benefit needs to be written down before the ribbon cutting.