Canadian Sovereign AI Infrastructure Act

⚠ This is a flightplan, not a commitment. Flightplans are hypothetical policy proposals modelled in the Ducklings causal simulation. They become projects only when voted into adoption through Consensus. Metrics shown reflect simulation model outputs, not real-world outcomes.

$3.50B CAD

Fiscal Impact

Long-term (5+ years)

Time Horizon

Medium–High

Implementation Risk

Epoch 95

Continuum Entry

Overview

Establish the Canadian AI Compute Corporation (CACC) — a Crown corporation capitalizing 300+ exaFLOPS of sovereign AI compute infrastructure across three hydroelectric-powered facilities in Gatineau QC, Winnipeg MB, and Prince George BC. Mandate federal workload migration, provide subsidized access for Canadian researchers, and create a Five Eyes secure compute export tier. Canada's cold climate and hydroelectric surplus give it a structural energy cost advantage over US hyperscalers that cannot be replicated.

Problem Statement

91% of Canadian AI workloads run on US-headquartered hyperscaler infrastructure (AWS, Azure, Google) subject to the US Cloud Act and FISA Section 702 — meaning the US government has legal access to Canadian government AI workloads without Canadian court involvement. Bill C-27 (AIDA) has zero enforcement teeth if compute is offshore. Canada has passed an AI regulation law it cannot enforce. Meanwhile, Canada's 20+ TWh/year of uncommitted hydroelectric surplus sits unused while data centres in Arizona consume water and burn coal. The energy cost moat — QC hydro at $0.04-0.06 CAD/kWh versus US grid average at $0.08-0.12 — is Canada's structural AI advantage. It is not being used.

Proposed Approach

Pillar 1 — Crown AI Compute Corporation (CACC): $2.1B capitalization over 4 years. Facilities: 200MW Gatineau QC (150 exaFLOPS, Hydro-Quebec surplus); 80MW Winnipeg MB (60 exaFLOPS, Manitoba Hydro surplus, co-located with Health Logistics Hub H-02); 120MW Prince George BC (90 exaFLOPS, BC Hydro surplus, Pacific Rim connectivity). Pillar 2 — Sovereign AI Access Program: $800M over 5 years for subsidized compute credits to Canadian universities, research hospitals, AI startups, and provincial governments. Pillar 3 — Federal Workload Migration Mandate: All Tier 1 federal AI workloads (personal data, health, national security, judicial) migrate to CACC within 5 years; Tier 2 within 8 years. Pillar 4 — Five Eyes Export Tier: Premium-priced secure compute access for allied intelligence partners outside US jurisdiction, generating $300-600M/year in export revenue at steady state.

Anticipated Impacts

sovereign_ai_compute_capacity rises from 0.5 to 300-400 exaFLOPS; data_sovereignty_index rises from 28 to 55-70 by Year 6; foreign_ai_data_processing_share falls from 91% to 60-70% as federal workloads migrate; hydro_surplus_utilization_ai rises from 0% to 18-25%; fdi_inflows +$3-6B/yr from allied partner investment; exports_to_us +$0.8-1.5B/yr in AI services; gdp_growth +0.15-0.25%; ai_datacenter_energy_consumption rises to 18-22 TWh/yr total as sovereign capacity attracts additional workloads.

Ducklings Simulation

This proposal is active in the Ducklings causal simulation (Epoch 95). The simulation models downstream effects using a BFS cascade engine with strength-weighted, time-delayed edges capped at 3-hop depth and ±25% per-hop limits. Cascade outputs are bounded by variable saturation thresholds.

Domain: National Defence | Proposal ID: 187 | Series: E-series

How to Engage

Discuss this flightplan in the Pond forum under National Defence. Vote on adoption through Consensus. Adopted flightplans become projects with real-world implementation tracking.

Contact: [email protected]