Regina's mayor expresses excitement over proposed AI data centre near city

Bell Canada has filed to rezone a 160-acre parcel south of Regina to build a phased AI data-and-research campus as part of its AI Fabric Project, with initial construction slated for 2026 and on-site SaskPower substation planned. The proposal, supported by Saskatchewan Polytechnic and the University of Regina and waived for a pre-development concept plan by the city, would provide regional digital infrastructure, potential research partnerships (including heat-reclaiming greenhouses) and local economic benefits; outstanding issues flagged by adjacent landowners center on water sourcing, drainage and noise. Servicing is planned to be private and roadway upgrades would be at the developer’s expense; approvals and environmental/resource clarifications remain the key near-term risks.

Analysis

Market structure: Bell’s Regina AI campus (BCE exposure) creates a localized demand node for compute, power and construction — winners are telecoms with managed-AI services, hyperscaler hardware suppliers (NVDA), and data-centre REITs (DLR, EQIX) that price facility premium for AI workloads. Losers are small local agricultural landowners (zoning/water risk) and any regional hosting competitors without low‑cost power; expect pricing power concentrated in operators owning onsite power/cooling and fiber. Expect incremental power demand modest at project start (single campus over years) but signals multi-year capex cycle: 100s MW incremental demand across Canadian heartland if replicated (+5–10% provincial peak load scenario). Cross-asset: modest positive for CAD on sustained tech capex, small upward pressure on provincial muni spreads if subsidies follow; oil/gas impact minimal, but electricity forward curves in SK could reprice if project scale-ups emerge.

Risk assessment: Tail risks include community/regulatory backlash over water and noise leading to rezoning reversal or multi‑year litigation (low prob, high impact), supply-chain blowouts for GPUs (NVDA) causing project delays, or an AI regulatory clamp reducing onshore compute demand. Time horizons: immediate (30–90 days) hinge on zoning and SaskPower substation agreements; medium (6–18 months) on construction contracts and partnerships with U of R/SaskPolytech; long (2–5 years) on campus scale and reuse-of-heat commercialization. Hidden dependencies: reliable low-cost power contracts, fiber routes, water rights and university research funding; failure on any raises opex 10–30% and reduces IRR. Catalysts: municipal zoning approval (expected within 30–90 days), SaskPower interconnection agreement, Bell public roadmap updates, and provincial incentives announcements.

Trade implications: Direct plays: small, staged long in BCE.TO (Bell) for strategic exposure; hardware longs (NVDA) for GPU demand; selective REIT exposure to DLR/EQIX for real-estate capture of AI rents. Pair trades: long BCE.TO vs short RCI.TO (Rogers) if Bell wins regional enterprise AI contracts within 12 months. Options: buy 6–12 month call spreads on NVDA to express high-conviction GPU upside while capping cost; consider 9–12 month covered calls on DLR/EQIX to harvest yield while holding for 15–25% upside.

Contrarian angles: Consensus underestimates operational frictions — water sourcing and power interconnection could delay 12–24 months, so early-stage construction beneficiaries (local contractors) may see lumpy revenue. The market may overvalue immediate local economic impact; real payoff is in repeatable, national rollouts of ‘made-in-Canada’ compute capacity — trade for 12–36 months, not days. Historical parallels: telecom capex waves (early 2000s datacentres) show hardware spikes then consolidation; expect similar consolidation among operators. Unintended consequence: aggressive local incentives could pressure provincial budgets — creating political risk and valuation compression for provincials if taxpayers subsidize private compute.

Key Decisions for Investors

• Establish a staged 2% net-long position in BCE.TO within 30 days; if municipal zoning and a formal SaskPower interconnection letter arrive within 60 days, scale to 4% (target hold 12–24 months, target return 15–30%).
• Allocate 3% portfolio to NVDA via a 9–12 month 10% ITM call / sell 30% OTM call spread (1:1) to capture GPU demand tied to AI Fabric projects while capping downside; size position so max premium outlay ≤0.5% portfolio.
• Initiate 3–5% sector allocation to data-centre REITs split equally between DLR and EQIX (1.5–2.5% each) on pullback or upon publicized SaskPower substation agreement (look for agreement within 60–90 days); target 12–24 month hold and a 15–25% upside, use covered calls to enhance yield if volatility rises.
• Implement a relative-value pair: long 2% BCE.TO / short 1% RCI.TO (Rogers) conditional on no competing Rogers large-scale data-centre announcement within 90 days; unwind if RCI announces equal-or-larger AI campus or if regulatory approvals stall beyond 180 days.