Toronto’s AI Data Centre Debate Is Really an Infrastructure Capacity Test
The proposed expansion of STACK Infrastructure’s data centre at 3650 Danforth Ave. is not only a neighbourhood controversy. It is a signal that Toronto’s land use framework is being forced to absorb a new class of infrastructure before the city has fully defined how that infrastructure should be governed.
As blogTO reported, the proposal would expand an existing 8-megawatt data centre into a 56-megawatt AI-oriented campus across roughly 19 acres in Scarborough Southwest. The site would add two new buildings and substantially increase electricity demand, raising questions from residents and Councillor Parthi Kandavel about power capacity, water use, noise, local infrastructure costs, environmental impacts, and community protections.
For developers and planners, the key issue is not whether data centres are needed. They are. AI, cloud computing, finance, health systems, logistics, and public services are all becoming more dependent on heavy digital infrastructure. The more important question is where these facilities belong, what infrastructure they consume, and whether existing zoning tools are precise enough to manage their impacts.
The Danforth application exposes a gap in the planning system. According to the local councillor, data centres are already permitted under the site’s existing zoning, which means the project can move through Site Plan Approval rather than a full rezoning. That distinction matters. Rezoning opens a broader political and policy debate about land use. Site plan review is narrower, focused on built form, access, servicing, landscaping, and technical matters. When the use is already permitted, the city has less leverage to ask whether the land use category itself still fits the scale of the proposed operation.
Data centres are no longer passive industrial users. At AI scale, they behave like major utility infrastructure and should be planned with that level of scrutiny.
This is where the development implications become broader than one Scarborough site. A 56-megawatt facility places a different kind of load on the urban system than a warehouse, office building, or traditional light industrial operation. It can affect local grid planning, backup power requirements, stormwater design, cooling systems, noise mitigation, and long-term municipal infrastructure coordination. Those factors all influence feasibility, timelines, capital costs, and public approval risk.
There is also a land economics issue. Large data centres can compete for scarce employment lands, especially well-serviced parcels with access to power, fibre, roads, and separation from sensitive uses. In a city already under pressure to protect industrial land while also delivering housing, transit-oriented growth, and climate resilience, every major land allocation carries opportunity cost. A 19-acre site used for compute infrastructure is not available for other employment-generating or mixed urban functions.
Toronto is working on its first city-wide approach to large-scale AI data centres, but that work is not expected to return to council until 2027. That timing creates a policy vacuum. Mississauga is considering a one-year moratorium, while Hamilton recently debated and rejected a temporary pause. The regional pattern is clear: municipalities across the Greater Toronto and Hamilton Area are being asked to approve a fast-growing asset class without a mature planning framework.
For investors, this means approval risk will increasingly depend on more than zoning conformity. Grid capacity, climate optics, community benefit expectations, energy sourcing, water strategy, and noise performance will become core due diligence items. For municipalities, the lesson is equally direct. If AI infrastructure is now part of city-building, it needs a policy framework that treats it as infrastructure, not just another permitted industrial use.
Source: blogTO
