When a nurse in Thunder Bay opens a patient’s chart or a shopper in Halifax taps a bank card, the request travels somewhere physical — a building humming with servers, storage, network equipment, backup power and cooling systems.
Data centres are not mythical “AI factories” or innocuous warehouses, but industrial infrastructure with real appetites for electricity, water and land.
Current debate about data centres seems to offer a false binary: stop these facilities or accept them unconditionally.
There is a better option. We need to develop projects that are transparent, low-carbon, grid-responsible, respectful of Indigenous rights and that leave durable public value where they stand.
A baseline, not a permit
On Sept. 3, the federal government and the Federation of Canadian Municipalities unveiled Canada’s Responsible Data Centre Development Principles, backed by 23 industry signatories. The principles say projects should:
- Create lasting local benefits.
- Avoid shifting project-driven electricity costs onto households and existing businesses.
- Minimize freshwater use and environmental impacts.
- Provide transparent, independently verifiable information on local impacts and commitments.
- Deliver lasting strategic value for Canada.
The principles are a common national framework, a baseline of expectations that complement provincial, territorial, municipal and Indigenous processes. They are not binding federal law, a permit or an approval.
These principles are a necessary start. Ottawa acknowledges that Canada relies significantly on infrastructure abroad for computing power, cloud and data storage. If we build domestic capacity, we can strengthen our resilience, keep critical services running and support innovation here.
However, location is not sovereignty. A Montréal server run by a firm subject to foreign law may still face foreign legal demands.
Read more:
Sovereign AI? Anthropic shutdown reveals Canada’s weakness
No two facilities are alike
A facility’s impacts depend on several things: its workload (AI or conventional); IT load (the power the computers themselves draw) and how fully it runs; cooling design; water source and local water stress; electricity mix; grid connection; backup generation and location.
Power usage effectiveness (PUE) is calculated by dividing a facility’s total energy by the energy its computers use; the theoretical ideal is one. Water usage effectiveness (WUE) counts litres of water consumed per kilowatt-hour of computing energy. Carbon usage effectiveness (CUE) does the same for emissions.
Natural Resources Canada’s best-practice guide notes that PUE ignores the useful work servers perform. It also reports that only about 22 per cent of Canadian data centres report their PUE. And these ratios measure efficiency, not absolute local impact or public value. An efficient university campus, for example, can still strain a regional grid or a stressed watershed.
In Alberta, companies have requested about 19,565 megawatts (MW) of new data centre load. That is more than the province’s record hourly peak of roughly 12,800 MW. The International Energy Agency projects that global data centre electricity use will roughly double between 2025 and 2030.
Energy and water stewardship
Principles and expectations are not enforceable outcomes, however. We need to test these principles on a project-by-project basis, using data that communities, regulators and Indigenous rights holders can inspect and commitments they can enforce.
This begins with mandatory disclosure before approval. Proponents should report:
- Peak and average electricity demand, and expected load shape.
- New transmission, generation and substation needs.
- Water withdrawal and consumption by source.
- Cooling design and backup generation.
- Noise, emissions and land impacts.
We then need independent measurement, public reporting and enforceable consequences when a project deviates materially from its commitments.
Those building data centres should bear the connection, grid-upgrade and reliability costs their projects directly cause, not households or existing businesses. Clean-power rules should fit the local grid. A project should be judged against the capacity and emissions of its regional grid, not a one-size-fits-all national rule.
They should also disclose whether a company’s clean-energy claims rest on annual matching (balancing yearly totals) or hourly matching (clean supply at the time power is used). Research has found that hourly matching could substantially cut long-run system emissions in the western United States, although the cost is contested.
Water stewardship should start from basin conditions and drought planning. Cooling choices trade water against energy, so no single technology is always best. Waste-heat recovery should be assessed in contexts where a viable heat customer and infrastructure exist.
Communities must benefit
Community benefits should be enforceable and should outlast the construction of data centres. Benefits can include training, Indigenous and local procurement, research partnerships, municipal infrastructure and local access to computing power.
When a government decision could negatively affect Indigenous or treaty rights, the Crown must consult the affected Indigenous Peoples and, where needed, adjust the decision or reduce the harm. Consultation is not consent. A data centre also does not automatically trigger a federal impact assessment. Engagement should begin early and be grounded in rights.
Advocates of rapid expansion have a point. Blanket bans on data centres could forfeit economic and resilience gains, push impacts to other jurisdictions and deepen dependence on foreign infrastructure. But that argument supports conditions, not their absence. It does not excuse opaque projects, subsidized grid costs or disregard for host communities.
Canada has an opportunity to lead, if we can turn principles into proof. The question is not whether Canada should build more data centres, but what it should ask of them before they draw on public systems and local landscapes.
Jake Okechukwu Effoduh does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.
