Photo: Brett Sayles / Pexels
Ask an AI assistant a question on your phone and the answer seems to come out of nowhere.
It doesn’t. Your question travels to a building full of specialized computers, often hundreds of kilometres away. Those machines do the work, and the answer travels back in seconds.
That building is a data centre.
AI is making these facilities more visible because some AI workloads require much more computing power than conventional digital services. At large scale, that turns a technology facility into an electricity, cooling and infrastructure project too.
Recent projects in Alberta, Saskatchewan and Nova Scotia show the scale of AI infrastructure now being developed across Canada.
Here’s the version that fits in your head.
So what is an AI data centre, actually?
A data centre is a building designed to house computers: rows of servers, the networking that connects them, and the power and cooling systems that keep them running around the clock.
Ordinary data centres already sit behind everyday life — email, online banking, streaming, cloud storage and government services.
An AI data centre uses the same basic infrastructure, but some are packed with specialized chips designed to perform enormous numbers of calculations at once. That work falls broadly into two categories:
Training is building an AI model by running it through large amounts of data. It can require concentrated computing power for days, weeks or longer.
Inference is using the finished model — every time someone asks a chatbot a question, summarizes a document or runs an AI agent.
Not every data centre is dedicated to AI, and not every AI model needs a huge facility. Some smaller models can run on a laptop or phone.
What makes large AI-focused facilities different is the amount of computing equipment they can bring together in one place. The International Energy Agency says a conventional data centre may draw roughly 10 to 25 MW, while a hyperscale AI-focused facility can exceed 100 MW.
The number to understand: megawatts
Data centres are usually described by how much power they are designed to draw, measured in megawatts (MW).
A megawatt describes power — the size of the electricity connection a facility needs at a given moment. A megawatt-hour measures how much electricity is actually used over time.
That difference matters because proposed capacity is not the same as operating demand.
Bell’s Saskatchewan AI data centre is designed for an initial 300 MW. Meta’s Alberta project received a 970-MW Phase 1 grid allocation. Those numbers describe the scale of the power connection, not electricity already being consumed every hour.
For another sense of scale, the International Energy Agency estimates that a typical AI-focused data centre can consume as much electricity annually as about 100,000 households.
At this size, a data centre is no longer just a building that uses electricity. It becomes part of how the electricity system is planned.
Why are companies looking at Canada?
Canada offers several things large data centres need.
There is access to large electricity systems, including hydro, nuclear and natural gas. The climate is relatively cool, which can reduce some cooling needs. Many regions have large industrial sites near transmission infrastructure.
There is also a strategic reason.
Ottawa says Canada still relies significantly on infrastructure outside the country for the compute, cloud and storage behind AI, and wants more capacity built here.
The Canadian Sovereign AI Compute Strategy includes up to $700 million to support commercial AI data-centre capacity and up to $1 billion for public supercomputing infrastructure.
That gives Canada a reason to attract these projects. It does not mean every project offers the same value.
Five questions to ask about a data-centre project
These questions also help explain what a large facility can mean for the community that hosts it.
1. Who pays for the electricity system?
A large facility can require new generation, transmission lines, substations or grid upgrades.
If those investments are built mainly for one project, the key question is whether the developer pays for them or whether some of the cost is spread across other electricity customers.
Canada’s Responsible Data Centre Development Principles say project-driven electricity costs should not be shifted onto households and existing businesses.
That does not mean every data centre raises electricity bills. A project that uses spare capacity and pays its fair share can add new demand without necessarily shifting its costs onto other customers.
The effect depends on what must be built and who pays.
2. Where does the power come from?
A data centre can reduce pressure on an existing grid by bringing new generation with it.
But “bring your own power” does not answer what kind of power gets built.
A facility supplied by hydro, nuclear, wind or natural gas can have very different emissions and infrastructure implications.
The source matters almost as much as the amount.
3. How much water will it use?
Computers produce heat, and that heat has to be removed.
Some cooling systems use evaporation and consume freshwater. Others use closed-loop systems that recirculate coolant and can sharply reduce direct water use.
The International Energy Agency also notes a trade-off: systems designed to reduce water use can require more electricity, while electricity generation itself can consume water.
There is no single water number that applies to every data centre. The cooling design, climate and local water supply all matter.
The useful question is what the design requires and, once the facility is operating, what is actually measured.
4. What does the community get?
Data centres can bring large construction projects, tax revenue, local purchasing and specialized jobs.
But construction employment and permanent employment are different. A facility may require thousands of workers to build and far fewer people to operate.
Communities may also weigh noise, land use, road traffic, water demand and infrastructure against those benefits.
5. Who controls it?
Putting servers in Canada does not automatically make the infrastructure Canadian.
A facility may be foreign-owned, serve customers outside Canada or depend on technology from international suppliers.
For Canada, the bigger questions are who owns and operates the facility, who can access its computing capacity, which countries’ laws may apply to the operator and the data it controls, and whether Canadian businesses, researchers and public institutions gain meaningful access.
Location matters. It is not the whole sovereignty question.
Who actually decides?
There is no single Canadian data-centre approval desk.
Ottawa sets national direction and has published voluntary principles covering local benefits, electricity costs, water, transparency and strategic value.
Provinces, utilities and electricity regulators make many of the decisions about whether large facilities can connect to the grid, under what conditions and at what price.
Municipalities control land-use planning and local permits. Indigenous rights holders may also have consultation, regulatory and economic interests depending on the project and location.
That division of responsibility is why a major announcement can still be a long way from an operating facility.
Our read
Canada has real reasons to build more AI computing capacity at home. Researchers, businesses and public institutions increasingly need access to powerful computing systems, and domestic infrastructure can provide more choice and resilience.
But a large data centre is not valuable simply because it is large.
The most useful way to think about these projects is as major infrastructure. They use shared resources, require long-lived investments and can create both local benefits and local costs.
The right questions are therefore familiar ones: who benefits, who pays, what resources are required, and which commitments can actually be measured.
A strong project should be able to answer those questions without relying on the size of its investment announcement.
Saskatchewan’s Data Centre Framework, explained
How one province turned questions about power, ownership and sovereignty into rules.
The Weekly Brief — Issue 16
Anthropic signs a lease at Nova Scotia’s proposed $10-billion AI and wind development.
The Weekly Brief — Issue 10
Ottawa sets national principles covering electricity costs, water, community benefits and transparency.
