Why Modern AI Data Centres Matter for Canada

Modern high-density AI facilities are rapidly adopting closed-loop liquid cooling. In a closed-loop system, the same fluid circulates continuously through servers and heat exchangers — it is not consumed or discharged. The facility's water circuit is filled once at construction and then recycled indefinitely.

Microsoft's newest AI data centres use a zero-evaporation closed-loop system that Microsoft's CEO says uses as little water annually as a restaurant — compared to millions of litres consumed by older evaporative cooling towers. Oracle is deploying the same approach across upcoming facilities in multiple U.S. states, and expects zero-water-evaporation cooling to be the industry standard for new builds by late 2027.

In colder climates like Canada, operators also rely heavily on free-air ("economizer") cooling for much of the year, further reducing water demand. Canada's climate is a direct competitive advantage in building low-water-intensity AI infrastructure.

AI data centres do require substantial power. Globally, the International Energy Agency (IEA) estimates data centre electricity demand will nearly double between 2022 and 2026, reaching 650–1,050 TWh annually. But the way large Canadian facilities obtain power is changing significantly.

Because grid connections for hundreds of megawatts can take years, many large projects now use a "bring your own power" model — on-site generation, dedicated power purchase agreements, and separate regulatory treatment — specifically to avoid burdening local residential feeders.

Canadian provinces are also setting hard limits. Ontario now requires Ministerial approval before large data centres can connect to the grid. British Columbia has capped AI data centre allocations at 300 MW for a two-year period starting in 2026. Alberta's grid operator capped new large load connections at 1,200 MW through 2028. These caps are designed to protect grid reliability and manage costs for existing customers.

The energy footprint of AI is real and must be taken seriously. But framing it as an unmanageable disaster ignores the pace of renewable energy adoption and efficiency improvement in the sector.

Globally, data centres consumed approximately 415 TWh of electricity in 2024 — roughly 1.5% of world electricity. The IEA projects this could rise to up to 4.5% of global electricity demand by 2026 in the highest-growth scenario. However, data centre CO₂ emissions are projected to remain below 1.4% of global CO₂ by 2030, partly because so much of that electricity comes from low-carbon sources.

Since 2025, leading AI companies have signed more than a dozen large solar energy contracts each adding over 100 MW of capacity. Renewable energy production for data centres is growing at an average rate of 22% per year and is expected to cover nearly half of additional electricity demand from data centres by 2030. Microsoft has already reduced its Scope 1 and 2 emissions by 30% against 2020 baselines while doubling its electricity consumption, through 34 GW of carbon-free energy contracts across 24 countries.

For Canada, building AI-capable data centres is not just about hosting foreign workloads — it is about keeping Canadian data, models, and critical infrastructure under Canadian law. Without domestic capacity, Canadian organizations are forced to rely on facilities in other countries, where foreign laws can govern access to that data.

The federal government has backed this position with concrete investment. Budget 2024 committed $2 billion over five years to a Canadian Sovereign AI Compute Strategy, structured across three streams:

• Up to $1 billion for public supercomputing infrastructure for researchers and government.
• Up to $300 million in an AI Compute Access Fund for Canadian startups and businesses.
• Up to $700 million to mobilize private-sector AI infrastructure investment.

In April 2026, Canada launched the AI Sovereign Compute Infrastructure Program, committing approximately $890 million over seven fiscal years to build a large-scale, Canadian-owned AI supercomputer. In June 2026, Prime Minister Carney launched AI for All, Canada's new national AI strategy.

It is accurate that operating a fully built data centre requires fewer workers than its construction phase. This is an honest criticism that the industry should address transparently. However, the full economic picture is broader:

• Canada's AI for All strategy targets the creation of 250,000 new AI-related jobs over the next five years, with up to 90,000 work placement opportunities for young Canadians.
• The strategy projects an additional $200 billion of economic growth.
• Construction, engineering, electrical, mechanical, and cybersecurity trades all benefit from major data centre builds — and from the ongoing infrastructure ecosystem they require.
• AI compute capacity enables Canadian universities, hospitals, and startups to run research workloads that would otherwise be processed — and billed — abroad.

The Bank of Canada, in a May 2026 analysis, identified AI as a significant driver of Canada's next productivity story — but noted that realizing that productivity requires domestic access to compute. Without Canadian AI infrastructure, economic gains flow elsewhere.