The Fractured Race for Canada AI Data Center Power

It was revealed earlier this month that ISED prepared a pitch deck for AI Minister Evan Solomon to use with international investors. The deck made two bold claims about Canada AI data center power: Canada had about 337 MW of AI data center capacity; and 20 GW of projects under planning or development, a figure that would rank Canada second only to the U.S. among G7 countries. The deck promoted Canada’s stable electricity grid powered by abundant clean sources and affordable prices backed by consistent regulations.

Those claims are directionally true, but they obscure who actually controls access to electricity. That answer is almost always the province. Canada does not run one grid. It runs thirteen, each controlled by a different provincial or territorial government, each with its own rules for who connects and how much power they get. No federal minister or strategy can promise power, because the federal government does not control provincial electricity systems. This article takes a snapshot of where things stand today. It covers the status of power in Canada’s four largest provinces and what public records reveal about who is actually receiving access to that power.

Our Insight Note on this topic includes additional information. You can download it here.

Canada AI Data Center Power: Local Grid Constraints

Canada is known as an energy powerhouse. In electricity terms, Canada had about 159 GW of generation capacity in 2023 and produced about 622 TWh of electricity in 2024. The United States generated 4,453 TWh in 2023, more than seven times Canada’s output. Canada remains a net electricity exporter, but net exports fell to about 12.5 TWh in 2024, down 76% from 2022, a drop driven by drought, lower reservoir levels, and rising domestic demand.

Provincial governments hold most of the authority over electricity in Canada. Each province controls generation, transmission within its borders, and distribution. The federal government steps in only for specific matters. These include parts of the nuclear sector, electricity exports, and a small number of designated international or interprovincial transmission lines.

Ownership models differ too. Most provinces run their electricity system through a government owned utility. Some also allow smaller private utilities or independent power producers to add generation or distribution capacity. Alberta and Ontario stand apart: both rely more heavily on competitive generation markets than the rest of Canada, although the two market designs differ substantially.

This structure matters for data centers. A company cannot negotiate one national power deal in Canada. It negotiates with whichever province, utility, and market structure sits under the site it wants to build on—a dynamic I examined in my framework for assessing AI data center power and grid risk.

Four Provinces, Four Different Gatekeepers

Canada’s four largest provinces are all experiencing growing interest from AI infrastructure developers. Each has responded differently.

Ontario: Planning Before Limits

When assessing Canada AI data center power needs, Ontario faces a forecast problem more than a shortage today. The IESO’s 2026 Annual Planning Outlook projects that data centers will reach 8.6% of Ontario’s electricity demand by 2050, a forecast that rose 60% in a single year. Current interconnection interest sits near 6,000 MW. Ontario has set no fixed cap. Instead, the IESO created a Major Projects Identification Committee and, in May 2026, posted draft technical requirements for large computational loads Municipal governments are also becoming increasingly active participants as communities evaluate the local impacts of large-scale data center development.

Quebec: Selective Access to Hydroelectric Power

Quebec possesses one of North America’s most attractive electricity systems for AI infrastructure, but abundant hydroelectric resources have not translated into quick access. Hydro-Québec stopped issuing new cryptocurrency mining permits in 2018 and made that moratorium formal in 2023. In June 2025, Quebec passed Bill 69, requiring provincial approval for any new load of 5 MW or more, decided case by case. Quebec plans to make its crypto ban permanent while taking a more open stance toward AI data centers, and proposed raising rates for large data centers to 13 cents per kWh. Quebec has chosen discretion over easy access.

British Columbia: A Controlled Competition

British Columbia took the most direct approach of the three with a structured allocation process. BC made its 2022 crypto mining moratorium permanent. On January 30, 2026, the province and BC Hydro launched a competitive bidding process for new AI and data center connections, capped at 400 MW pilot allocation over two years. AI projects can claim up to 300 MW, conventional data centers up to 100 MW, and no single project can exceed 145 MW. The cap tightens further after 2028. BC Hydro expects to notify successful applicants in September 2026.

Alberta: The First Province to Hit the Wall

Alberta stands apart from every other province in how directly it courts data center investment. The province has pursued the most aggressive strategy to attract AI infrastructure investment. Its message has been straightforward: abundant natural gas, existing energy infrastructure, and a competitive electricity market position Alberta to become Canada’s leading destination for AI data centers.

In December 2024, Alberta released its AI Data Centre Strategy, a first among Canadian provinces. The document demonstrated strong political commitment, but it functioned primarily as a vision statement rather than an implementation plan. It established no measurable deployment targets, no employment targets, no dedicated budget, and no framework for evaluating success.

The practical rules governing AI data center development emerged elsewhere. During 2025 and 2026, the Alberta Electric System Operator (AESO), together with new provincial legislation governing cost allocation and large electrical loads, established the framework that now determines how projects connect, how costs are shared, and how limited capacity is allocated. In effect, Alberta announced the vision, while its technical institutions wrote the operating rules. The shift shows up clearly in AESO’s own published data. Total data center demand capacity in Alberta’s queue grew from almost nothing in early 2023 to close to 20 GW by late 2025. Then, within a single month, most of that demand moved from Active to an open ended status called ISD Under Review, the moment AESO’s cap took effect.

The Pattern Across Canada

The landscape of Canada AI data center power is defined by how provinces ration electricity to the same industry using completely different methods. Alberta set a hard MW cap after the fact, once nearly 20 GW of demand had already applied. British Columbia set a smaller cap well before its own demand grew that large. Quebec kept its process fully discretionary, deciding case by case. Ontario currently manages demand through planning and connection studies rather than explicit capacity caps.

Canada is not fundamentally short on electricity generation. It generates more power than it uses and still exports the surplus. The constraint is where and when that electricity can be delivered.

National AI Strategy Meets Provincial Reality

This disconnect becomes particularly evident when examining Canada’s AI sovereignty objectives. Canada’s National AI Strategy, released in June 2026, estimates that the country needs 5.5 GW of AI compute for domestic commercial use by 2030, a target backed by real federal programs, including a $1 billion AI Compute Access Fund and programs supporting sovereign AI infrastructure. The 20 GW figure in the ISED pitch deck describes something else. It is a list of privately proposed projects, most of them designed to serve global hyperscalers looking to build massive clusters with high-speed data center connectivity to train next-generation models. Whether those projects proceed ultimately depends not on Ottawa, but on provincial electricity allocation decisions.

Following the Megawatts

Public interconnection data rarely identify the ultimate customer behind proposed AI projects. AESO, for example, publishes project status and requested capacity but not tenants. Identifying who actually stands behind Alberta’s queue requires cross checking project names against public announcements.

That exercise produces a clear answer for Alberta’s only confirmed capacity. The province’s entire Phase 1 allocation of 1,200 MW went to two projects. The larger one, 970 MW, belongs to the Greenlight Electricity Centre, a gas plant built by Pembina Pipeline, Morgan Stanley Infrastructure Partners, and Kineticor Asset Management. Meta confirmed in July 2026 that it is the customer. The smaller one, 230 MW, belongs to TransAlta’s Keephills site, developed with CPP Investments and Brookfield. No tenant has been named yet. TransAlta’s own chief executive said in February 2026 that the company hopes a hyperscaler eventually shows up. Thus, 81% of Alberta’s confirmed allocated AI capacity has already been committed to a named U.S. hyperscaler, while the remaining 19% has no publicly announced customer.

This does not diminish the economic value of these investments. They bring construction, infrastructure spending, and long-term employment. But they also illustrate an important policy distinction: attracting AI infrastructure investment is not necessarily the same as securing domestic AI computing capacity.

A Broader National Pattern

Recent announcements outside Alberta reinforce the same theme.

TELUS is developing a multi-site sovereign AI data center program in British Columbia under the federal Enabling Large-Scale Sovereign AI Data Centres initiative. The initial deployment begins around 85 MW and is expected to expand toward 150 MW by 2032. It remains unclear whether this federally supported project draws from BC Hydro’s separate 400 MW competitive allocation or operates under a different approval process.

Bell has announced a 300 MW AI data center development in Sherwood, Saskatchewan. The facility’s anchor customers are Cerebras and CoreWeave, both U.S.-based AI computing companies, with CoreWeave expected to lease approximately 140 MW.

These examples highlight the broader reality. Federal programs can support investment, define national AI objectives, and finance sovereign computing initiatives. They cannot determine which projects receive electrical capacity or who ultimately leases it.

Concluding Thoughts

Canada possesses enormous advantages in the global AI infrastructure race: abundant energy resources, political stability, clean electricity, and a growing pipeline of AI data center projects. Yet Canada’s AI strategy rests on an inherently decentralized foundation. The federal government can establish national AI ambitions and provide funding. Provinces and the electricity institutions they oversee determine how, when, and to whom new electrical capacity is allocated. That jurisdictional gap is becoming the defining factor shaping Canada AI data center power allocations and long-term market growth. The race to build AI infrastructure in Canada is being decided by the provincial interconnection queues.