Canada is often viewed as an ideal destination for data centers. It offers abundant natural resources, low-carbon electricity, and more than 159 GW of installed generation capacity. On paper, power should not be a constraint.
Yet that headline number masks a more complicated reality. For large AI and cloud campuses, the challenge is increasingly not access to energy, but access to firm, deliverable capacity. Utilities across the country are grappling with rising demand from electrification, industrial development, and now AI. While many provinces produce more electricity than they consume over the course of a year, peak demand, transmission bottlenecks, reserve margins, and local grid constraints can limit their ability to serve large new loads.
As a result, Canada’s data center opportunity may ultimately be defined less by the availability of electricity than by the availability of deliverable capacity. That shift is changing who pays for infrastructure. Rather than socializing the cost of new generation, transmission, and distribution assets, provinces increasingly expect large customers to fund some or all of the capacity required to serve their loads, whether through transmission upgrades, dedicated generation, or behind-the-meter power systems.
The impact varies considerably across the country. Each province brings a different resource mix, demand profile, and regulatory framework. Understanding the Canadian data center opportunity therefore requires looking beyond installed capacity and examining what utilities can reliably deliver when a developer requests hundreds of megawatts of new load.
Canada’s Power Paradox
Canada’s challenge is not a lack of electricity, but a lack of firm, deliverable power. While many provinces generate surplus energy annually, peak demand, transmission capacity, reserve margins, and local grid limitations increasingly constrain large data center loads. Canada’s dependence on hydroelectric generation adds another layer of complexity. Unlike thermal generation, hydro output varies with water conditions and can be affected by prolonged droughts. The consequences are already visible: in 2025, Canadian electricity imports rose to 25.5 TWh while exports fell to 31.9 TWh as hydro-dependent provinces such as Quebec and British Columbia faced tighter supply conditions and rising demand. Yet hydro variability is only part of the story. In Ontario and Alberta, the core obstacle is not generating electricity, but delivering firm capacity where and when large loads require it. For hyperscale and AI data centers, annual energy production matters far less than the ability to secure reliable, 24×7 capacity. In other words, not all megawatt-hours are created equal.
| Province | Installed capacity (MW) | Peak demand (MW) | Annual generation 2025 (TWh) | Annual consumption (TWh) | Installed minus peak (MW) | Peak as % of installed | 2025 trade position |
|---|---|---|---|---|---|---|---|
| Quebec | 46,300 | 41,502 | 196.7 | 184.9 | 4,798 | 89.60% | Net importer on US trade (7.4 TWh in, 4.2 out) |
| Ontario | 42,000 | 24,862 | 173.1 | 145.6 | 17,138 | 59.20% | Net exporter (21.1 TWh exports, 3.1 TWh imports) |
| Alberta | 23,242 | 12,785 | 91.7 | 90.4 | 10,457 | 55.00% | Net exporter in 2024; 2025 not found |
| British Columbia | 18,800 | 11,300 | 65.5 | 56 | 7,500 | 60.10% | Largest importer; 14.5 TWh imported in 2024 |
| Manitoba | 6,800 | 5,111 | 26.6 | 22.6 | 1,689 | 75.20% | Exports down 42% in 2025 |
| Saskatchewan | 6,191 | 3,910 | 25.2 | 24.3 | 2,281 | 63.20% | n/a |
| New Brunswick | 4,500 | 3,394 | 12.9 | 13.6 | 1,106 | 75.40% | n/a |
| Nova Scotia | 3,000 | 2,481 | 9.7 | N/V | 519 | 82.70% | n/a |
| Newfoundland & Labrador | 8,200 | 1,770 | 37.1 | 10 | 6,430 | 21.60% | n/a |
| Prince Edward Island | 400 | 404 | 0.7 | N/V | -4 | 101.00% | Import-dependent |
Not All Provinces Face the Same Constraints
The headline numbers mask a very different reality across the country. Because electricity falls largely under provincial jurisdiction, Canada does not have a single power market. Instead, each province has developed its own generation mix, regulatory framework, planning process, and infrastructure priorities. Consequently, the ability to support large data center loads varies significantly from one province to another:
- Quebec: Quebec has traditionally benefited from abundant hydroelectric resources and substantial export capability. However, drought conditions in recent years have reduced hydro generation and periodically turned the province into a net importer of electricity. Quebec’s system is strongly winter-peaking, reflecting widespread electric heating. In response to growing power demand and tightening supply conditions, Hydro-Québec has proposed a new tariff for large data centers averaging 13¢/kWh, approximately double the current large-power rate, ensuring these facilities bear the costs associated with their substantial electricity consumption.
- Ontario: Ontario is a summer-peaking grid with a strong, diversified baseload (nuclear, hydro, gas). One of its challenges is the capacity of the transmission network. New capacity is planned in large part through additional nuclear power to accommodate projected data center and EV growth.
- Alberta: Alberta has added substantial wind and solar generation capacity and, on paper, possesses significant energy resources. However, transmission constraints and system reliability considerations have limited the ability to serve large new loads. AESO imposed an interim 1,200 MW limit on new large-load connections, which has already been fully allocated. Going forward, Alberta is increasingly encouraging data center developers to pair large loads with dedicated generation resources.
- British Columbia: British Columbia is facing growing pressure on its hydroelectric system as electricity demand rises and drought conditions affect hydro production. The province has responded by creating a competitive allocation process for AI and data center projects, limiting the initial allocation to 400 MW over two years. Developers of large facilities should expect heightened scrutiny regarding economic benefits, grid impacts, and energy supply plans. Projects that can secure dedicated generation resources may enjoy a strategic advantage.
- Saskatchewan: Saskatchewan continues to rely on coal generation while evaluating future additions that may include natural gas and nuclear power. The province recently approved Bell’s 300 MW AI data center near Regina, one of the largest data center projects in Canada (additional 900 MW planned for a later phase under BYOP). While the continued operation of coal facilities has increased the province’s ability to serve large new loads, SaskPower has stated that the decision to extend coal generation predated approval of the Bell project.
- Manitoba: Manitoba has a strong hydroelectric base and is a consistent net exporter. However, recent droughts have reduced export volumes. Manitoba Hydro is closely scrutinizing large new loads to ensure they do not compromise the province’s reserve margin.
- New Brunswick, Nova Scotia, & Prince Edward Island: The Maritime provinces generally operate smaller systems than central and western Canada and rely significantly on interprovincial energy transfers. Nova Scotia is retiring coal generation while expanding renewable resources and regional interconnections. The recently announced EverWind-Anthropic data center project illustrates the scale challenge facing the region: Phase 1 (95 MW) is expected to be accommodated on the existing grid, while Phase 2 (325 MW) will require substantial infrastructure planning and dedicated generation resources. EverWind and Anthropic have committed to funding required grid upgrades and associated infrastructure.
Provincial Policy Mechanisms
Provinces rely on three primary mechanisms to respond to large data center loads:
- Bring-your-own power and capacity (BYOP/BYOC): Alberta and Saskatchewan lead this model. Additionally, provinces like Nova Scotia and Ontario actively enforce bring-your-own-capacity and bring-your-own-generation principles, requiring developers to fund grid upgrades, transmission connection costs, or dedicated power generation.
- Rationing and gatekeeping: Ontario enforces a centralized government approval regime to vet and prioritize large load connections. British Columbia caps conventional projects at 100 MW and AI data centers at 300 MW (with a 145 MW per-project limit), while Quebec mandates a selection process for loads exceeding 5 MW.
- Targeted pricing: Quebec proposes a 13¢/kWh tariff, Manitoba relies on targeted levies, and Ontario is establishing a separate, higher rate class for data centers.
Over the longer term, Canada is shifting from a utility-provided power model toward a customer-provided capacity model for large AI and data center campuses. Utilities no longer absorb the cost and risk of serving large new loads. Instead, they require developers to fund transmission upgrades, secure dedicated power generation, or supply self-generation solutions. Consequently, Canada’s most attractive locations for AI infrastructure will not necessarily be the most populous provinces, but those willing to embrace customer-funded capacity. Saskatchewan and Alberta are increasingly aligning with this trend, leveraging policy flexibility to position themselves for the next wave of data center development.
| Jurisdiction | Current Position | Assessment |
|---|---|---|
| Quebec | Large projects compete for Hydro-Québec allocations. The grid remains the primary supply source, though high tariffs (13¢/kWh proposal) and a 5 MW selection process restrict access. Self-generation improves reliability but is not mandatory. | Grid-access controlled |
| Ontario | The August 2026 Data Centre Playbook establishes a formal vetting regime for large loads. The province actively encourages behind-the-meter generation and full developer cost-recovery, alongside creating a higher (Class C) rate class. | Grid-first; BYOP encouraged |
| Alberta | AESO has exhausted its initial 1,200 MW grid allocation. Under the June 2026 Data Centre Regulation (Alta. Reg. 117/2026), connection priority goes to dedicated off-grid "tethered" facilities. Additionally, the provincial hardware levy scales from 2% down to 0% for power generated off-grid. | BYOP explicitly incentivized and prioritized |
| British Columbia | BC Hydro enforces a 400 MW provincial allocation cap (145 MW per-project limit). While dedicated supply improves bid scoring, private self-generation faces steep environmental permitting, land-use, and BCER regulatory hurdles. | Grid-rationed; BYOP heavily constrained |
| Saskatchewan | The August 2026 Data Centre Framework explicitly requires proponents to bring their own power (BYOP), while prioritizing Canadian ownership, data sovereignty, and Indigenous equity partnerships. | BYOP explicitly required |
| Manitoba | Bill 39 imposes a 100% default electricity levy on data centres over 5 MW and crypto operations, effectively doubling power costs. Manitoba rejected off-grid gas-backed proposals and with zero large facilities operating, it deters large AI campuses in favor of smaller-scale loads. | Grid-hostile / Punitive pricing model |
| New Brunswick | The September 2026 policy framework mandates that projects "pay the costs they create" to protect ratepayers. Limited reserve margins make dedicated generation highly attractive even without a strict BYOP mandate. | BYOP encouraged |
| Nova Scotia | Large AI developments must fund connection upgrades or bring dedicated supply, demonstrated by the $10B EverWind/Anthropic wind-and-battery data center campus. | BYOP strongly preferred |
| Prince Edward Island | Import-dependent transmission infrastructure requires major data center developments to bring self-generation and dedicated transmission capacity. | BYOP effectively required |
| Newfoundland & Labrador | Surplus hydro capacity exists, but the province has not established a data-center-specific BYOP requirement or specialized allocation regime. | No specific BYOP requirement |
| Federal Government | The September 2026 Responsible Data Centre Development Principles urge developers to fund grid upgrades and private generation to protect ratepayer costs and support clean energy goals. | Supportive of BYOP |
The emerging winners in Canada’s AI infrastructure race may not be the provinces with the largest populations or the largest economies. Instead, jurisdictions that can provide incremental capacity, accommodate customer-funded generation, and streamline power procurement may attract a disproportionate share of future investment.
Concluding Thought
BYOP is likely to reinforce the role of natural gas in the near term, particularly in Western Canada, where abundant domestic resources support the development of dispatchable generation for AI data centres. Nuclear energy, while facing longer development timelines, offers a viable longer-term option, particularly in Ontario and New Brunswick, while Quebec can leverage its substantial hydroelectric resources. Expanding east-west natural gas transportation infrastructure is another consideration for meeting Eastern Canada’s growing energy needs and potentially supporting future LNG exports. These regional dynamics will also intersect with the federal government’s ambition to double Canada’s electricity system by 2050, backed by more than $50 billion in Budget 2026 commitments to expand generation, modernize infrastructure and connect provincial grids. The challenge will be to align these investments with the pace and location of new industrial and AI infrastructure demand.