Stop exporting Canada’s computing advantage one electron at a time

Electricity is valuable. Compute is a value-added export.

In 2024 Canada exported 35.64 TWh of electricity to the United States for C$3.13 billion, according to the Canada Energy Regulator. All Canadian electricity exports went to the U.S. The CER reports an average export price of C$81.42 per MWh; its published volume, value, and average-price figures do not arithmetically reconcile, so they should be read as the regulator’s reported series rather than a calculation from one another. Canada also imported 23.21 TWh, leaving net exports of only 12.43 TWh and a net export value of C$1.79 billion.

Those figures do not prove that every exported megawatt-hour could profitably run a data centre. Power is provincial, seasonal, location-specific, and constrained by generation, transmission, interconnection, and reliability requirements. They do show the scale of the raw input Canada sends south while the United States races to convert energy into the defining productive capacity of this century.

A megawatt-hour sold across the border is a commodity transaction. A megawatt-hour used in Canada can support a facility, a long-lived electrical interconnection, construction and electrical trades, property and corporate taxes, Canadian operators, domestic suppliers, and an asset that Canadian businesses can use. AI inference can then be sold globally over fibre. Bitcoin mining converts power into a globally liquid digital commodity without asking a foreign utility or cloud platform for permission.

Canada does not need to move every marginal electron to the customer. For suitable workloads, it can move the computation to the generator and move the result over fibre.

Put flexible machines near constrained power

Moving electricity is not free. Long-distance transmission requires corridors, substations, transformers, approvals, capital, maintenance, and time. Congestion can leave generation trapped behind a limited interface. Line losses also rise with distance, although losses are not the main economic argument. The larger opportunity is avoiding or deferring some network expansion by placing an appropriate industrial load where usable capacity already exists.

The strongest official articulation now comes from the G7 Energy and AI Work Plan led by Canada: strategically locate data centres according to energy ecosystems to optimize and stabilize grid use and expansion. Ontario’s IESO reaches a similar technical conclusion. It says AI training models can be located farther from users because processing happens on site and the answer travels back. That does not mean anywhere is automatically viable. Fibre, transformers, cooling, security, skilled service, and grid studies still matter.

Bitcoin mining is even more location-flexible. A miner does not need to sit beside a population centre, and its revenue is not meaningfully degraded by an extra few milliseconds of network latency. It can be modular, rampable, and interruptible. Proper contracts can turn that flexibility into a grid service: consume when generation is abundant or constrained, reduce load when households and critical industry need the capacity, and monetize sites that would not support a conventional factory.

The jobs argument should be honest—and still confident

Data centres and mining facilities are not labour-intensive in the way an auto assembly plant is. Canada should not pretend otherwise. Their employment impact is front-loaded and ecosystem-wide: civil works, concrete, steel, electrical construction, substations, transformers, switchgear, fibre, HVAC, liquid cooling, controls, security, commissioning, maintenance, repair, and power-system operation.

That is precisely the labour Canada can deploy at scale. The build-out creates a lasting stock of physical infrastructure, not just software subscriptions on a foreign invoice. The permanent workforce is smaller, but the installed asset remains, demands recurring technical service, and makes other Canadian firms more productive. A domestic AI facility also supports researchers, model operators, security professionals, and businesses building products on predictable Canadian capacity.

Do not confuse a strategy with a giveaway

Canadian households and existing employers must come first. No project should receive cheap public power while shifting grid-upgrade costs to ratepayers, crowding out housing, weakening reliability, or consuming water without disclosure. Provinces should auction scarce firm capacity transparently and evaluate the value created per unit of constrained electricity.

The policy should reward additionality and flexibility: new generation, private investment in substations and transmission, location at constrained resources, measurable curtailment, heat reuse where practical, Canadian operation, beneficial ownership disclosure, cybersecurity, and credible end-of-life plans. A hyperscale logo is not sovereignty. A Bitcoin label is not a public benefit. Projects must prove what they add.

Interties should remain for emergencies, balancing, and sales that are genuinely advantageous. But proposed new exports should face a domestic-value test. If a long-term U.S. sale and a Canadian-controlled compute project compete for the same incremental capacity, the Canadian project should receive a fair opportunity to demonstrate greater total value.

A practical Canadian power-to-compute policy

1. Map usable energy, not theoretical generation. Publish node-level ranges for available capacity, congestion, curtailment, firm versus interruptible service, fibre, water, and upgrade timelines without compromising critical infrastructure security.
2. Create a flexible-load tariff. Let qualified AI and Bitcoin operators buy interruptible power under clear curtailment rules, with no expectation of household-grade firmness.
3. Make developers fund causation. Where a project triggers a substation or network upgrade, allocate costs transparently instead of socializing them by default.
4. Prefer Canadian control and access. Score domestic operation, Canadian customer capacity, data jurisdiction, tax presence, workforce development, repair capability, and vendor portability.
5. Keep a reliability reserve. Never contract export or compute loads so tightly that provinces lose the ability to serve Canadians through drought, cold, heat, maintenance, or generation outages.

Policymakers do not need to search for the buyers alone

D-Central is ready to advise governments, utilities, regulators, municipalities, and Indigenous economic-development organizations that want to design a serious power-to-compute program. We can help identify credible buyers and operators in both AI compute and Bitcoin mining, distinguish flexible loads from projects demanding firm service, and translate power-system constraints into commercial terms the market can actually use.

The buyer interest is not hypothetical. In 2018 Hydro-Québec faced what its own quarterly report called an unprecedented number of blockchain connection requests. The Quebec government and Régie responded with interim orders, special conditions, and a selection framework; current Rate CB rules still distinguish capacity confirmed before June 7, 2018 and later allocated capacity. The institutional choice may have been understandable under the forecasts and reliability obligations of the time, but the result was clear: willing compute customers were constrained while Quebec continued serving export markets.

We should not rewrite that history as proof that every proposal was good. Many were speculative, undercapitalized, or unwilling to accept curtailment. Nor can a historical export average be compared casually with a retail tariff at a specific node. But Canada had an early signal that location-flexible computing wanted our power. We treated it mainly as a problem to contain. Eight years later, the United States is demonstrating what happens when energy, capital, and compute are treated as one strategic industrial system. Canada should reopen the door with better screening, firmer public-benefit tests, and contracts designed around the grid.

The United States asked for less dependency. Canada should oblige.

The current U.S. administration says it wants smaller trade deficits and is willing to use tariffs and threats to get them. Canada does not have to accept its diagnosis to respond rationally. We can reduce the sale of a strategic input to the United States and reduce our purchase of U.S.-controlled cloud and AI services at the same time.

That is not autarky and it is not revenge. Canada should continue trading where trade is mutually beneficial. It should simply stop exporting the raw advantage while importing the finished dependency. Use more Canadian power to create Canadian compute. Let Canadian firms buy Canadian inference. Let Canadian operators secure the facilities, maintain the hardware, own the customer relationships, and pay taxes here.

Frequently asked questions

Would ending electricity exports make Canadian power cheaper?

Not automatically. Provincial market design, generation mix, contracts, transmission constraints, weather, and demand all affect rates. The case is to compare domestic strategic value against export value, not to promise an instant bill reduction.

Can data centres really locate directly at generation?

Some can locate much closer to generation than ordinary urban loads, especially AI training, batch compute, and Bitcoin mining. Every project still needs an engineered interconnection, fibre, cooling, security, access, and a reliable operating plan.

Are AI and Bitcoin loads interchangeable?

No. AI has more demanding networking, hardware, customer, uptime, and data-governance requirements. Bitcoin mining is generally more modular and interruptible. A strong regional strategy can use each where its characteristics fit.

Primary sources reviewed August 26, 2026: Canada Energy Regulator, 2024 electricity trade summary; Natural Resources Canada, Powering Canada’s Future; G7 Energy and AI Work Plan; IESO, Large Step Loads technical paper; IESO, large computational load requirements; Régie de l’énergie, docket R-4045-2018; Hydro-Québec, second-quarter 2018 report; Hydro-Québec, Rate CB.