The Queue Is the Policy

Commentary on "Powering AI: Canada's evolving electricity grid connection policies" · Osler, Hoskin & Harcourt

Osler's outlook, written at the end of last year, argued that Canada was moving away from first-come, first-served grid connection toward selective, policy-aligned frameworks, and called this a structural change in how large loads would reach the grid. Six months on, the prediction has landed harder than the piece suggested. Canadian electricity connection is no longer being administered. It is being rationed.

The distinction matters. A queue processes everyone eventually. A ration does not.

Alberta ran the experiment first

Alberta's system operator froze new large-load connection requests in mid-2025 after receiving proposals it could not accommodate. The scale of the mismatch is the number worth holding on to: twenty-nine proposed projects seeking more than sixteen gigawatts combined, against an interim allocation of twelve hundred megawatts for projects at or above seventy-five megawatts. Roughly thirteen times more demand than supply.

That allocation has now been fully distributed. Load contracts are executed, the successful projects have moved into the connection process, and the operator has published a guide setting out technical requirements for transmission-connected facilities. Phase two, examining the long-term framework across connections, planning, operations, markets, tariffs and reliability, is underway.

Read the sequence plainly. A public system operator received applications for thirteen times the capacity it had, allocated all of it, and closed the window. The projects that got in are proceeding. The others are waiting for a framework that has not been written yet, on a timeline nobody has published.

Quebec is doing the same thing with more candour

Quebec's approach is a competitive call that caps two-year allocation at three hundred megawatts for AI projects and one hundred megawatts for data centres generally, with a separate undetermined amount for hydrogen export. Other industrial loads, including resource development, face no such limits.

That is not a technical constraint. It is an industrial policy expressed in megawatts. The province has decided how much of its electricity this sector may have, ranked it against other uses, and written the answer down. Whatever one thinks of the numbers, the mechanism is honest about what it is.

Ontario holds the same power under legislation passed in December and has not yet published the criteria it will apply. British Columbia's framework is developing. The federal sovereign compute program runs on memoranda of understanding rather than a published allocation rule.

Rationing without criteria is just discretion

Here is the part that should concern everyone in this industry, including the people who won allocations.

When capacity is scarce and distributed by administrative process, whoever designs the process determines the shape of the sector. Not the market, not the customers, not the siting economics. The allocation rule. And in most Canadian jurisdictions that rule is either unpublished, or published only in its technical dimension while the substantive test remains at ministerial discretion.

That produces two failures at once. Applicants cannot price the risk of not being selected, which raises the cost of capital for every project including the good ones. And the public cannot audit a decision whose criteria were never stated, which is precisely the grievance driving the municipal fights in Toronto, Hamilton, and Vancouver. A community can accept losing an argument on stated grounds. It cannot accept a decision it has no way to examine.

The engineering side of this has been done properly. Alberta and Ontario have both drafted detailed technical requirements, consulted publicly, and published the results. The substantive side, which projects deserve scarce capacity and why, has been handled with far less rigour, and it is the side that carries the political weight.

What developers are now competing for

The practical consequence for anyone building in this country is a reordering of what actually determines whether a project happens.

For most of the history of this industry, a developer competed on site, power price, construction cost, and customer relationships. In a rationed system, all of that is downstream of one prior question: can you obtain an allocation. A project with perfect economics and no allocation does not exist. A project with mediocre economics and an executed load contract is real.

Which means that in a rationed system the scarce asset is not land or capital. It is a position in a queue that in two provinces is already closed, and in the others is being designed right now in system operator consultations that almost nobody outside the industry is watching.

It also means the projects best positioned are those that reduce their claim on the rationed resource. Capacity that does not draw from the contested pool, generation brought to the table rather than requested from it, and load that can be curtailed when the system needs it are no longer efficiency talking points. They are the difference between being in the allocation and being in the queue behind it.

What this means going forward

Osler was right that this is structural, and it is worth being clear that the shift is defensible. A system operator with sixteen gigawatts of requests and a grid that cannot serve them has to say no to somebody, and doing so through a published process is better than doing it through delay.

The gap is that Canada has built the rationing machinery well ahead of the reasoning that is supposed to govern it. Four provinces are now deciding which projects get scarce public capacity, and only one has written down the test. Until the rest do, every allocation decision will be contested, every unsuccessful applicant will suspect the process, and every community will assume the answer was arranged somewhere they could not see.

VOLTEDGE

Reference: "Powering AI: Canada's evolving electricity grid connection policies" · Osler, Hoskin & Harcourt · December 2025 · read the article