A recent article by a thought leader in this space named the constraint directly. The biggest problem currently sits with the grid and the utilities, bogged down by connection requests from data centres that do not intend to start. The test in that piece is a construction commitment: if a new development cannot commit to a shovel in the ground within six months, it is speculation.

That is a reading of a filing, not a clause in the tariff. Alberta has not published a six-month shovel rule for large-load connections. Methodology does not score one. The test is still the cleanest question to put to a new request.

Study capacity is the scarce object

The published gap is already large. As of 30 July 2026, about 19,565 MW of large-load requests sat against about 12,800 MW of 2025 maximum demand. AESO Phase I allocated 1,200 MW and stopped: 970 MW to GLDC for Sturgeon / Meta, and 230 MW to Keephills, now the Boreal Technology Hub. The longer comparison is in how much of the power queue is likely to be built.

Each of those requests is also a claim on staff time. Studies, models and queue administration land on the same planning desks that have to clear projects which will actually energize. A filing from a sponsor that does not intend to start still occupies that work. The megawatt total is the visible symptom. The bog-down is the process.

Public materials do not sort the 19,565 MW by construction commitment, so this article does not estimate how much of the book fails a six-month test. What they do show is that allocated load is a small fraction of requested load, and that most names in the tracker remain proposed. As of 13 October 2026 the tracker lists 37 projects: four under construction, one approved, 29 proposed and three inactive. Announced capacity of 17,900.6 MW mixes facility power, computing load, on-site generation and connection requests. Those figures are not a forecast of who breaks ground.

Six months is a screen for a new development

Apply the test before treating a new connection request as a project. Can the sponsor commit to a shovel in the ground within six months? If they cannot, the filing is speculation: a claim on the connection process, ahead of a start.

The screen asks about commitment, not about the size of the rendering. A development can publish a large megawatt figure and still have no date on which someone will turn soil. That figure will appear in announcement totals. It should not be underwritten as near-term load.

Power Certainty remains what the score already says it is: allocation, interconnection and credible generation. An AESO request is not an allocation. The six-month line answers a prior question: whether the request deserves to be read as a development that intends to start.

A request, a contract and a shovel are different facts

The public record can show process. It cannot show that a named sponsor does not intend to start. That inference belongs to the thought leader’s test. This article will not turn it into a finding against a named row.

Three facts are enough to keep the screen honest:

  • Goldfinch carries 950 MW as AESO connection request P3102. Land-use progress is real. The request is still a request. Nothing in the public file is a commitment to break ground within six months.
  • Boreal Technology Hub, at TransAlta’s Keephills site, holds the signed 230 MW Phase I load contract. BGRE leads development for Brookfield and CPP Investments, and Parkland County has not yet received a development-stage application. That is a contract without a hall. It is a different object from a speculative connection request. A shovel test applied carelessly would flatten the two. The queue problem the thought leader described is the second one.
  • The Highly Credible set is already past the question. Meta is approved, with site preparation underway and vertical construction still ahead. eStruxture CAL-3, Bitdeer Fox Creek, AHI Red Deer and K-Tech / Aurora are under construction. Prairie Sky Strathmore is no longer in that set: its Q2 2026 opening target passed and public registries now list it as proposed. The shortlist is in projects most likely to proceed.

Withdrawn requests are the clearest public proof that a filing can occupy the process and then leave it. Crusoe Crossfield / Lone Pine withdrew the municipal proposal, and the related AESO requests were cancelled. A cancelled request is study work the queue no longer has to do. The filings that never cancel, and never start, are the ones bogging the utilities down.

What to do with the test

For a new Alberta development, ask the six-month question before its megawatts enter a pipeline view. If the sponsor cannot commit to a shovel in the ground within six months, read the connection request as speculation. It may still be a land position, a rendering, or a placeholder in a utility study. It is not evidence that load is coming.

Keep three cuts separate. Announced MW mixes different bases. Phase I allocation is 1,200 MW. Construction evidence is the Highly Credible set. A request without a near-term shovel commitment belongs with the first cut.

The same line will mis-classify some real projects that are still in permitting, generation hearings or definitive agreements. Beacon’s campuses, and Boreal Technology Hub, are proposed for reasons that have nothing to do with a placeholder interconnection filing. Use the test on new connection requests that lack a start commitment. Leave pre-construction names that already hold permits, hearings or a signed load contract in their own category.

This article does not name a megawatt total that will be withdrawn, and it does not forecast which requests AESO will study next. Utilities spending their connection process on developments that will not start is the current grid problem. A new development that cannot commit to a shovel within six months should be read as speculation.