Two provinces have now written grid scarcity into regulation. The read-through for developers is not what most of the market assumes.
Alberta's system operator is processing roughly 20.7 gigawatts of data centre connection requests. The interim framework it introduced allows up to 1,200 megawatts of large load to connect between now and 2028, and that allocation has already been assigned. For scale: the entire City of Edmonton draws around 1,400 MW. The requests exceed the available capacity by a factor of roughly seventeen.
Two provinces have now written grid scarcity into regulation. The read-through for developers is not what most of the market assumes.
The number that reframes everything
Alberta's system operator is processing roughly 20.7 gigawatts of data centre connection requests. The interim framework it introduced allows up to 1,200 megawatts of large load to connect between now and 2028, and that allocation has already been assigned.
For scale: the entire City of Edmonton draws around 1,400 MW.
So the requests exceed the available capacity by a factor of roughly seventeen. This is not a queue in any meaningful sense. A queue implies that if you wait long enough, you reach the front. What Alberta has is a closed allocation with a very long line outside it, and no published commitment on what capacity might open up beyond 2028.
None of this reflects an unwillingness to serve the industry. Alberta's government has been among the most enthusiastic in the country about attracting data centre investment. The constraint is physical: large loads of this character behave in ways the grid was not designed to accommodate, and the generation and transmission required to serve them at this scale has not been built yet. The system operator is being responsible, not obstructive.
British Columbia took a different route to the same place
On 1 February 2026, BC's Data Centre and Hydrogen Production Facility Power Supply Regulation came into force. Rather than manage scarcity through interconnection process, BC legislated it directly.
The regulation caps new electrical capacity that BC Hydro may make available for data centre purposes: 100 MW per two-year period for conventional data centres, 300 MW per two-year period for AI data centres. These are system-wide aggregate limits, not per-project ones.
The definition of an AI data centre is deliberately wide. Any facility where ten percent or more of supplied electricity goes to AI computation, the storage and processing of data related to that computation, or the equipment and infrastructure supporting it, falls inside the category. A great many operators who do not think of themselves as AI facilities will find that they qualify.
Most consequentially, BC replaced first-come-first-served connection with a competitive process administered by BC Hydro. Capacity is now awarded on provincial criteria: employment, revenue contribution, alignment with economic objectives. The province was explicit in its reasoning — data centres generally provide fewer jobs and less revenue per megawatt than natural resource projects, and BC intends its clean electricity to go where it delivers most.
That is a defensible policy position. It is also a decisive change in what it means to develop compute infrastructure in British Columbia.
What Saskatchewan tells us
The most instructive data point in the region is not a constraint at all. It is what happened when a very large, very well-capitalised counterparty went to build in Saskatchewan.
Bell's announced AI data centre outside Regina is being served by SaskPower with 300 MW of interconnection capacity built in two phases through 2027. But the project also includes SaskEnergy developing natural gas infrastructure — a new high-pressure pipeline and a high-volume meter station — for on-site gas-fired generation to serve peak operational demand and backup.
Read that again. A national telecommunications incumbent, with a supportive provincial government and a Crown utility actively building transmission for it, is still putting gas generation on site.
If the largest and best-connected participants in this market are self-supplying part of their load, the question for everyone else is not whether behind-the-meter generation is legitimate. It is why anyone would expect to succeed without it.
The read-through most developers are getting wrong
The dominant industry response to grid scarcity has been to get better at competing for scarce grid capacity. Larger land positions. More sophisticated queue strategy. Longer option periods. Relationships with system operators and provincial governments.
All of this is rational and some of it will work. Somebody wins the BC competitive allocation. Somebody holds the Alberta capacity that has been assigned.
But it is worth being clear about what that strategy is: it is a bet that you will be selected, on criteria set by a third party, on a timeline you do not control. For a developer with the balance sheet to hold land for a decade, that is a perfectly good bet. For anyone trying to deliver capacity to a customer within the next thirty-six months, it is not a plan.
The alternative is unglamorous. Build smaller. Generate your own power. Site at the fuel rather than at the load. Accept operating burdens that grid-connected developers get to outsource, in exchange for a delivery timeline that belongs to you.
The trade, stated honestly
Behind-the-meter development is not free. You become a generation operator, with everything that implies: fuel supply management, emissions compliance, maintenance cycles, spares inventory, and the permanent obligation to keep machines running that a utility customer never thinks about. You carry fuel price exposure directly. You have no grid to lean on when a unit trips, so redundancy must be engineered rather than assumed. Your permitting path is different, not necessarily easier.
Anyone presenting this as a shortcut is not describing the same business we are in.
What it buys is control. Your power pathway does not depend on a regulator's discretion, a competitive process you have not won, or transmission that has not been built. That is currently the scarcest asset in the industry, and it is not available for purchase at any price.
Where this leads
If the constraint is real and durable — and two provinces writing it into regulation suggests it is — then the shape of what gets built changes. Not one large facility waiting on an interconnection, but many small ones sited where fuel is available and delivered on a manufacturing cadence rather than a construction one.
That is a different asset class with different economics, a different customer, and different risks. Over the coming weeks we will work through each of those in turn: the energy economics that only close at small scale, the segment of the market that nobody currently serves, and why inference workloads are suited to distributed infrastructure in ways training never will be.
The arithmetic is the starting point. Twenty-point-seven against one-point-two. Everything else follows from taking that seriously.
Forward-looking statements. This content contains forward-looking statements within the meaning of applicable securities laws, including statements regarding business strategy, planned infrastructure development and market opportunity. Forward-looking statements involve known and unknown risks and uncertainties, and actual results may differ materially. Readers should not place undue reliance on them. See the Company's filings with the Securities and Exchange Commission for a discussion of risk factors. The Company undertakes no obligation to update forward-looking statements except as required by law.
