The pitch is that Alberta’s deregulated market is the reason data centres come here, while Quebec makes them queue for a minister’s signature. For that to be true, a developer arriving in Alberta with a 400 megawatt load would have to be able to buy a connection rather than apply for one. That is not what happened. The Alberta Electric System Operator capped interim large load connections at 1,200 megawatts in June 2025, handed the entire allocation to two projects, and sent everybody else to a second phase that still has no completion date.
Key facts
- The International Energy Agency report Key Questions on Energy and AI, published 16 April 2026, puts global data centre electricity consumption at 485 terawatt hours in 2025 and projects roughly 950 terawatt hours by 2030, about 3% of global electricity demand.
- The Alberta Electric System Operator announced an interim cap of 1,200 megawatts of additional large load through 2028 on 4 June 2025, applying to projects of 75 megawatts or larger that require no new transmission reinforcement.
- That 1,200 megawatts was fully allocated to two projects, a 970 megawatt load and a 230 megawatt data centre phase, against transmission service requests exceeding 16 gigawatts.
- Alberta’s record electricity demand peak is 12,785 megawatts, set on 11 December 2025, per the Alberta Electric System Operator.
- Quebec requires authorization from its Minister of Economy, Innovation and Energy for new loads of 5 megawatts or more, a requirement reinforced by Bill 69, adopted in June 2025.
Power is the constraint, and the numbers are not close
Start with the demand side, because it is the part nobody disputes. Global data centre electricity use ran at 485 terawatt hours in 2025 and the International Energy Agency’s April 2026 update projects roughly 950 terawatt hours by 2030. Whatever you think of the business case for artificial intelligence, the electricity is not speculative. It is contracted, scheduled and in many jurisdictions already unavailable, with grid connection waits running five to ten years.
Texas is the clearest illustration because it is the closest analogue to Alberta’s market design. The large load connection queue in the Electric Reliability Council of Texas grew from roughly 63 gigawatts in December 2024 to over 230 gigawatts by January 2026. The state’s all-time peak demand is about 85 gigawatts. A queue nearly three times the size of the entire system is not a market signal. It is a waiting list with a spreadsheet attached.
Alberta’s version of the same arithmetic: more than 16 gigawatts of transmission service requests against a record peak of 12,785 megawatts. By our own arithmetic that is roughly 1.25 times everything this province has ever drawn at once, and the 1,200 megawatts actually released is about 9.4% of that peak. The requests are not all real, and anybody in this field knows that a connection request costs a developer far less than a substation. But even discounting heavily, the gap between what has been asked for and what has been granted is the whole story.
The strongest version of the Alberta case, and where it breaks
Here is the optimistic argument at its best, and it is better than its critics allow. Alberta has an energy-only market where a generator can build to serve a load without waiting for a rate base decision, abundant natural gas, a cold climate that cuts cooling load for a large part of the year, and a government that has been willing to legislate decision deadlines rather than let regulators drift. Bring your own generation is a genuine structural advantage, and the Sturgeon County project pairing a $13 billion campus with on-site generation is the proof that the model works at scale. None of that is marketing. It is materially easier to power a large load here than in most of North America.
Where it breaks is the claim that Alberta does not ration. Alberta rations. It rationed by administrative cap and pro-rata assignment above a 75 megawatt threshold, which is a different mechanism from Quebec’s comparative evaluation above 5 megawatts, and a much higher threshold, but it is rationing all the same. The honest version of the provincial pitch is that Alberta rations at a higher ceiling, on published technical criteria, and lets a developer who brings generation skip the line entirely. That is a real advantage. It is not the same as a market, and the difference will matter to the twentieth applicant in a way it did not matter to the first two.
What a host community should be asking before the announcement
The questions that get asked at a town hall after a project is announced are the wrong ones, because by then the connection is already allocated. The useful questions are earlier and duller. How much of the load is served by grid supply and how much by on-site generation, and in which years, because a campus energizing before its dedicated plant is finished runs on everybody else’s electricity in the interim. What happens to the connection if the project stalls. Whether the local distribution work is paid by the proponent or socialized. And what the operator’s own guide, published 12 June 2026, actually requires of a transmission-connected data centre, because that document sets the technical obligations long before any council votes on anything.
Communities near Indus and elsewhere have been handed the debate at the point where only the visible parts remain negotiable. The decisions that determine whether a project is good for a county were made in a connection queue nobody outside the industry reads. That is not a conspiracy. It is a sequencing problem, and sequencing problems are fixable if somebody writes the process down in plain language before the next wave of AI infrastructure arrives.
Here is the prediction, so it can be checked. If the second phase of the large load framework is not published with firm allocation rules before the end of 2026, at least one announced Alberta project of 200 megawatts or more slips its energization date into 2030 or later, and the reason given will be transmission rather than capital.
