Bring your own generation: how Alberta decides which data centres get built

When demand for grid connections exceeds available capacity, someone has to decide who connects first. Alberta’s answer is written into the Data Centre Regulation, in force since June 2026: projects that pair their load with dedicated new generation or storage are prioritised in the AESO’s connection process.

The logic is straightforward. A data centre that brings its own supply does not deepen the province’s capacity shortfall. One that does not is asking the existing system, and the people who already pay for it, to absorb an industrial load equivalent to a small city.

Alberta is unusual among Canadian provinces in having the market structure to attempt this at all. It runs a deregulated, energy-only electricity market in which generators are paid for the power they produce rather than for holding capacity available, and in which private developers build generation on their own commercial judgement rather than at a Crown utility’s direction. In most of the country a data centre developer would negotiate with a single provincial utility that owns generation, transmission and distribution. In Alberta the developer can go and build a power plant. The bring-your-own-generation rule is only workable because that option genuinely exists here.

The practical consequences are significant. Bring-your-own-generation changes the skill set a developer needs. Securing land, fibre and a construction contractor is no longer sufficient; a competitive project now needs a generation strategy, an interconnection plan for that generation, and financing that survives the years between commitment and energisation.

It also changes who wins. Firms with power-development experience (independent power producers, midstream and energy companies, and utilities) hold capabilities that pure compute developers have to buy in. In an energy province, that is less a distortion than a return to type, and it is the clearest reason Alberta’s data centre sector will look different from Virginia’s or Ohio’s.

The Meta campus in Sturgeon County illustrates the timelines involved. Reporting on the project describes power arriving late in the decade and a grid position secured under a connection process that has since closed. Whatever one thinks of the project, the sequencing is instructive: the grid position was the scarce asset, and it was locked in early.

Two open questions remain. The first is cost allocation: how much of the new transmission required to serve these projects is recovered from developers rather than the general rate base. The second is what happens to the roughly 18,000 MW of requests sitting behind the first tranche. Some will convert to bring-your-own-generation projects, some will relocate to other jurisdictions, and some were never real.

It helps to understand what "bring your own generation" means in practice, because the phrase covers several quite different arrangements. At one end is a genuinely islanded facility with on-site generation and no reliance on the transmission system, which is rare and expensive. More common is a project that contracts new generation elsewhere in the province and connects both to the grid, so that the net effect on system capacity is close to neutral even though the electrons do not travel directly from one to the other. Between those sits a range of hybrid designs using on-site gas turbines for a portion of load, grid supply for the rest, and storage to manage the difference.

Each version carries a different risk profile. On-site gas is fast to build and gives a developer control, but it exposes the project to fuel price and to emissions policy that may tighten over the asset’s life. Contracted renewables are cheaper per megawatt-hour and carry no fuel risk, but they are intermittent, and a data centre is the least intermittency-tolerant load there is. That mismatch is why storage keeps appearing in these proposals: it is the component that makes a variable supply look like a firm one to a customer that cannot tolerate a flicker.

The regulation also creates a timing problem that is easy to miss. Generation and transmission are permitted through different processes on different clocks, and a data centre needs both to line up. A developer can hold an approved generation project and a signed interconnection agreement and still be unable to energise because the transmission upgrade serving the site is scheduled two years later. Sequencing risk of this kind does not appear in any announced capital figure, and it is one of the more common reasons large industrial projects slip.

For municipalities the regulation has a second-order effect worth noting. Because projects that arrive with generation are advantaged, the sites that become attractive are the ones where generation can also be built: land near gas supply, near existing transmission corridors, or with wind and solar resource. That is a different map from the one you would draw if you were optimising for fibre routes and proximity to population. It is part of why so much of this activity is landing in places like Sturgeon County and the Grande Prairie region rather than in the two large cities.

For readers tracking this sector, the distinction that matters is between a project with a signed interconnection agreement and a project with a press release. Only one of those is a commitment, and only one of them has survived contact with a regulator whose job is to say no.