A Hundred Data Centres Is an Infrastructure Portfolio
Texas stopped issuing new state data-centre permits this week. Alberta now has more than 100 projects proposed.
Capital Power says the Meta project has raised Alberta’s credibility with other hyperscalers, that it is talking to several more, and that the province could become a multi-gigawatt market. On its own, that sounds like an attraction problem being solved, and it mostly is. But success changes the problem you are solving.
Texas has paused new state-issued permits while regulators audit what the industry is doing to the grid, with more than 474 GW of connection requests, over five times the state’s record peak demand. California chose rules rather than a pause, tightening requirements around electricity costs, water use and local oversight so the cost of new generation and grid upgrades does not simply fall on other ratepayers. Europe wants data-centre capacity to more than double by 2030, while proposing more reporting on energy use, water use, local water stress and waste-heat recovery.
None of that means attracting the investment was a mistake. Once enough projects arrive, the problem stops being project-by-project.
One data centre is a project you can assess on its own. A hundred proposed data centres behave more like an infrastructure portfolio, because they do not arrive independently. They compete for the same transmission, the same gas, the same turbines and transformers, the same construction crews, the same water, and eventually the same public patience.
Alberta has already started responding to that problem. AESO capped the first phase of large-load grid connections at 1,200 MW, now fully allocated, while data-centre connection requests total roughly 19,565 MW. That helps sequence access to the grid. The harder question is whether everything around the grid is being sequenced on the same clock.
The dedicated 932 MW Greenlight plant serving Meta is expected online in 2030, while Capital Power’s 250 MW agreement starts in 2028. Generation has a lead time. So does transmission. So do gas connections, turbines, transformers, water infrastructure and construction capacity. The slowest constraint is the one that moves the date. Multiply transitional supply arrangements and equipment orders across even a fraction of the proposed projects and they start colliding with each other.
Alberta does not have to learn this the expensive way. Texas, California and other mature data-centre markets are already showing what gets difficult once the money starts arriving, so Alberta can watch which constraints become binding first, and which questions communities start asking after the concrete is poured. Being a few years behind might be the advantage here.
So the number I am watching is not how many projects have been proposed. It is how many can turn into operating capacity before the hundredth one is waiting on infrastructure the first ninety-nine already used up.
Which constraint do you think Alberta hits first?
Comments