Alberta Just Learned What “Social Licence” Actually Costs
Ottawa's five voluntary data centre principles landed the same week Alberta's town halls turned into shouting matches. What Lacombe and Sturgeon County actually prove about sequencing consultation before approval, not after.
AI POLICY & GOVERNANCE
Alberta Just Learned What “Social Licence” Actually Costs
Lacombe, 19 August. A community hall built for a few hundred people is packed past capacity. Nate Glubish, Alberta’s minister of technology and innovation, has spent two years courting the hyperscalers — Meta, the cloud giants, the AI compute buyers — with cheap power and a fast approvals process. Tonight he’s standing in front of the people whose province he sold. Dozens line up at the microphone. Not one of them endorses anything. A man from neighbouring Ponoka gets applause for four words: “Albertans do not want data centres.” A woman tells the minister to “get your boxing gloves on” if another one is coming. When Glubish tries to explain that a data centre “is just a warehouse with computers in it,” the heckling drowns him out. When someone asks for a land acknowledgment, he gives one — and gets booed for that too.
The next night, in Sturgeon County, it gets worse. Glubish is “peppered with profanity” by residents fighting Meta’s proposed $13-billion complex, backed by its own $4.6-billion gas plant. A woman tells him, to loud applause: “Your words are nice, you explain things well, but I don’t trust you.” An activist who has spent eight months opposing a similar project in nearby Olds calls the fight “pure hell.”
This is not a story about Albertans hating computers. It is a story about what happens when a government writes an industrial contract with a company before it writes one with the town that has to live next to the substation. And it lands nine days before Ottawa published a national framework meant to prevent exactly this — Canada’s Responsible Data Centre Development Principles, five rules, voluntary, signed by 23 companies including OpenAI, Google, Meta, Microsoft and Amazon. I want to take the five rules seriously enough to ask what they actually require — and then hold them up against what just happened in Lacombe, because the gap between the two is the whole argument.
The five lines, and what they actually demand
It’s worth being precise here, because “principles” documents are usually written to be agreeable rather than enforceable, and this one is no exception on paper — but each line does carry a real technical demand underneath the soft language.
Create lasting local benefit. This sounds like jobs. It mostly isn’t. A hyperscale data centre is compute-dense and labour-light — hundreds of millions in capital, dozens of permanent roles once construction crews leave. That is structurally different from the oil-and-gas plants Alberta’s Industrial Heartland has hosted for decades, which employ operators, technicians, and trades on an ongoing basis and pay royalties into a system communities already understand. The principle’s actual content, once you strip the word “jobs” out, is procurement, skills training, research partnerships, and — critically — the property tax base a facility of this size throws off to a municipality with a small population and a large new asset on its books. That tax base is real money. It is also the one lever a small town actually controls, which is why fights over assessment class and mill rate are where these disputes usually end up, long after the ribbon-cutting press release is forgotten.
Don’t shift electricity costs to Canadians. This is a rate-design problem dressed as an ethics statement. A gigawatt-scale load dropped onto a regional grid forces new transmission build-out, and the question is who pays for the wires: the developer, through embedded or self-supplied generation, or the ratepayer base, through socialised transmission costs. Saskatchewan’s answer is blunt — developers must supply their own power, full stop. Ottawa’s principle is softer: contribute new supply, storage, or demand flexibility “when required.” Demand flexibility means a facility that can curtail or throttle its own load during grid stress — a real, buildable feature, not a slogan — and it is the difference between a data centre that stabilises a grid at 2am and one that trips it. Alberta hasn’t picked a lane. That ambiguity is precisely what residents in Sturgeon County were shouting about: nobody in that room believed “the polluter-pays principle” Glubish invoked actually pins the bill on Meta rather than on their own power bill five years from now.
Minimize water use and environmental impact. The polite version. The technical version is a fight over cooling architecture. Air-cooled systems use almost no water and burn more electricity. Evaporative and adiabatic cooling use far less power but consume water at scale — millions of litres a day for a large campus. Liquid and immersion cooling, increasingly standard for AI training clusters because GPUs run hotter than general compute, sits somewhere between the two depending on the loop design. “Minimize water use” without a stated technology standard or a disclosed Water Usage Effectiveness figure is not a commitment, it’s a preference. None of the coverage out of Alberta’s town halls suggests anyone has been shown a WUE number for any of these projects. That is the actual gap between the principle and the practice.
Be transparent about local impacts. Here is where Alberta’s failure is most naked, and it isn’t really about the content of the principle at all — it’s about sequencing. Ottawa’s language is specific: engage “early,” disclose information that is “independently verifiable.” Alberta’s residents told Glubish, in as many words, that the consultations came after the approvals and the announcements — Meta’s $13-billion Sturgeon County project and its power plant were already public before anyone asked the county what it thought. A disclosure delivered after the decision isn’t transparency. It’s a debrief. The Alberta Utilities Commission’s own rejection of the Olds gas plant, on setback grounds, after eight months of what Bek MacIntosh called “pure hell” for the community fighting it, is the clearest evidence available that the province’s own regulatory sequencing was backwards from the start.
Bring strategic value to Canada. The catch-all — sovereign compute capacity, data residency, research access — and the one principle nobody in a community hall is going to stand up and argue against on its merits. It’s also the one doing the least local work. Strategic value accrues to Ottawa’s AI strategy. It does not show up on a Sturgeon County tax notice.
Does it matter whether the pipes are already there?
I think this is the more interesting empirical question, and it cuts against the intuitive answer. Sturgeon County sits inside Alberta’s Industrial Heartland — the province’s most industrialised strip, decades of refineries and petrochemical plant, a workforce that has lived alongside heavy industry its whole life. The area’s own industrial association describes it as having enjoyed “very strong community support and social licence” for exactly this kind of development — right up until the AI data centre showed up. If existing infrastructure bought goodwill, this is where it should have shown. It didn’t.
What existing infrastructure actually buys a developer is a shorter engineering timeline, not a shorter political one. Sturgeon County already has gas pipelines and grid capacity nearby — that’s precisely why Meta and its power partner picked it. But the community has also had forty years to learn what an industrial project is supposed to deliver in exchange, and a hyperscale data centre delivers a different bargain: less permanent employment, no royalty structure, and a power and water draw with no precedent in the local ledger. Greenfield sites like Olds carry the opposite trade: no existing grid capacity, which means the developer needs the community’s consent to build the substation and the peaker plant in the first place — genuine leverage, used to real effect when the AUC turned the project down on proximity grounds. Infrastructure doesn’t neutralise the politics. It just decides which side is negotiating from strength.
The toolkit only works before the ribbon is cut
This is where I land, and I want to be direct about it rather than hedge it into mush: a real toolkit handed to a mayor and a council — the standing to demand disclosure, benefit-sharing, water limits, before a project is approved — makes a data centre easier to build, not harder, provided it arrives before the decision, not after.
Alberta is the live demonstration of what happens when the sequence runs backwards. Every quote coming out of Lacombe and Sturgeon County is a trust problem, not a technology problem. Nobody at that microphone was arguing GPUs are dangerous. They were arguing that the government signed the deal and is now performing the consultation. That is the exact failure mode Ottawa’s transparency principle is trying to legislate against with the word “early” — and I’ve watched it play out on the same continent that will one day host its own data-centre gold rush. Building community health information systems across East and West Africa taught me the same lesson twice. In Uganda, the districts that trusted the rollout were the ones brought into the design before the donor funding was locked. In Liberia, where I later worked on national CHMIS deployment, the fault line ran even deeper, because it mapped onto a much older grievance: the same tension between Americo-Liberian settler institutions and indigenous county authorities that has shaped who gets consulted and who gets informed in that country for two centuries. A county health team that felt like the last to know behaved exactly like Lacombe — polite in the room, and quietly uncooperative for the next eighteen months. A community given real leverage before the contract is signed will extract a harder bargain — and then hold up its end of it. A community handed a microphone after the contract is signed will do exactly what Lacombe did: line up, one after another, and refuse to endorse anything at all.
Ottawa’s five principles are not wrong. They are simply describing, in policy language, the order of operations Alberta just proved you cannot skip.
← My Written Thoughts