The AI Data Centre Rush Is Outrunning Public Consent

AI data centres are being expanded faster than public policy, utility planning, water planning, zoning, and community consent processes can responsibly absorb.

That is the issue.

Not AI itself.

Not technology.

Not innovation.

The issue is that a new class of heavy industrial infrastructure is being pushed into municipal lands, rural regions, industrial corridors, and Indigenous territories at a speed that does not match the scale of its demands.

And the demands are massive.

Meta has announced a new 1-gigawatt AI-optimized data centre in Sturgeon County, Alberta. It is Meta’s first data centre in Canada, its largest outside the United States, and represents more than CAD $13 billion in investment. Meta says the project will support about 3,000 construction workers at peak and more than 300 operational jobs once completed.

That is the headline.

Investment. Jobs. Infrastructure. Opportunity.

But behind the headline is the harder question: What exactly are we approving?

A 1-gigawatt data centre is not just a building. It is an industrial-scale power demand. It is land consumption. It is transmission planning. It is cooling infrastructure. It is backup generation. It is water and wastewater servicing. It is road upgrades. It is noise. It is heat. It is air emissions tied to power supply. It is public risk attached to private growth.

In Sturgeon County, the project is being located in Alberta’s Industrial Heartland, an area the County says was planned for heavy industry and is already zoned for large-scale industrial activity. The County also says data centre development is directed away from residential neighbourhoods and that projects must show how they will manage impacts such as water use, transportation, and construction activity.

That matters.

This is not a data centre being dropped into the middle of a residential subdivision.

But that does not end the public-interest question.

Industrial zoning answers one question: where can it go?

It does not fully answer:

How much power will it require?
Who pays for the grid and generation consequences?
How much water and wastewater capacity is being reserved?
What happens during drought or peak demand?
What emissions come from dedicated power generation?
What does continuous industrial noise do to nearby people, wildlife, and land users?
What are the cumulative impacts if more projects follow?
What public process is enough for infrastructure of this scale?

Sturgeon County’s own public material says Alberta’s deregulated electricity market encourages data centre projects to generate their own power on-site and that large industrial projects bringing their own generation reduce pressure on the shared system. It also says the County will operate separate potable and process water systems to protect drinking water and prevent high-volume users from straining the potable system.

That is the official reassurance.

But it also exposes the core issue.

If the project requires its own power strategy, its own water-service planning, its own wastewater planning, its own road upgrades, and its own regulatory path, then we are not talking about a normal commercial development.

We are talking about industrial infrastructure at a scale that should trigger a much deeper public test.

Meta says the Sturgeon facility will use a closed-loop, liquid-cooled system with dry cooling, designed to avoid the need for continuous water supply for cooling. The company also says it will invest about CAD $60 million in local infrastructure improvements and will pay the full cost of energy, water, and wastewater services for the data centre.

Good.

Those commitments should be written down, audited, enforced, and publicly reported.

Because corporate commitments are not the same thing as public accountability.

The question is not whether Meta can publish a responsible-sounding fact sheet.

The question is whether residents, municipalities, Nations, farmers, land users, and future councils will have transparent, enforceable evidence that the facility is not transferring hidden costs onto the public.

This is the pattern we need to name.

For years, ordinary people have been told to conserve.

Use less water.
Reduce emissions.
Pay more for energy.
Change consumption habits.
Accept restrictions.
Adapt to climate policy.
Use paper straws.
Shrink your footprint.

Then, almost overnight, governments welcome industrial-scale AI infrastructure that can require power at the scale of cities.

That contradiction is not lost on people.

The public is not anti-technology. The public is tired of being asked to sacrifice at the household level while massive industrial users are fast-tracked under the language of innovation, competitiveness, and economic growth.

Alberta’s own AI Data Centre Strategy says the province is committed to reducing red tape and positioning Alberta as a premier destination for AI data centres. It also describes a concierge service to work directly with potential project proponents and help facilitate industry connections.

That is not neutral language.

That is recruitment language.

Again, name it clearly: Alberta is not simply regulating this industry. Alberta is courting it.

That means the public-interest test has to be stronger, not weaker.

A government that is actively recruiting an industry cannot rely on ordinary approval language and expect people to trust that all impacts have been fully weighed.

The electricity issue alone is enough to demand scrutiny.

Alberta filed its Data Centre Regulation on June 9, 2026. Legal analysis of the regulation states that it was created because large data centres can create significant new electricity demand on timelines faster than traditional generation and transmission planning cycles.

That sentence should stop every municipal council in its tracks.

Faster than traditional planning cycles.

That is the problem.

AI companies are moving at market speed.

Utilities move at infrastructure speed.

Water systems move at watershed speed.

Municipal planning moves at public-process speed. Which if you don’t know, is slow. Very slow.

Communities move at trust speed. And right now, trust is a rare commodity.

When one of those systems outruns the others, the public carries the risk.

The Alberta Electric System Operator has also stated that data centres present novel technical and operating challenges and require information different from a typical load to assess grid access.

So let’s drop the soft language.

These are not ordinary loads.

These are not ordinary warehouses.

These are not ordinary employers.

They are large-load industrial compute campuses with power, cooling, water, land, noise, emission, transmission, and public-cost implications.

The Sturgeon County case is important because it shows both sides of the issue.

On one side, it is located in a heavy industrial area. It brings investment. It brings construction jobs. It brings a non-residential tax base. It may be designed with closed-loop cooling. It may pay for major local infrastructure. It may be better sited than many other proposed data centres.

On the other side, it is a 1-gigawatt AI facility tied to dedicated natural gas-fired power. AP reports the facility will be powered by a natural gas-fired plant developed by a consortium including Pembina Pipeline, and identifies the Greenlight Electricity Center as a 932-megawatt plant expected to begin operating in the second half of 2030.

So do not call this clean just because it is digital.

Do not call it weightless because the product is data.

Do not call it sustainable because the building is modern.

AI has a body.

That body is land, steel, chips, water, electricity, cooling, generators, transmission, substations, emissions, minerals, construction, waste, and noise.

The cloud is not in the sky.

It is on someone’s land.

That is why this conversation matters for municipalities.

And it matters for Indigenous Nations.

Across North America, data centre developers are looking at rural lands, industrial lands, agricultural lands, and sovereign lands because those places can offer space, power access, fewer adjacent residents, faster approvals, and sometimes more flexible governance pathways.

For Nations, the issue is even more serious.

Economic development matters. Jobs matter. Revenue matters. Sovereignty includes the right to say yes.

But sovereignty also includes the right to say no with full information.

The Seminole Nation of Oklahoma passed a moratorium on generative AI technology and hyperscale data centre development in March 2026 by a 24–0 Tribal Council vote. Reporting says the moratorium followed concern about data centre development inquiries and the speed at which communities were being asked to respond.

That is not fear of progress.

That is governance.

That is a Nation saying: slow down, disclose, assess, and do not rush us into an infrastructure decision that could affect land, water, energy, culture, and future generations.

Every Nation and every municipality should have the same posture.

Not automatic rejection.

Not blind acceptance.

A hard public test.

Before any AI data centre is approved, the proponent should have to answer, in public, in plain language:

How much electricity will the facility require at full buildout?
What new generation is tied to it?
What emissions will that generation create?
Will residents or businesses absorb any grid cost, rate pressure, or reliability risk?
How much water will be used directly?
How much water is used indirectly through power generation?
What happens during drought, river stress, or peak summer demand?
What wastewater will be produced?
What chemicals are used in cooling or maintenance systems?
How loud will the facility be at the property line, including low-frequency hum?
How far will noise travel at night?
What are the effects on wildlife, livestock, nearby residents, and land users?
How many permanent jobs will actually remain after construction?
What local skills will be built?
What happens if the AI boom slows and the facility becomes stranded infrastructure?
Who pays for decommissioning?
What benefits are guaranteed, not merely promised?
What cumulative impacts occur if three, five, or ten more projects follow?

That is not anti-AI.

That is basic stewardship.

The industry wants to talk about the future.

Fine.

Then let’s talk about the future honestly.

A future where AI improves medicine but drains watersheds is not responsible.

A future where AI increases productivity but raises local power burdens is not responsible.

A future where AI companies receive fast-track public support while communities receive after-the-fact consultation is not responsible.

A future where Nations and municipalities are asked to trade land, water, silence, and energy for speculative promises is not responsible.

This is the line:

If AI infrastructure cannot prove that it strengthens the place that hosts it, then it is extraction.

Digital extraction is still extraction.

The Sturgeon County project may become a model of better industrial siting and stronger infrastructure planning.

Or it may become the first major Alberta example of the AI gold rush being normalized before the public fully understands the cost.

That depends on what happens next.

The approval process cannot be treated as a communications exercise.

It must be treated as a public covenant.

If the project is as responsible as promised, then the reporting should be public.

Power use should be public.
Water use should be public.
Wastewater data should be public.
Noise monitoring should be public.
Generator testing should be public.
Emission impacts should be public.
Infrastructure cost agreements should be public.
Emergency planning should be public.
Community benefit commitments should be public.

Not buried.

Not summarized.

Not hidden behind commercial confidentiality whenever the numbers become inconvenient.

Public land-use decisions require public evidence.

That is the standard.

And this standard should apply everywhere: municipal lands, rural counties, industrial parks, reserve lands, settlement lands, Métis lands, agricultural regions, and edge-of-city growth corridors.

The issue is not whether AI has benefits.

It does.

The issue is whether the infrastructure required to power AI is being governed with the seriousness it deserves.

Right now, the answer is no.

AI data centres are being expanded faster than public policy, utility planning, water planning, zoning, and community consent processes can responsibly absorb.

That is the thesis.

And Sturgeon County is now the case study.

The public should not be asked to accept a new industrial era on faith.

We need proof.

Proof that water is protected.
Proof that power costs are not shifted.
Proof that emissions are counted honestly.
Proof that noise is controlled.
Proof that communities benefit beyond construction jobs.
Proof that Nations and municipalities have real consent, not managed consultation.
Proof that this infrastructure serves the place it occupies, not only the companies that own it.

Progress is not proven by what we can build.

Progress is proven by whether what we build leaves people, land, water, and future generations stronger than before.

That is the test AI data centres must meet.

Until they do, the rush should slow down.