Alberta’s New Data Centre

Meta Is Building the Biggest Data Centre in Canada North of Edmonton. I'm Not Sure How I Feel About It.

On July 8th, Danielle Smith stood in Calgary next to a Meta executive and announced that Facebook’s parent company is putting more than $13 billion into a data centre in Sturgeon County. It’s Meta’s first in Canada. It’s their largest anywhere outside the United States. And it’s one of the biggest private investments this province has ever seen.

Smith called it a big deal. Her tech minister, Nate Glubish, said it “did not happen by accident. It happened by design.”

He’s right about that part. Alberta spent two years chasing this.

Here’s where I land before I get into any of it. I’m not against data centres. I’m not going to pretend the AI buildout is some conspiracy against Alberta. But I’m not ready to call this a win either. There’s a lot of noise on both sides right now and not much of it survives contact with the actual numbers.

So let’s go through what we know, what we don’t, and who’s guessing.

What Alberta Says It’s Getting

The facility goes in Sturgeon County, about 35 kilometres north of Edmonton, in Alberta’s Industrial Heartland. That land has been zoned for heavy industry for decades. It’s about 7 square kilometres of land and roughly 270,000 square metres of building.

The province’s numbers:

• More than 3,000 construction jobs at peak

• 300 permanent jobs once it’s running

• Roughly $250 million a year to Alberta in royalties, taxes, levies and fees

• $60 million from Meta into local roads and water infrastructure

• Meta paying about $100 million a year in transmission fees

That last one matters. Because Meta is paying into the transmission system, the fixed cost of the grid gets spread across a bigger base. The province says Alberta ratepayers could see up to a six per cent drop in the transmission portion of their power bill.

Not the whole bill. The transmission portion. That’s an important distinction and it’s the one everyone is skipping.

How the Power Actually Works

You need to understand this part before you can judge the rest.

Alberta passed two bills in December 2025. Bill 8 requires large data centres to bring their own generation and pay for their own transmission upgrades instead of dumping the cost on ratepayers. Bill 12 puts a levy on the value of computing equipment for any data centre pulling 75 megawatts or more. Two per cent if you’re grid-connected. One per cent if you generate your own power but stay hooked to the grid. Nothing if you’re fully off-grid.

The province calls this “bring your own power.” It’s the whole pitch.

So Meta is doing both. About 970 megawatts grid-connected, including a ten-year deal with Capital Power for 250 megawatts. And a new gas plant called the Greenlight Electricity Centre, built by Pembina Pipeline, Morgan Stanley Infrastructure Partners and Kineticor. That plant costs $4.6 billion, starts at 932 megawatts, and is designed to scale toward 1,800.

Meta doesn’t own the plant. They’re buying the power under a long-term contract.

Meta calls the data centre a one gigawatt facility. The entire city of Edmonton draws about 1.4 gigawatts.

Now here’s the thing nobody says out loud, and I want to be careful with it because it cuts both ways.

Bring your own power is a rule about who pays. It is not a rule about what gets burned. The emissions don’t disappear. They move from the grid to the property line. So don’t let anyone sell you this policy as an environmental win. It isn’t one. It’s a cost-allocation rule and it should be judged as a cost-allocation rule.

But don’t let anyone sell you the reverse either, because the Pembina Institute is already trying. Their statement called it a “missed opportunity” to include wind and solar in the mix. Think about what’s actually being asked for there. A one gigawatt data centre runs at full tilt twenty-four hours a day, three hundred sixty-five days a year. Wind and solar can’t serve a load like that. Not without enormous overbuild and a scale of battery storage that doesn’t exist yet at any price. That’s not an ideological objection. That’s arithmetic.

So let’s do the arithmetic, quickly.

Travers Solar in Vulcan County is the biggest solar farm in Canada. 465 megawatts on about 13.5 square kilometres, and over its first three years it ran at a 25 per cent capacity factor. That’s the best number in the country. The Alberta average is closer to 20. I’m using the generous one on purpose.

To make the same annual energy as this one data centre, you’d need 8.6 Travers Solars. Call it 116 square kilometres, sixteen times the footprint of the entire Meta campus. The gas plant doing the same job sits on about a fifth of a square kilometre. Wind is better on capacity factor but worse on spacing, and to be fair about it, most of the land between turbines stays in crops.

Then comes the part that actually settles it, and it has nothing to do with land.

A data centre doesn’t need annual energy. It needs power at three in the morning on January 12th. At noon on this past winter solstice, Alberta’s entire solar fleet was running at 33 per cent, at the moment the sun was highest in the sky. By evening, zero, same as every night. And on January 5th, 2022, during an extreme cold warning across most of the province, wind and solar together produced one tenth of one per cent of their rated capacity. Turbines shut down around minus 30 so the blades don’t shatter. That’s the same weather that has everybody cranking the furnace.

Carrying one gigawatt through a single winter night takes roughly 16 gigawatt hours of storage. Alberta’s entire grid battery fleet is about 271 megawatts, most of it short duration. That’s one night. The 2022 event lasted days.

So when the Pembina Institute calls this a missed opportunity for wind and solar, ask them where the 116 square kilometres goes, and ask them what powers the building in January.

That’s not a defence of Meta. It’s just what the numbers say.

And gas is not the villain in this story. Greenlight is a combined-cycle plant, which runs around sixty per cent thermal efficiency. Coal runs thirty-three to forty. So gas burns less fuel to make the same electricity, puts out roughly half the CO2 per megawatt hour, and emits essentially no sulphur dioxide, no mercury, and almost no particulate. Every one of those things kills people. Carbon doesn’t kill anybody directly. If you’re going to burn a hydrocarbon to run a data centre, gas is the good answer, not the compromise.

The green alternative isn’t free either. Solar panels, turbine blades and grid batteries take mining, smelting, concrete and steel. That cost is real, it just gets paid somewhere you can’t see it from Sturgeon County.

So the honest read is this. Gas is the right fuel for this job today. The policy doesn’t reduce emissions, it relocates them. Both of those are true at the same time, and anybody telling you only one of them is selling something.

Where the Numbers Get Slippery

You’re going to see a lot of figures thrown around this week. Be careful with them.

The Pembina Institute, a clean energy think tank in Calgary that has nothing to do with Pembina Pipeline, called this an 1,800 megawatt data centre that will use more power than the City of Calgary. Reuters compared it to 800,000 homes. Meta and the province say one gigawatt and compare it to Edmonton.

Those aren’t the same claim. The 1,800 megawatt figure is the maximum potential buildout of the gas plant, not the announced size of the data centre. Depending on which number someone picks, this project either uses less power than Edmonton or more than Calgary.

I’m going with the conservative version because that’s what was actually announced. But the bigger number isn’t made up. It’s measuring something different.

What Data Centres Actually Do to the Environment

This is where I think both sides are wrong, so let’s be careful.

Globally, data centres used about 460 terawatt hours of electricity in 2025. That’s roughly 1.8 per cent of world electricity demand and about two per cent of global emissions. For scale, that’s about what France uses in a year. Transportation, agriculture, heating and heavy industry each dwarf them.

So no, data centres are not currently a top-tier driver of climate change. If somebody tells you AI is burning the planet, they’re ahead of the evidence.

The trajectory is the real story. The IEA projects data centre consumption could pass 800 terawatt hours by 2028. Goldman Sachs has floated a 165 per cent jump in demand by 2030. That’s the curve worth watching, and it’s steep.

But here’s the part that actually settles the question, and it’s the reason I can’t give you a clean verdict on this project:

“How bad is a data centre” is the wrong question. The environmental cost is almost entirely the energy mix and the local water situation. It’s not the servers. The same building running on Quebec hydro in a wet climate is close to negligible. The same building running on gas and pulling municipal water in a dry one is genuinely damaging.

Alberta is a strange case here. Cold climate is a real advantage, and it’s not a talking point. When outside air is near freezing you can cool servers with air instead of running chillers and evaporating water. That’s a genuine win on electricity and on water.

But the energy mix is the bigger lever, and Alberta’s grid runs on gas. Our electricity is far more emissions-intensive than the national average. Ottawa’s national AI strategy assumed Canadian data centres would run on a clean, hydro-heavy grid. Alberta is the exception, and the bulk of Canada’s planned data centres are being built here anyway.

So Alberta gets one of the two wins. The cold, not the clean power.

The Water Question, and the Part Meta Isn’t Answering

Meta says the facility uses closed-loop liquid cooling with dry cooling and needs no operational water for cooling at all. On-site water is limited to drinking, fire suppression and equipment maintenance. Their VP said the site will use less water in a year than a regional golf course. Meta also says it’ll be water positive by 2030 and is conserving about 80 hectares in the North Saskatchewan River watershed with ALUS and local farmers.

I believe the technology claim. Closed-loop dry cooling is real and it’s not the same thing as the evaporative systems that made data centres a water story in Arizona and Texas.

Two things worth knowing anyway.

First, “zero water cooling” is a trade, not a miracle. Water cooling is far more energy efficient than air cooling. When you eliminate the water, you spend more electricity. The cost doesn’t vanish. It moves.

Second, and this is the one nobody covering this announcement has mentioned: the power plant uses water. Thermal generation, gas included, boils water for steam and uses more of it for cooling. A data centre can honestly report almost no on-site water use while still driving real water consumption upstream at the generating station.

Meta’s number counts the building. It doesn’t count the 932 megawatt gas plant being built to feed it.

I’m not saying that number is huge. I’m saying nobody has published it, and nobody has asked. That’s a fair question for a reporter or an MLA to put to Pembina Pipeline before that plant fires up in 2030.

There’s also a broader verification problem here. Companies in this industry routinely operate under non-disclosure agreements and don’t publish audited water, energy or emissions data. That cuts both ways. The scariest activist figures are hard to verify. So are the corporate “we use zero water” claims. Almost everything you’re reading this week, from both camps, is an estimate.

The Concerns Are Real

The grid. In June 2025, AESO had 29 large-load connection requests representing over 16,000 megawatts. That queue is now 41 projects and 19.5 gigawatts, up from five gigawatts in early 2024. Alberta’s entire peak demand is about 12,000 megawatts. We’re not building one data centre. We’re being asked to build a demand base bigger than the province currently uses. And a lot of Albertans remember the grid alerts during the January 2024 cold snap.

Noise, light, land. Land-use ecologist Brad Stelfox filed a report with the Alberta Utilities Commission warning about noise, light and air pollution near these facilities. Sturgeon County residents are raising the same concerns. The county says the site sits in a pre-zoned industrial area next to existing heavy industry, not near neighbourhoods. Forty-four people left angry reactions on the county’s own announcement post.

The jobs number. 300 permanent jobs for $13 billion. That’s the number that makes people squint.

Data centre employment accounts for something like one hundredth of one per cent of all jobs in the United States, while the industry eats around four per cent of the country’s electricity. Construction jobs are real and they’re good jobs. But construction jobs end.

There’s a second problem with the jobs pitch that doesn’t get said out loud, and it’s the one that should matter most to anybody who works with their hands in this province.

A lot of these jobs don’t go to locals.

Good Jobs First went through data centre subsidy agreements looking for local hiring requirements and couldn’t find one that had them. A development agency in Genesee County, New York expects sixty per cent of the construction workforce on a proposed data centre to come from outside the fourteen-county region. Amazon promised Indiana a little over a thousand jobs, but the actual subsidy agreement commits to four hundred Amazon employees. The rest are subcontractors. The New York Times reported that data centre electrical work is largely done by crews who travel across the country from project to project, because that’s who has the specialized experience. Electrical is anywhere from forty-five to seventy per cent of the budget on these builds.

Meta already knows how this looks. Its Hyperion campus in Richland Parish, Louisiana peaked at around five thousand workers in a parish with about twenty thousand people. Around the Stargate site in Abilene, Texas, thousands of out-of-state trades flooded in and rents went up.

Anybody who’s worked a big project knows this pattern. The crew shows up, the camp fills, the rents jump, the work finishes, everybody leaves.

Now, the honest counterweight, because I’m not going to hand you only the numbers that agree with me.

A Brookings study published this past May actually found the opposite of what most critics claim. Counties that land their first big data centre see total private employment rise four to five per cent over five or six years. Construction employment climbs eleven per cent. Wages go up three to four per cent for existing workers and new hires both, with no jump in house prices. Those are real gains for real people.

Two catches. First, the researchers found that the industry’s own sponsored studies overstate the employment effect by a factor of three. Second, the effects that make data centres worth having are concentrated specifically in hyperscale projects and in clusters, not single facilities. A one-off data centre generates modest gains. A cluster generates the real thing.

Which, if you take it seriously, is an argument for Glubish’s dominoes. It’s the same fact pointing in both directions depending on what you’re worried about.

And here’s where Alberta might genuinely be different. The general contractor is Frost Collective, a joint venture between Clark Builders and PCL Construction. PCL is headquartered in Edmonton. This province has an industrial trades workforce sitting an hour down the road in the Heartland, which is not something you can say about Richland Parish.

But nobody has published a local hiring commitment on this project. Not the province, not Meta, not Sturgeon County. Three thousand construction jobs is a number in a press release until somebody puts a floor under how many of them go to Albertans.

That’s a question your MLA can ask and get an answer to. It’s a better question than most of the ones being asked this week.

The Carney Angle

Here’s what makes this bigger than one project in Sturgeon County.

On July 2nd, Mark Carney stood in Vancouver with David Eby and said Ottawa would fast-track LNG projects that will “more than triple Canada’s LNG production over the next decade.” That same week, the Greenlight gas plant reached its final investment decision. Six days later, Meta was announced.

That is not a coincidence of scheduling. Look back further. The November 2025 memorandum of understanding between Carney and Smith committed both governments to supporting data centre development, promised a federal framework by July 1, 2026 to incentivize large data centres and Canadian sovereign computing, and put the federal Clean Electricity Regulations in abeyance. Pembina Pipeline said so directly in its own press release announcing the plant.

Alberta didn’t land Meta alone. Ottawa cleared the runway.

Now run the arithmetic that David Pickup at the Pembina Institute is running. You’re adding a 932 megawatt gas plant, and eventually maybe 1,800. You’re adding 41 more data centre applications behind it. And you’re simultaneously tripling the LNG exports that ship Canadian gas to buyers in Asia and Europe.

All of that is competing for the same natural gas. The same gas that heats your house and powers the grid you plug into.

Pickup’s conclusion is that Albertans will see rising energy costs in the years ahead. His words: Alberta’s energy policy is “being designed to structurally lock in demand for natural gas above all other options, even if it means higher and more volatile costs for consumers.”

I don’t have to agree with everything the Pembina Institute says to notice that’s an argument nobody in government has answered. Smith’s response is the six per cent transmission discount. That’s a claim about one line item on your bill. It says nothing about the commodity price of the gas that generates the power in the first place.

And it’s worth remembering where we are on the calendar. Alberta votes in October on whether to hold a referendum on leaving Canada. Carney has spent the summer signing agreements with Smith. A pipeline. LNG. A data centre framework. Some of that is good policy on its own merits. All of it is also politics.

What the Evidence Says About Your Power Bill

Everybody is telling you data centres will wreck your electricity bill. It might be true. The evidence is worse than either side admits.

Senator Elizabeth Warren said people near large data centres have seen bills rise 267 per cent over five years. PolitiFact rated that Mostly False. The figure was wholesale power prices, not residential bills.

A Rutgers researcher looked at data centre openings across 24 states from 2014 to 2024 and found the effect on residential bills was “weak, mixed, and small in magnitude.” A separate study found no evidence of a cost shift from data centres to residential customers in Virginia, the largest data centre market on earth. That study was funded by the Data Center Coalition, which is the industry’s lobby group. Take it accordingly.

On the other side, Harvard Law’s Ari Peskoe lays out the mechanism clearly. Utilities build infrastructure for the data centre and spread the cost across every ratepayer. Wholesale prices rise when demand outpaces supply. Both are real. Both take years to show up.

American residential power prices are up about 42 per cent in five years. Data centres are part of that. So are aging grids, equipment costs, and everything else.

Nobody has clean numbers yet. The people telling you data centres will definitely wreck your power bill are ahead of the evidence. The people telling you they definitely won’t are usually being paid by someone.

Where I Land

Alberta’s model is genuinely different and I’ll give credit where it’s due. Requiring these companies to build their own generation and pay their own transmission costs is a better structure than what Virginia did. Rocky View County rejected a data centre last fall. The AUC threw out the Synapse application in Olds in March. Hamilton paused development. Manitoba’s premier killed a project outright. Alberta didn’t just open the door and hope.

But better structure is not the same as good deal.

The levy gets credited back against corporate income tax, so the $250 million a year needs a harder look than a press release gives it. The gas demand argument hasn’t been answered. The power plant’s water use hasn’t been published. And if that 19.5 gigawatt queue turns into anything close to reality, bring your own power gets tested in a way one project never will.

Glubish said this will be “the first domino of many to fall.” He meant it as a good thing. That’s also exactly what worries me.

I want Alberta to build things. I want investment here instead of Texas. I’d rather we be the province that figured out how to do this properly than the province that watched it all go somewhere else and complained about it.

But I’ve watched this place get told a project was a sure thing before. I’d rather be the guy asking questions in year one than the guy explaining in year five why nobody did.

The jury’s out. It should be.

* * *

No Signal Out Here covers Alberta politics, independence, and the issues Ottawa doesn’t want you thinking about. Subscribe for more.