The true cost of AI data centres

Summary

Podcast Summary: Front Burner

Episode: "The True Cost of AI Data Centres"

Host: Jayme Poisson (CBC)
Guest: Ellen Thomas (Business Insider investigative reporter)
Date: March 24, 2026

---

Overview

This episode of Front Burner explores the rapidly expanding footprint and complex impacts of AI data centres in Canada and the US, with a particular focus on a massive new development outside Regina, Saskatchewan. Host Jayme Poisson interviews Ellen Thomas, who has extensively reported on the AI data centre boom, to examine the promises, environmental and economic costs, and growing community pushback associated with these facilities. The discussion broadly questions what is gained and lost in the race to power the world’s artificial intelligence—and at what environmental and social cost.

---

Key Discussion Points & Insights

1. The Scale of AI Data Centres ([02:37]–[03:30])

• The upcoming Saskatchewan facility:
  • Set on 150–160 acres (roughly 120 football fields)
  • 300 MW capacity—enough to power 200,000–300,000 homes, or all of Detroit.
• US data centres can be three or four times larger, but 300 MW is still “very big.”

"300 megawatts is a very big data center. Roughly speaking, 300 megawatts is enough electricity to power 2 to 300,000 homes." — Ellen Thomas [03:02]

2. Economic Benefits & Job Creation ([03:30]–[05:01])

• Initial jobs boom during construction (approx. 800 temporary jobs).
• Only about 80 long-term, higher-paid positions post-construction—mostly technicians and security.
• US experience: job growth is transient as workers move from project to project.

3. Environmental Costs: Water and Power ([05:01]–[08:59])

• Data centres require enormous amounts of electricity (running 24/7) and water for cooling.
• In the US, 40% of data centres operate in water-stressed regions, consuming “millions of gallons” per day.
• Community resources are stretched:
  • Who pays for electricity and grid improvements?
  • Water supply concerns, especially critical in already-stressed regions.

"Data centers are guzzling millions of gallons of fresh water a day. 40% of US data centers are in water-stressed areas of the country." — Ellen Thomas [05:56]

4. Community Pushback and Organizing ([08:59]–[12:12])

• Widespread grassroots opposition:
  • Citing water, air, and noise pollution; land use; and impact on quality of life.
• Notable Canadian disputes: Alberta (water scarcity concerns), Lorneville, New Brunswick (drinking water and forests).
• US examples:
  • Memphis, Tennessee organized against Elon Musk’s “super cluster” as a health hazard.
  • Local opposition has delayed or cancelled $98 billion worth of data centre projects in just three months (March–June 2025).

"Local communities delayed or canceled $98 billion worth of projects between March to June 2025 alone." — Jayme Poisson [11:47]

5. Electricity Prices & Infrastructure ([12:45]–[14:51])

• Surging electricity demand from data centres is driving up costs for everyone by basic supply & demand.
• Utilities often pass the costs of new infrastructure onto all ratepayers, not just data centres.
• Example: Dominion Energy in Virginia spreads costs across state ratepayers, even when centres are clustered in one area.

"Just the fact that data centers are driving so much demand is making the cost of electricity go up for everyone." — Ellen Thomas [13:02]

6. Policy Responses: Ratepayer Protection & Power Plants ([15:55]–[17:25])

• US political response: Mandating tech companies to build their own power plants, aiming to relieve public grids.
• Companies are already building “behind the meter” natural gas plants and experimenting with alternative power sources (e.g., jet engine turbines).
• Nuclear (modular reactors) is being researched but is years from scaled deployment; most centres still run on natural gas.

"We're seeing jet engine turbines being installed to power data centers." — Ellen Thomas [17:11]

7. Environmental Impacts Beyond Water and Power ([18:42]–[20:48])

• Data centres affect wetlands (example: 6 acres impacted by Amazon campus in Virginia).
• Potential increase in “forever chemicals” (PFOAs) in water due to runoff.
• Rising air pollution from increased reliance on diesel backup generators, which some operators want to use more often as grid reliability declines.

8. Big Tech’s Response ([20:48]–[21:41])

• Companies like Microsoft are pledging to be “better neighbours”:
  • Promising transparency, ending use of NDAs, and covering their own infrastructure costs.
• Competing firms (Anthropic, OpenAI) issue similar promises to address negative perceptions and community concerns.

9. The Debate Over Data Sovereignty ([21:41]–[24:07])

• In Canada, advocacy for data sovereignty is framed as a patriotic issue: Canadians should control their own infrastructure/data, not rely on US companies.
• Some argue sovereignty is a legal/compliance matter, solvable by regulation, not location.
• Economic trade-off: building “walled-off” systems may undermine competitiveness because US-based global cloud platforms dominate the industry.

"The more separated you are from the global cloud computing system, the more resources you have to pour into it and ultimately it will make your business model much less competitive." — Jayme Poisson paraphrasing IBM Canada CTO [24:43]

10. Long-Term Necessity and Uncertainty ([24:07]–[26:43])

• Uncertainty over longevity/necessity:
  • Data centres may only be useful “for the next 25 years,” according to Amazon’s CEO.
  • Once AI systems are trained, some experts say the current scale of infrastructure may become obsolete.

"The useful life of a Data center is 25 years. There are experts out there who say that...once the AI is trained, we won’t need these big data centers anymore." — Ellen Thomas quoting Andy Jassy [25:53]

---

Notable Quotes & Memorable Moments

• On scale:
  "300 megawatts is a very big data center. Roughly speaking, 300 megawatts is enough electricity to power 2 to 300,000 homes." — Ellen Thomas [03:02]
• On public pushback:
  "Local communities delayed or cancelled $98 billion worth of projects between March to June 2025 alone." — Jayme Poisson [11:47]
• On tech company response:
  "Microsoft...promises to be a better neighbor. It's going to ensure that it pays its own way for its electricity. It's going to...stop using NDAs when they first come to town." — Ellen Thomas [21:00]
• On environmental uncertainty:
  "There was a higher than normal concentration of forever chemicals or PFOAs in the water...that to me was one thing that I think we should continue to watch." — Ellen Thomas [19:18]
• On long-term need:
  "The useful life of a Data center is 25 years...once the AI is trained, we won’t need these big data centers anymore." — Ellen Thomas referencing Andy Jassy [25:53]

---

Timestamps for Key Segments

| Time | Segment | |----------|--------------------------------------------------------| | 02:37 | Scope of the Saskatchewan data centre | | 03:30 | Breakdown of job numbers and types | | 05:01 | Environmental costs: water and electricity | | 08:59 | Community pushback in US, Canada, grassroots organizing | | 11:47 | Data Center Watch stats: $98B of projects delayed/cancelled| | 12:45 | Electricity pricing and infrastructure costs | | 15:55 | US policy—requirement for power self-sufficiency | | 17:47 | Exploring nuclear and alternative power sources | | 18:42 | Other environmental impacts (forever chemicals, diesel) | | 20:48 | Tech companies’ new PR and transparency strategies | | 21:41 | The argument for/against data sovereignty in Canada | | 25:44 | Long-term uncertainty about data centre necessity |

---

Conclusion

This episode delivers a comprehensive look at the “true cost” of AI data centres—far beyond economic headlines or promises of tech-driven prosperity. The conversation, rich with reporting and community voices, exposes the deep trade-offs between economic development, environmental stewardship, and national sovereignty. Listeners are left with a clearer sense of both the scale of this new infrastructure cycle and the public debate only now beginning to unfold in Canada as it has in the US.

Transcript

[A]

There's one place for the newest drops in wellness and performance and the biggest sale of the year. It's the Drop by gnc. Curating the best of what's new, handpicked by the pros who know what works. And right now, get it all. Buy one, get one 50% off during the semi annual LiveWell sale. From crushing workouts to leveling up your nutrition and everything in between, get the best deals on the latest innovations. All the newness is all on sale right now during the LiveWell sale on the Drop by GNC. This is a CBC podcast.

[B]

Hey everyone, I'm Jamie Poisson. This spring, just outside Regina, construction is set to begin on Canada's largest data center so far. Saskatchewan Premier Scott Mo says it is a huge economic boon for his province and a big, big win for Canadian data sovereignty.

[C]

Bell Canada has chosen Saskatchewan as the place to construct an estimated 12 billion dollar data center, one of the most significant sized data centers on Earth. An investment that means hundreds of new jobs resulting in a significant contribution to national Canadian data sovereignty, which I think

[B]

we're seeing lots of these data centers being built or proposed. Bell has at least seen at various stages in bc, including at least one that is already operational. There's a slew of them set to open this year in rural parts of Alberta. Of course, this is all part of a global infrastructure super cycle. In the US alone, data centers have nearly quadrupled since 2010. Amazon, Microsoft, Google, Meta plan to spend more than $600 billion on their expansion in 2026. Just to put that into context, that is the GDP of some nations. These massive server farms are the backbone of training and powering AI. But as they grow, so does resistance to them. As communities begin to ask, what are we giving up to power the world's chatbots? Ellen Thomas is an investigative reporter with Business Insider. She's been covering the AI data center boom for years. We're going to the US to see what happened when an AI data center opens up in your town and discuss the true cost of building them and keeping them open. Ellen hey, thanks so much for coming on the show.

[D]

Thanks for having me.

[B]

So let's start with the scale of this one center that's going to be built in Saskatchewan, Canada. This is going to be built on 150 and 60 acre piece of land. So that's 120 football fields and it'll be a 300 megawatt facility. That sounds pretty big to me. But how does that Compare to most AI data centers that are up and running in the US right now, 300

[D]

megawatts is a very big data center. Roughly speaking, 300 megawatts is enough electricity to power 2 to 300,000 homes. So, so about the size of Detroit. Granted, in the US we're seeing much bigger data centers being built like three, four times the size. But we shouldn't downplay how big 300 megawatts still is.

[B]

The company behind this project, Bell, is promising 80 full time jobs at this facility on top of around 800 temporary construction jobs. And just one. What are these full time jobs likely to be? Doesn't, it doesn't sound like a lot to me.

[D]

Yeah, so the boom in data center jobs really comes at the beginning because you of course need a lot of workers, construction workers, electricians, to get the data centers up and running. And in the US there's so much data center development that you have these professionals kind of going from site to site. You know, once the data center is done in Ohio, they, they might move on to another project in Indiana or some other state. However, once the data center is built, it's pretty self sufficient and you don't need a lot of workers to maintain it. That being said, the 80 jobs that are on offer are likely to be kind of a mix of technicians. They do need people who are around to manage the physical infrastructure, whether that's the racks of servers, the computers, the H vac systems, electricity wiring. They'll need security personnel that that's what's going to comprise the 80 jobs. And data centers will say, and this is true, that they often, you know, these jobs are higher paying than, you know, they're not minimum wage jobs, but you know, they're, there's a limited amount of, of them to go around.

[B]

We're still waiting on a lot of details about this particular data center and how it will work. But one of the big concerns in the lead up has been possible impact on municipal water and electricity. And I know that's come up quite a bit in your reporting about U.S. data centers. Like in Virginia, home to the largest concentration of data centers in the world, interestingly. And what was it like in the state when different communities were debating the pros and cons of data centers moving in?

[D]

Data center developers will come in with the reasoning that data centers are great neighbors. They're these big giant buildings that sit relatively empty, they use a ton of resources and they contribute to the tax base. What we at Business Insider found is that that isn't necessarily always the case. We did an investigation called the True Cost of data centers. And one of our findings was, you know, data centers are guzzling millions of gallons of fresh water a day. 40% of US data centers are in water stressed areas of the country. So water is a big concern. Electricity is another concern, straining the local power grid. With the amount of electricity that data centers need, they need to be on 24, 7. They're constantly using electricity. And that raises questions like, who's going to pay for that electricity? And, you know, if there's new infrastructure that needs to be built, who pays for that? There are questions over the amount of taxes that data centers use. In Northern Virginia, which is kind of the most mature data center market in the US There is a debate right now. The local state legislature is stalled currently over the state budget for the next couple of years because they are unsure as to whether to continue giving data centers these really generous tax breaks that the state found has been costing Virginia a billion dollars a year. Like they could be getting a billion more dollars in taxes if not for the data center tax incentive.

[B]

And just to be clear, they were given the tax breaks in the first place. Why? Because Virginia thought that it would be a boon to the local economy to have these data centers there being built. And yes, even though they don't employ a lot of people, in the end,

[D]

yes, 10, 15 years ago, when data centers, not much was known about them, um, it was relatively easy for a tech company or a data center company to come in and tout all of the opportunities that could come with this kind of economic development. But now that we're seeing that start to play out, it brings up some questions.

[B]

I have kind of an obvious question for you that might sound a bit silly, but why do they use so much water? Electricity, you know, makes sense, but why so much water?

[D]

Yeah, no, not silly at all. So when you think about a data center, it is literally a room filled with servers. It's just racks and racks of computers. And those computers get really hot. And AI data centers specifically get really, really hot because they're working extra hard. The general rule of thumb is that an AI workload takes like 10 times the amount of energy to, to process as, you know, like a traditional cloud computing function, you know, like data storage. So AI data centers are just super energy intensive, and with that comes, you know, the generation of heat. And they need these really intense cooling systems to maintain the data center. And those often take up a lot of water.

[B]

So increasingly, we're seeing more and more pushback to these projects. Right. Our producer, Elaine, was doing a Quick survey from just the last few months and there are communities trying to shut down these projects in rural New York, Alabama, California. Here in Canada, the proposal for a natural gas power plant in Olds, Alberta, meant to supply power for a 1 GW data center development, was just rejected by provincial regulators. And residents raise a lot of concerns about air pollution and noise pollution that is going to be guaranteed to happen, but to what level, we don't know.

[A]

I don't know if we quite understand

[C]

the full scale of what we're getting into.

[E]

My main concern is the emissions, the

[B]

water use, the land use you were talking about before. I think it's worth me mentioning. Alberta has challenges with water supply and consumption without any of this. And three quarters of the data centers planned in Alberta are in areas with high water stress, according to a recent analysis by the National Observer. In Lorneville, New Brunswick, residents are also worried about how a data center will affect drinking water and also old growth forests nearby.

[F]

But it's really hard for us to

[D]

believe a word you say.

[E]

Tension filled the packed Lorneville Community Center.

[D]

You can't replace these weapons. You can't replace our quality of life.

[E]

For nearly two hours, Beacon AI Center's co founder Joseph Shovlin and Voltigrid CEO Nathan oh took questions from the community about building an AI data center at the recently expanded Spruce Lake Industrial Park.

[D]

We still feel like there's a lot of details that we're missing.

[B]

Can you tell me more about some of the pushback that you have witnessed and how much it has grown?

[D]

You know, it's. I guess it's a chicken or the egg question. But all of a sudden, as data centers became a bigger news story, we started seeing more local coverage crop up. And all of a sudden it became the story. You know, these movements are often very grassroots. Sometimes they're started by the people who like, literally live next door to the data center. Like a neighborhood association will get together. Sometimes they're citizens actions groups or, you know, in Memphis, Tennessee, where Elon Musk has a big, what's called an a super cluster of computers for his company, X AI a recent health department hearing

[B]

turned into a shouting match. An executive from XAI ducking out a side door.

[D]

There's a lot of local organizing there on the ground to begin with. And the Memphis community organized really quickly around, around Elon Musk's data center. It starts at a very local level

[B]

and I mean, they've been having quite a bit of success or at least some success. Right. I was looking at Some stats from Data Center Watch, which is this nonpartisan group that's tracking opposition to data center projects in the US Only. And, and they're reporting that local communities delayed or canceled $98 billion worth of projects between March to June 2025 alone. It's just like a couple of months.

[D]

Yes. You know, the one really powerful tool that communities have when they're engaging in these pushback campaigns is they, they can exert pressure. They, I think communities are in a bigger position of power than they realize. These companies, the data centers, the tech companies, they want to get these data centers up and running fast. They do not want to deal with local opposition, any kind of delays, pushback, and they can go other places.

[B]

And when we're talking about the impacts here, I mean, electricity, for example, what impact does the level of demand that we're seeing from these data centers actually have on, like, the price of electricity, you know, in an area?

[D]

There are so many factors that go into determining the cost of electricity for consumers. I will say that the general consensus of all the power grid experts I've spoken with is that just the fact that data centers are driving so much demand is making the cost of electricity go up for everyone. Just as kind of like a general supply and demand economics thing. On a more local level, utilities are spending a lot of money to build new infrastructure to serve the data centers. And that cost is usually borne by a utility's entire ratepayer base.

[B]

Yeah, give me an example of that.

[D]

Okay, so a nonprofit group here in the US Power Lines did a study that came out recently, and they found that utilities last year asked for $31 billion more in rate increases than they had the year before. And they found that this was the case across, you know, multiple states, multiple utilities. One big example, of course, is Dominion Energy. They supply energy to about 75% of the state of Virginia. Even though data centers are really clustered in the northern part of the state, that cost is being borne by Dominion's entire rate base, no matter where they live. Because of the way that costs are spread out and allocated.

[A]

There's one place for the newest drops in wellness and performance. And the biggest sale of the year, it's the drop by gnc, curating the best of what's new, handpicked by the pros who know what works. And right now, get it all. Buy one, get one 50% off during the semi annual LiveWell sale. From crushing workouts to leveling up your nutrition and everything in between, get the best deals on the latest innovations, all the Newness is all on sale right now during the LiveWell sale on the drop by GNC.

[G]

If journalism is the first draft of history, what happens if that draft is flawed? In 1999, four Russian apartment buildings were bombed, hundreds killed. But even now, we still don't know for sure who did it. It's a mystery that sparked chilling theories. I'm Helena Merriman and in a new BBC series, I'm talking to the reporters who first covered the this story. What did they miss the first time? The History Bureau, Putin and the apartment bombs. Listen on BBC.com or wherever you get your podcasts.

[B]

This year in the State of the Union, Trump introduced a mandate for data centers to build their own power plants.

[F]

Essentially tonight I'm pleased to announce that I have negotiated the new rate payer protection pledge. You know what that is? We're telling the major tech companies that they have the obligation to provide for their own power needs. They can build their own power plants as part of their factory so that no one's prices will go up and in many cases, prices of electricity will go down for the community and very substantially.

[B]

And is that something that companies are actually doing, building their own power plants?

[D]

There's still a lot we don't know about the ratepayer protection pledge and what that might entail. But even, you know, before the State of the Union, companies here were already building their own power plants. In Texas in particular, which has a very strained power grid, data centers have been building, you know, since last summer, big natural gas plants. Sometimes they're called behind the meter projects, sometimes they're called off the grid. They're basically just private, on site power plants that connect directly to the data center and give them power. There's been so much demand for these, these kinds of plants that some data center companies are getting really creative with what they're using. We're seeing, you know, jet engine turbines being installed to power data centers.

[B]

That data center that we talked about earlier, the big one, well, big for here in Saskatchewan. They're building next to an electrical substation. But the province is also looking into nuclear power from large scale reactors. And is that something that we're seeing in the US Too? What are some of the alternative sources of power that are being looked at right now?

[D]

We see that a lot in the US we see a lot of announcements about data centers and, you know, localities researching funding, looking into small modular nuclear reactors. And the thing that I have heard experts say over and over again is that this technology is years away from being deployed at Scale. And you know, the fact the data centers have this enormous electricity demand now does create the opportunity to, you know, put money towards studying new ways to generate power. The reality is that most of these facilities are going to be powered by natural gas, which is what all of the utilities are building to serve the new, the new infrastructure.

[B]

And this I guess also feeds into the environmental concerns here that this will be kind of a continued reliance on fossil fuels to keep this industry going. Right?

[D]

Yes.

[B]

Just to keep going a little bit more with the environmental stuff. I know in Virginia, an environmental impact report done in 2024 by government concluded that an 11 building Amazon data center campus would permanently affect nearly 6 acres of wetlands. You know, I wonder if you could just elaborate for me a little bit more on some of the other environmental impacts that you're coming up against in your reporting.

[D]

In 2024, an independent legislative committee put out a study in Virginia just looking at all of the different effects that the data center industry has had on the state. And one thing that was interesting was one of the local water companies that has a reservoir that gets a lot of the data center runoff said that, you know, there was a higher than normal concentration of forever chemicals or PFOAs in the water. And you know, that should be looked at. So that, that to me was one thing that I think that we should continue to watch. I'll be interested to see what happens. We'll have to continue to watch and see if, if that eventually is studied. You know, other concerns are the diesel powered backup generators that data centers run on. You know, data centers again have to be powered 24, 7 and so they have these diesel powered backup generators that are ready to go at all times. In the US Some people in the data center industry are advocating to, because power is so constrained, they're advocating to, you know, have the data centers run their backup generators more, which of course would, you know, release more pollutants into the air.

[B]

How are the big tech companies, Amazon, Google, et cetera, OpenAI kind of broadly responding to all of this pushback.

[D]

Microsoft for one, has been pretty proactive about trying to change the narrative around data centers. You know, Microsoft has come out and said that it, you know, from now on it promises to be a better neighbor. It's going to ensure that it pays its own way for its electricity. It's going to, you know, I think they're going to stop using NDAs when they first come to town. And we've seen, you know, anthropic OpenAI come out with similar, very similar Statements, you know, they're. They're getting a lot of negative attention, and they, they. They definitely want to flip that narrative.

[B]

Yeah, I mean, it's just really interesting listening to you talk about all of this, because the United States is just, you know, further ahead, really, than Canada. And, like, I'm looking at some polling that's literally just been released now from a polling from here, abacus data. And really kind of the headline is that Canadians don't really know what to think about these things right now, that they're still kind of forming their opinions on whether or not we should have them here. Right. At a broad level, they're kind of divided on whether building more would be a good idea or not. And, you know, one of the arguments that you hear from people who are really pushing these centers here is that they are being pitched as models of data sovereignty. Right. As a way of appealing to people's sense of patriotic duty to take ownership of our data. Right. So that, frankly, you guys don't house it all and control it. And, you know, what would. What do you make of that argument? I mean, I don't mean you guys, you know, the United States and are data centers really away for Canadians to control our own data?

[D]

I understand that one, you know, one way to ensure that that data is sovereign and not subject to any foreign laws or interference. You know, one argument for that is to just build all of the data centers in. In the place where they're needed and have, you know, all of the Canadian data housed in the Canadian data centers. The flip side of that is that there are people who say the idea of data sovereignty is really a legal and a compliance issue. And you can. You can get around those issues by, you know, terms and contracts and, you know, new. New laws, new regulations, and, you know, making sure that foreign governments, foreign adversaries, can't access a country's data by law and not. Not just because of where the data center is physically located. So that's something that we're still seeing really play out. You know, I was kind of looking up some of the local coverage. It's not something that I've heard a lot about in the US But I can see how that would be kind of a compelling idea for Canadians.

[B]

You're probably not hearing about it on your side as much because, you know, you have all the big companies there, and we're quite reliant on those companies. And sort of in the face of threats of annexation, this has become like a much louder conversation. Could the United States, for example, use economic force or impose kind of controls on critical AI infrastructure, disrupting our own economy in order to extract something from, from us. You know, the, one of the other arguments I, I've heard around this important topic is that the push for sovereignty could mean kind of giving up stuff too. Like, the chief technology officer at IBM Canada essentially argues that the more separated you are from the global cloud computing system, the more resources you have to pour into it and ultimately it will make your business model much less competitive. Like this is where the United States is, where all the biggest and most fast moving companies are on this stuff. And they have like a lot of services that are kind of built around all of this. And it's, it's complicated. AI is developing at this exponential rate and the requirements for that development could be changing fast. You know, what do we know about how needed these data centers will be in the long term? Even like a decade from now?

[D]

Well, it's something that I'm starting to hear business leaders talk more and more about. The short answer is we just don't know. The data center industry has not been around long enough for us to, you know, really see whether these data centers will be necessary in 20, 30, 40, 50 years. One thing that I like to point out is on an Amazon earnings call sometime in the last couple of years, the CEO Andy Jassy said something to the effect of the useful life of a Data center is 25 years. There are experts out there who say that, hey, these big data centers that are being built for AI training, once the AI is trained, we won't need these big data centers anymore. You know, we'll, we'll move on to another, another need. So we'll, we'll just have to see.

[B]

Okay, that feels like a good place for us to end. Ellen, thank you so much for this.

[D]

Thank you so much.

[B]

All right, that's all for today. I'm Jamie Poisson. Thanks so much for listening. Talk to you tomorrow.

[A]

For more cbc podcasts, go to cbc ca podcasts.