Narrator: The federal government recently unveiled its $2 billion AI strategy. Now it promises new jobs, literacy training, a push for businesses to scale up. Oh, and new data centres.

Government Official: We'll build sustainably. Linking new data centre development with expanding clean energy and robust environmental standards.

Jeyan: Large scale AI data centres, to be specific now. Mark Carney is clearly in favour of them, but what does the public think?

Protester: Battleford Campbellton.

Jeyan: That was protesters outside Hamilton City Hall last week, demonstrating against a planned data centre. Okay, but do they reflect the majority of Canadians? Well, according to a recent poll by Angus Reid, 68% would oppose a large AI data centre near their home. So are they right to push back? Well, we look at what's driving the opposition and separate fact from fiction. Welcome to The Rundown. There are already an estimated 100 data centres in Ontario. But are they all created equal? And is there a risk in adding more? Kathleen Koth is the chief operating officer of Mantle Climate. Mark Winfield is a professor of environmental and urban change at York University, and Craig Tavares is the president and chief operating officer of buzz H.P.C. Great to have all three of you in our studios. Really appreciate it. Kathleen, I'm going to start with you. Help! Give me the lay of the land, essentially. Do we know exactly how many data centres already exist in Ontario and how many more are in the planning stage?

Kathleen: We thank you for. It's great to be here. And from a lay of the land perspective, we actually don't have really solid numbers on that. We know that there's a lot of interest in digital infrastructure companies building here in Ontario and more broadly in Canada. Our independent electricity system operator has very good numbers of of how many of that how much of that interest is serious. But we also know that whatever is in the pipeline could be much higher as well as, as, as this plays out. And we put we put more of an emphasis on building, you know, data centres here in Canada.

Jeyan: Are we talking like a handful in the pipeline or are we talking about tenfold? What are we looking at?

Kathleen: We're looking just sort of ballpark figures. So to put this in perspective, Ontario has a current sort of peak electrical demand of 2525GW. And what we're seeing is probably at a minimum, about 1.5GW of just data centre development happening in the next decade. That's at a minimum. If things play out a little differently, it could be as much as three gigawatts or more. So you're starting to look at a pretty big percentage of our overall load.

Jeyan: All right, Mark, I think there's a when we talk about data centres, I think a lot of people have this image. And I'm hoping you can help clarify a little bit. What's the big difference between the data centres being built here in Ontario versus the ones that are being built south of us, in the US?

Mark: Principally one of scale. Okay. In terms of, of, of how large the centres are in the US and what's being proposed relative to what we're seeing here. But we are starting to see very large data centre proposals in Canada. There's there's the largest ones are in Alberta and Saskatchewan at the moment. But there is interest here in Ontario as well and in surprising places. I had people calling me from Burk's Falls last week, of all places where someone's proposing a data centre. Okay. So so there there we are, somewhat behind the curve in terms of scale. And so the US, this conversation has been unfolding for quite a while now. It's only beginning to land here in Canada at this stage.

Jeyan: All right. Craig I want to get you in here. There's something that as I was doing research, I'm learning about hyperscalers. Is that did I get that right? Yes. It sounds like a scary villain, but help me understand a little bit. Is that what we're going to be seeing here in Ontario?

Craig: We. Yeah. First of all, Hyperscaler are the large cloud operators Google, Microsoft, Amazon Web Services. So they predominantly got the title Hyperscaler because of the size and scale they build at. If you roll back the clock, just over ten years ago, we were building data centres sub 20MW, and that was the average size. When the hyperscalers came into the market, we started at 20, went up to 100. And then now what we're seeing are gigawatt scale facilities. So I just yeah, put that in context from what a hyperscaler is. We are seeing hyperscale demand in Canada more. So now the shift has become what does AI need from a compute perspective and how do we cater to that? Just as we build out these large systems across the world.

Jeyan: When it comes to what exactly is being proposed here in Canada and here in Ontario, what exactly are the main facilities related to you talk about cloud service, but is this AI training? Is this actual applications?

Craig: Yeah, it's across the board. So again, if you think about the context of a data centre, traditionally was to host a broad set of applications. The demand that we hear about today has shifted. And a hyperscaler can deliver traditional services. Or it can deliver AI based services. Buzz high performance computing. We actually focus on building AI factories. So in the context of data centres, these are large supercomputers for the purpose of using AI models and then creating innovation, research and development. We work with academia across the board Start-ups businesses, and they're integrating models into all their daily processes. And to support that, again, you need a special compute chip. It's called a GPU. So I think that's the delineation to when you think about the traditional data centres and where we are now, traditional data centres use CPUs and they were built at scale. Again, there were broad facilities that definitely took a lot of land and space. The GPU, though more dense, we build them around more efficient systems, so we use special cooling techniques. One of the misconceptions is sometimes water usage, which we use.

Jeyan: Which we'll.

Craig: Get into. Yes, yes.

Jeyan: All right. Kathleen, does that sound pretty accurate in terms of what we're sort of seeing, what the applications are here as well?

Kathleen: I don't know if this is as entertaining, but unfortunately I'm in complete agreement with everything that's been said.

Jeyan: All right. Mark, you were talking about receiving calls Burk's Falls, but you actually recently hosted a seminar, I believe, at a local Toronto community centre about the environmental impact of data centres and a lot of questions and perception as well from the crowd about data. What kind of questions were you fielding?

Mark: The big questions are principal ones about energy and about energy demand and how that demand is being met, and what the implications may be more widely, especially in terms of energy transitions, because we were anticipating we would see for decarbonisation purposes, widespread decarbonisation of space heating, transportation, industrial processes, and we expected demand growth and electricity for that purpose. We're not seeing that. We're seeing demand growth in electricity principally to serve data centres, and we're seeing that in the United States as well as Canada. And the real concern there is that load is being served overwhelmingly by fossil gas fired generation, so very carbon intensive here in Ontario. We've already been seeing a very dramatic increase in the role of gas fired generation at the grid level anyway. So there's a double concern to sort of direct impacts in terms of emissions to serve these loads, but also concerns around are we displacing capacity that would otherwise be used for decarbonisation. So that's that's the energy piece. There's a water piece as well. And there's land and and other sort of dimensions of this, but it's the energy and water pieces which are probably got the most attention so far.

Jeyan: Well, you talk about Lance. So Kathleen, I'm curious, what are the factors, whether it's here in Ontario or in Canada, that go into determining where a data centre should be built?

Kathleen: Yeah, first and foremost, it comes down to where is there power. So really over the last two years, you know, cloud growth and centralised data centres have been really a thing and growing quite steadily for the past 15 to 20 years. What's different now is this race to have the best AI, the large language models themselves, as well as enable the compute for people to use them. And that requires so much compute and data centre capacity that really it's a race to build them as quickly as possible. And that means power has become one of the limiting constraints because that's the number one input data. The data data centre is very complex, but they're also simple. In goes electricity, out comes compute whatever, whatever application it is. And a lot of heat, a lot of waste heat. That's really the quiet and some water for cooling, but that's really the equation. And so power is just first and foremost. So the number one thing is where can they find power. Okay. Now it's much deeper than that for anybody serious. It's is there fibre connectivity? Is there a community that I can start to work with early on? Hopefully this is the approach to work with them to understand what what needs to be true in order for this site to be here. That's, I think, where we're starting to see the conversation go. But first and foremost starts with power. And then one other element is the type of training, the type of compute. So Craig was alluding to this earlier. But if you're going to train new language models, you can kind of do that almost anywhere. You don't need to be as close to the user. But when you're starting to rely on the users to use your compute, you're going to want to be close to where the users are. So we're going to see a big push. Now that AI is good enough to use, we're going to see a big push push of data centres being close to city centres.

Jeyan: Okay. I want to pull up a couple of charts from an Angus Reid poll. This is about Canadian data centres and it was released last week. Let's have a look at that. So as you can see, 46% of respondents said that we need domestic AI infrastructure to ensure digital services stay under Canadian control. You can see right beside that, 25% say this is not really a concern in a global technology economy. Then 28% not sure. Can't say obviously lots of questions there. But in response to the question, if a large AI data centre was proposed within a few blocks of where you live, would you support or oppose it being built there? Well, a little bit of a different story there. More than two thirds of respondents, 68% were opposed, 15 support and 17 not sure, can't say. So. With that being said, Craig, two charts. A lot of Canadians accept that data centres are necessary, but vast majority don't necessarily want it nearby. So where should they be built.

Craig: Just to build on what Kathleen was saying, it's very application dependent, so some data centres need to be close to city centres. Why? Because the application demands the speed of connectivity to be at a certain level or metric. So in most cases, and think of the internet as that one thing that pretty much runs in data centres. So that that experience that you have today, where you go on to the internet, you want a page to load fast. Well, guess what, it's probably because the data centre is not too far away from you. And when you think about what's happening in scientific research, when you think about all the advances in technology, if you have an application that's real time where maybe you're doing diagnostic imaging and you want to get a result really fast, well, you want that short connectivity. So again, that's a very application based example of where you want that data centre close to you. But the real answer to this is yes. You want to be very conscious of where the data centre goes. You don't necessarily want to build the biggest data centre in a dense urban area. So industrial parks are ideal for data centres, maybe one hour outside of the city centre is ideal. So you still have good proximity, you still have a good you know, when you look at speed of connectivity, that's there. But again, you're not in that Neighbourhood.

Jeyan: Do you mind me asking how did you decide where to build your data centres?

Craig: It was a combination of all factors, so it was proximity and closeness to power. It was again looking for an industrial area. And then also, you know, where can I build that? Connectivity was again, a big factor that the other thing that we do too is we we actually look for renewable energy. So, you know, similar to Mark, what was saying, you know, like, let's look at the carbon effect and, you know, how do we say net neutral? Well, yeah, again, a big part of our philosophy and thesis is, yeah, we want to use renewable energy sources and participate in grid balancing. Look for things that we can do when we talk about demand response systems, which is participating in the grid and using surplus energy when there's surplus energy, but being able to create a variable load when there's not a whole bunch of energy, then maybe you could switch to an alternative energy source.

Jeyan: Okay, Mark, a lot of a lot of headlines about a proposed high tech industrial park in Hamilton include a data centre that was on the land that was once owned by Stelco. The data centre plan was rejected quite recently by city officials after opposition and protest there do help us understand, do Ontario municipalities understand what data centres are and the resources that are required?

Mark: That's a good question. I think. I think there is a process of discovery going on, and we are seeing expressions of public discomfort. Some of those are for tangible concerns over noise, air pollution, other things. There are concerns around costs, given the experience in the United States, in terms of what serving these new loads has ended up costing inside electricity systems. And indeed, we've seen utilities rejecting proposals for those reasons. I do think there's another thing going on here too, which is that that other polling shows relatively high levels of discomfort with AI, very high numbers in terms of desire to see regulation 80% plus. So I think there is there is also a little bit of an agency by proxy thing going on here that this is this is a this is the tangible expression of, of the presence of AI. And it's a way in a form that people feel they do have some agency around that they can protest, they can go to their council and sort of express some degree of control. So I think that may also be a part of the sort of political mix that's playing out here. There's very real issues around energy, water, those kinds of things. But there is also in the background these questions around the role of AI and social acceptance, which I think play into the dynamics of these conversations in a locational context.

Kathleen: Kathleen.

Jeyan: Yeah, I completely agree.

Kathleen: Yeah, I completely agree. But I also think that there's this element of this is a new industrial load that most folks haven't really been that aware of. Like we said, it's been going on for about 15 to 20 years, but mostly under the radar right now because of the sheer size and speed and scale of these things. And to your point, the AI, you know, discomfort, it's really just come on so fast. And if you look to in the past, other types of industrial development in communities, be it factories or mines or what have you, that social contract was built over decades. Yes. We are going to be extracting some things from the local area, but here's the things we're giving in return a tremendous amount of local employment for the community, support for grassroots programs that are meaningful to those communities. They're a part of the community. And so that social contract was built and tested and grown over time. This is a new type of thing that hasn't quite been figured out yet. That element, and I think that we're on the cusp of, of people that are building these things and trying to host these things. There's a great conversation to have right in the middle. It just needs to start earlier in the process.

Mark: Communities are not seeing the value proposition, okay. At a pretty fundamental level.

Jeyan: How much is that on the businesses to be a little bit more transparent as to what exactly they are going to see?

Mark: I think there are many, many layers to this. You know, that this is this is in many ways a highly disruptive technology. And people are very concerned about the social, cultural, political, employment implications. I mean, this whole the Pope speaking about these kinds of questions, you know, those things are out there and people are not seeing, you know, given that they're not seeing the sort of value proposition that you saw, perhaps when Stelco first arrived in Hamilton and translates very clearly into very high value jobs in the community. So that, I think, is part of the political dynamic as well, that you've got a combination of misgivings, that you don't see clarity on benefits. We are seeing evidence in terms of costs, in terms of energy costs, and there's underlying question of what does this do in terms of energy transitions, given that this load overwhelmingly, as I say, is being served at the moment by fossil gas fired generation, either from the grid or, as increasingly is happening in the United States, the generation is moving behind the metre to on site, but still overwhelmingly gas fired. So, so air pollution, greenhouse gas implications, noise implications, all part of the equation.

Jeyan: As you get Greg in first on some of this as well.

Craig: No, absolutely. So yeah, I think we see isolated cases in certain areas. So it's not a it's not a, you know, one, one paint brush to coat them all the kind of thing, but especially to when you see the headlines coming out of the US because again, yeah, it has been like a massive land grab in the US to build these gigawatt campuses. Canada has been more discrete. Again, what we see in Canada is actually, and coming from the federal level, a mandate, a mandate to build sovereign infrastructure for Canada. And that's super important because we're talking about data sovereignty. We're talking about compute sovereignty. We're talking about owning our own intelligence rather than renting it from someone else. And we saw this actually happen not too long ago when the Russia Ukraine war broke out. Well, guess what? We were dependent on another nation for energy. They shut off the export of that energy in Europe. They didn't know how they're going to heat their houses in winter because they shut off that energy export. Canada. Our price of energy went from $2 per gigajoule for gas to $12 per gigajoule for gas overnight. So again, that that independence and reliance on foreign entities becomes a big issue. When you think about GDP growth. That's the other effect of it now too. So, you know, back to what Kathleen was saying, we have to think of this new economy or AI infrastructure as, you know, what we're building for the future in the next generation. Previous generations built railways, telecom power stations. The future generation is actually building AI infrastructure as critical infrastructure now.

Jeyan: All right, I want to talk a little bit about regulation here. Ontario's Protect Ontario by Securing Affordable Energy for Generations Act is now law and will support what they say, quote, responsible growth of energy intensive industries like data centres, and that align with Ontario's economic priorities and benefit local communities. Is this a good policy or not, when it comes to actually regulating the sector, as we've talked about, this is something that has ramped up over the years, has been sort of done behind the scenes. If we talk about from from members of the public, but is this good policy or is there more that can be done here?

Kathleen: I think the notion of it is very, very good. What the details of of the regulation will be. It remains to be seen, but. But I think so that's let's let's kind of pan out a little bit. Last week, the federal government, Mr. Mark Carney released the AI for all strategy. 3 or 4 weeks before that, they also released the federal electricity policy within a month of each other. We said we have to double the size of Canada's electricity grid. And oh, by the way, we need to build a digital sovereign digital infrastructure as well. And all of these other things related to AI. I don't think those two things are separate at all. And I think that's what you're seeing with this regulation here in Ontario, is, as we talked about earlier, the digital infrastructure is also an energy story at heart. It's an energy and power story. How do we collectively build regulation and unleash the innovation that we have latent in Canada to, to, to conquer those two strategies together? And that will come through some of this regulation, hopefully, to your point about, you know, it, you know, in urban centres using a lot of electricity for a data centre. Well, guess what we talked about? One of the things coming off of that is waste heat. If in some of these urban environments, you're able to leverage some of that waste heat to heat nearby mixed use residential homes, that's a true that's a truly Canadian innovation that we can we can build.

Mark: I think a number of things are going on in the electricity space. I think that this has taken utilities and government somewhat by surprise in terms of the pace of growth. We see governments giving themselves the power to arbitrate access to grid supply, which is relatively new. We generally have had universal service mandates, and now we're seeing ministers in Quebec and Ontario, for example, giving themselves the authority to basically arbitrate demand. I think behind that, there's there's there's no analytical frame for doing that. There is an underlying calculus, a realisation, you know, we may need to make choices here about what what loads are we actually going to serve? And there's layers of questions there in terms of potential economic benefit. There's also a question on the part of utilities how real is this? I mean, if if we build infrastructure, is the demand going to actually materialise or are we going to end up with stranded assets? Because there are concerns that the sector is is in a financial bubble. It's hyper capitalised. And and it's not clear how that is that bubble going to last? And what's going to happen if the bubble bursts? Are we going to be left with assets and costs that we we we didn't that we now no longer have a way of paying for. There's another question there too, which of course was the the electricity strategy, which its own conversation started out being about decarbonisation. And that seems less and less part of that. And it's more about serving AI loads, especially given the emphasis on fossil gas.

Jeyan: Does this undermine Canada's climate goals?

Mark: Well, I think I think the concern is, is is this the is this the technology that's eating the energy transition we thought we were going to have? And especially given, as I say, these loads are being served by fossil gas, the electricity strategy anticipates expansion of fossil gas. In fact, says we're going to rewrite the clean electricity regulations to allow for more gas in the system, which does tend to suggest that the emphasis on decarbonisation is pretty much out the window. And we're going somewhere else.

Jeyan: I'll get you a quick response here, and I've got one question, but quickly.

Kathleen: It's of note, though, that the large tech companies, the cloud companies, hyperscalers we talked about while they are installing and investing in in firm 24 by seven natural gas to power for speed to power, that we've also seen a breathtaking pace of investment in gigawatt scale renewables from them, paired with long duration energy storage for the first time as long, as well as lithium ion storage. We're seeing huge investments in nuclear, we're seeing geothermal, we're seeing carbon capture and storage all being invested in by digital infrastructure companies. It's not to say that it's enough, but it is to say that there's never been a better opportunity to underwrite some of the major energy innovation we would have needed anyway to achieve our our electricity strategy.

Jeyan: Craig, I want to pick up on something that Mark had said. It was about the bubble. I think a lot of people will have this conversation about, you know, when something is picking up so much momentum in speed, will this bubble burst? What happens to existing data centres and those in the planning stages if this bubble bursts or and if there's less funding available to build them?

Craig: Yeah. So the capital intensity around these is massive. That that is there's no doubt we see billion dollar announcements happening again on a weekly basis. The the investment in infrastructure, again, is twofold. One, the data centre itself requires a huge amount of capital, but the compute infrastructure that goes into it requires a huge amount of capital as well too. So, I mean, outside of that, really what happens is, you know, you have a demand based environment and the demand is actually being driven about the researchers and innovation. So, you know, one assumption is, well, will research innovation ever stop and slow down? And I think the answer is no. Right. We're not we're on a pace right now. We're on trajectory right now where investment in innovation is just ramping. And AI itself is a perpetual motion now, because again, the more compute capacity that we actually harness is, the more this AI goes forward. So ultimately, if there was a case where again, we didn't have, again, the, the data centre demand load that we were building to right now, I think the shift would happen very organically where it just go from one application to another. One thing I want to call out too, when you think about investment, it's not, again, just in the data centre itself. You have to think about the all the investment going into everything around the data centre. So for example, and Kathleen kind of touched on this as well too, because, you know, when we talk about energy infrastructure, we're actually working in concert with the regulators to say and the utilities to say, okay, well, if we're going to build something somewhere, you know, how do we contribute to that capital bill that you guys have as well too, and build into areas that never saw infrastructure growth as yet? So you're building to new, new economies, new neighbourhoods. And that, that that comes with telecom, that comes with sewers, that comes with roads. So there is like this by-product that happens.

Jeyan: Something that we will continue to follow and see. Craig, Mark, Kathleen, thank you so much. I really appreciate your time and your insights on this. Thank you.

Kathleen: Thanks for having us.

Jeyan: I'm Jeyan. Thanks for watching. "The Rundown" would you be okay with a data centre in your neighbourhood? If not, why not? Let us know. Send an email to rundown@TVO.org or leave us a comment on our YouTube page. Until then, I will see you next week. Coming up on The Rundown.

Doctor: Mobile health clinics are incredibly effective at providing the sort of care that can keep people out of resorting to the emergency department. For 37% of the folks who are able to be seen, they said that actually averted an emergency department visit. So there's a long standing history and evidence base for mobile options, which is really pushing back against the mantra we've had in healthcare is that if you build it, people will come. The reality is, for folks who've been marginalised from the system, who have obstacles and barriers to reach the system, you have to go out to where people are.

Jeyan: That's coming up on the rundown.