The Real Costs and Benefits of AI Data Centers
October 6, 2026 · Steve Laska, Matt Gomoll · 30 min

The Real Costs and Benefits of AI Data Centers
October 6, 2026 · 30 min
Episode summary
In this October 2026 episode of Plain Talk, Steve Laska and Matt Gomoll weigh the trade-offs of AI data centers, a story Steve says has two sides with the truth somewhere in the middle. Matt explains that AI drives demand for computing power both to train new models and to answer users' requests, creating a shortage as companies race to build. He says efficiency gains in AI models are likely to be offset by more usage, so total electricity demand keeps rising.
On costs to communities, Matt says data center operators may be willing to outbid households for electricity unless regulators or local deals cap prices, and some projects bring their own generators as a stopgap. On water, he explains that evaporative cooling is cheaper but uses a lot of water, closed-loop systems cost more but do not waste water, and air cooling uses more electricity. Steve pushes back on the idea that data centers can simply go in the desert, since the grid is not there, and both say the US is behind on power generation and needs grid upgrades.
Steve points to construction jobs and local spending, while Matt adds the ongoing tax base. Matt says he does not expect overbuilding soon because there is no idle compute today and projects can be stopped if demand fades. For responsible development, both favor community benefit agreements that require standards and local benefits, which Steve describes as reversing the usual tax incentives.
Key takeaways
- Matt Gomoll said AI needs computing power both to train models and to respond to users, which has created a shortage of data center capacity.
- Matt expects gains in AI efficiency to be offset by greater use, keeping total electricity demand rising.
- Matt said data center operators may outbid households for power unless regulators or local agreements cap what residents pay.
- Matt explained that evaporative cooling is cheaper but loses a lot of water, while closed-loop systems cost more and avoid wasting water.
- Steve Laska said placing data centers in remote deserts is harder than it sounds because the grid and connections are not there.
- Matt said there is no idle compute today, unlike the unused fiber laid during the dot-com era, so he does not expect overbuilding in the near term.
- Both hosts favored community benefit agreements that set building standards and secure local benefits, such as protections on residents' power costs.
Questions this episode answers
Do AI data centers raise electricity bills?
They can, according to Matt Gomoll in this October 2026 episode. He said data center operators may be willing to pay more for electricity than households, and if regulators let the market set prices, companies could bid up power costs for consumers. He said local governments can negotiate caps, and that some projects bring their own generators to avoid stressing the grid while new power plants are built.
How much water do data centers use?
Matt Gomoll said in October 2026 that it depends on the design. Evaporative cooling is cheaper to build and uses somewhat less electricity but loses a lot of water. Closed-loop systems cost more and use more electricity but do not waste water, and air cooling avoids water use but needs a lot of electricity. He expects communities to push back on the cheapest, water-heavy designs.
Do data centers create jobs?
Steve Laska said in October 2026 that data center projects create many construction jobs for electricians, steelworkers and others, plus business for restaurants, services and housing nearby. Matt Gomoll agreed there is an economic boom during construction, said ongoing staffing is modest, and pointed to the long-term tax base as a key local benefit.
Are companies overbuilding AI data centers?
Both hosts said they do not see it soon. In October 2026, Matt Gomoll contrasted today with the unused fiber laid in the late 1990s, saying there is no idle compute now and a backlog of demand. He said demand should outpace building, and that if excess capacity appears, unfinished projects can be stopped. Steve Laska said he believes, in his opinion, the industry is far from overbuilding.
Plain Talk reflects the views of the speakers as of the recording date and is provided for informational purposes only. It is not a recommendation to buy or sell any security or a guarantee of future results. Market conditions change; talk with your advisor before acting on anything discussed.
Full transcriptRead the full conversation
Automatically transcribed from the episode audio; may contain minor errors. Select a timestamp to play from that point.
Welcome to Plain Talk with Trillium Wealth. I'm Steve Laska, the managing partner here at Trillium Wealth. I'm here as usual with my partner and chief investment officer, Matt Gomoll. Matt, I think we're going to talk about something that is ever present in the news for the last couple of years, data centers. Yes, the media has gotten a hold of it. It is cross-current of politics, finances, economics, AI. This story has got it all. So, yeah, we're going to dive in today and talk a little bit about it. So, you know, we have just even talked to our producer about this a little bit, and my posture even coming into this since I did some more research and things like that is it was more apparent than ever of the old adage. There's two sides to every story and the truth is in the middle because if I listen to some, I'd be like, we should have none of these. And if I listen to others, I'd be like, why don't we just have them on every street corner.
So, first, like, why do they need so many data centers? Yeah, so we've data centers have been around for a long time, and they're just, they're pretty common and the internet runs on them, right? So, picture, if you're watching Netflix movies, those Netflix movies are housed on servers that are in a data center somewhere and you're just accessing it to stream it, right? So, that stuff has been out there, but in the last few years with the launch of AI, it is AI consumes a ton of compute both for training, right? So, if they're training a new model, they're just compute intensive for that. And then also, if I'm sitting here, you know, I'm ChatGPT or Claude or Gemini or whatever, and I'm asking questions or I'm having it do work, that is also using it. So, both aspects of the AI companies using a ton of compute to build new models, and then the users out there using it, all of that stuff runs through data centers.
And so, the demand for compute has just skyrocketed in the last few years. And because of that, right, so growing AI adoption and people finding more things to do with it. And the companies themselves using more and more compute to create these more and more advanced models, all of that leads to there's a shortage and people are trying to build these data centers as fast as possible. Yeah, and I think, you know, along the lines, what you just said, people have to understand if something's actually doing work versus just putting something out. So, for example, if you ever watch your power on your iPad, you know, I travel a lot, you know, you're on planes and all. If you download a movie and then you watch it after it's downloaded, it eats up very little power. If you stream a movie, it eats up more power.
Now, if you sit down and send emails and open some work apps, it takes even more power. So, AI is working for one. That's one of the big differences out there. What about electric bills? That's what everyone wants to know, right? Like, I mean, I thought this was the truth too. Like, we're going to go into a grid. We're going to put a serious, serious stress on the grid. Right. And it's going to cause everybody's bills to go up because I live in a town where we have our own grid. And normally that doesn't happen in a small town like Winter Park, Florida. But as a guy that has an electric car, when I lived in Orlando, it's like, don't even try to charge your car during the day because the cost will be double because there's so much need for the electricity that they're going to charge you more for it. What's been the talk about kind of consumer cost on electricity being like, I'm on the same grid as this data center is, and they're making it work overtime, and they're going to charge me more during prime hours. Yeah, that's, it's a major concern for. Yeah, there's, there's several concerns that we'll get into that, that the public has that if there's going to be a data center in your area, you're going to sit there and say, I don't think this is a good idea.
The company is are willing to pay more for the electricity because they really need it. And they might be willing to pay more dollars than you're willing to pay or are comfortable paying. And that could if if the electrical companies and the regulators right these these power companies are regulated, if they allow that to happen and just let market the cert determine the price. And so, you know, the people paying for compute and Amazon and and meta will just bid up the price of electricity and that's a real problem for consumers. Now, there are either, you know, you can, you can make, if you're a local government, you can make deals where you tell them that's not allowed we're not we're going to cap the price. And then there's a tax amount in there. The other part here is, there are behind is called behind the meter. It's basically you show up with a data center project and a giant generator.
Right. That's why some of these electrical component companies GE and Caterpillar makes giant turbines that you can generate electricity with. And so, you know, they're bringing power their own power to the site. And this is kind of a stopgap measure to get a data center up and running without stressing the grid while the power grid and the electrical companies build more power plants. Right. We have a shortage of power in the country and we we know we're going to need more. Yeah, and I think that's kind of where I see some things are disingenuous out there. Right. So I think it was Sam Altman. I was watching this past week on a video go, you know, hey, data centers don't care where they are. And I understand why people don't want to have them near him. He said, you know, I don't want a nuclear reactor next to my house, even though I know it's safe and I. It's the best kind of energy. So he's like, we should just put him in the desert.
Where I think that's disingenuous. It's like, well, there's not a grid out there right now. Right. It's not snap your fingers and it goes there. And the other two parts I think we're missing here is I still don't think they have a handle on how many do they need. Like what is it like? What's the actual number of these things that they need? Right. Like, we don't. We're already using a lot for all the reasons that that you just said. So those kind of projections and I think those those bother people. And again, I think this just kind of put it wherever that would be a great idea. That would be fantastic if you could put them out in the middle of nowhere, even, you know, our country has a lot of public lands and I'm not trying to do damage to stuff but it's like we have millions and millions of acres of public lands. It's just not that easy. And then finally the other point where I think it's disingenuous is there's kind of this. Like just whistle past the graveyard and go, I'll solve that energy problem.
Well, AI can't change the laws of physics, right? They can't like energy has been an issue for a long time. Most things with AI, they can do it faster than humans but it's not like we haven't looked at these things and you if people think suddenly, AI will fix itself and it'll barely use any energy. That's just not, you know, science is still science. So these are things we can't necessarily overcome. Yeah, the, you know, AI is the models themselves are getting more efficient and less costly, right? So they're using less compute because they're getting more efficient. However, if something's more efficient and it's working for you, you're going to use more of them. Right. So the efficiency gains are going to be offset by increased usage, which just is going to land us at more total compute more total electricity being used. And yeah, there's, there's a real physical problem here of electricity needs to be generated and then it needs to be distributed to the places, right? So just to build the data center in the middle of nowhere.
You got to get electricity to it. And now you got to get signal, right? The wiring that goes back and connects to everything else. It's, it's not an easy problem. And that also gets into what the, you know, like the environment, these things are built in, right? Has an influence on the design and the cost and all these things. So like, put it, put it in, you know, hot Arizona or Nevada. If we think we got problems cooling the data center and the chips right now. Put them in a hot desert. Going to be pretty rough. Right. So what about that? Like, what, what are the real environmental and, and a cost to neighborhoods out there because they're not put it by the way, if it was that easy, they would have done that. Like, I talked to a friend of mine about this. So like a lot of times I do research, just real people. He lives in Virginia. He was like, see, we're the data center capital of the world right now, at least of the United States. And he's like, like he's all for him, by the way.
But he's also like, cause a lot of traffic. You know, maybe they could build more trees around him and stuff. But again, we're looking at it through a lens of, oh, there's one of them over there. Would have suddenly, and like I said, obviously they could just put them out in the Nevada desert. They'd have done that. But instead, they're in Tysons Corner of Virginia. So what's the real cost? Yeah. And the, and this gets into this, this. There's also a. Call it a style factor, but just the scale of things has expanded, right? Like I've looked up data centers and there's websites. You can look up data centers near you and see what it is. And a lot of them are just in office buildings. They're just, you know, it's, it's a room. Building with a bunch of stuff in it. And that's it. You don't really know what it is. Now they're doing these mega projects that look like an Amazon. Warehouse, right? The giant footprint.
Tons of compute packed in there using power. The, I was listening to a really good podcast the other day talking about these trade offs, right? You'll hear people say they use a ton of water. They use a ton of power. Well, yes and no. Some designs use a ton of water, right? If you have evaporative cooling, which is just, Hey, we're pumping water in. It, you know, gets evaporated and the vaporation is the cooling mechanism, just like sweat, you know, evaporates off your body and helps you cool it down. That's using a bunch of water. It's, it's lost, right? That water is evaporating and it's getting lost. That is cheaper to build uses a little bit less electricity, but waste a ton of water, right? That's not a great design. Those designs are absolutely out there. I think there's going to be a lot of pushback on those designs, right? A community that's potentially thinking about adding one of these data centers might say, no, no, no, no, you're not going to build a cheap version and use all our water.
Now, there are other versions that are more expensive to build use more electricity, but don't waste any water, right? Close, close loop systems where they're, you know, it's, it's a water mixture with other stuff that helps cool it down. It's just more expensive, right? So now that's a, that's a, you got to get the company to just pony up for the more expensive version and negate the water issue. And then lastly, like you could just air condition, right? We air conditioning our homes and we're not spending waste in any water. And you got the electricity issue. A ton of electricity. Yeah. I mean, it's through the roof, right? So, so like there's these trade offs and then you think about where are you putting these again? The climate matters in a cold climate, you might have to cool some of this stuff less. But you might, you know, then there's, there's the transmission issues, there's power issues. So it's these things are solvable and a data center by itself isn't like an evil thing.
It's just a building with computers in it. But the choices that we make and what to approve and what style and where's the electricity coming from and in what type of water usage it is. And it's just matter a lot to a community and should, you know, these companies have a lot of money and they probably should spend on doing things that make it sustainable and right for that to fit in that community rather than the cheapest thing possible. I think they're going to and I think they're willing to because we see the amount of money they're paying for land right now. And so, I think that's the way that we're not trying to negotiate. They're just like, Hey, how can we get this done faster? And I think that most tech companies look at some of these things you're talking about as it's kind of a one time cost and then things run very efficiently. The questions that I'll have in the future, like when I again when I'm doing research on this and I see a very smart guy up there and say, well, you know, it only takes up a data center takes up the same amount of water as a golf course. And I'm like, Hey, we're not going to look golf courses.
It's like, yeah, but I don't have eight golf courses in my town right now or 10 golf courses in my town. Number one, I'm a golfer, as you know, golf golf golf. There are times when golf courses go. Yeah, we can't use water right now. It's all going to dry out and a data center doesn't get to do that. So when they use the golf course analogy, again, those are the things I find disingenuous. So what I want is honesty and what what's the solve. There are plenty of solves for this. I saw something. It's on Netflix recommended to everybody. It's called Inside Bill's Brain. It was about Bill Gates. And he's a great problem. So people don't know a lot of things. And part of the problem they were solving was with nuclear reactors and him and a team of scientists basically nobody had thought about it before said, cool them with liquid metal liquid metal boils at 1100 degrees water boils at 212 degrees. You're never going to overheat and have an explosion with liquid metal. You can make them small and more efficient. I think these are the kind of things they could come up with. Right. Now then again, you know, if you had a building full of mercury, I'm not sure people will be happy about that either. Right. But but other way there are solves and I do believe in it.
You know, there is a part of me that wishes that that actually because I agree with you. I think there's a lot of stuff we need to do on our grid today. And by the way, we're still more advanced in most countries out there like we are distraught when there is a brown out even right or a blackout. I mean, that happens in other places in the world. All the time, all the time during the summer we're just a little spoiled here and I like that by the way I like being spoiled. So, yeah, we should be right. I do think we are we are woefully behind on power generation. We've we've kind of we've made it very hard post, you know, the 70s with the kind of backlash against nuclear power since then. That's come, you know, it's extremely hard to build and really costly to build that stuff. For environmental reasons, the cheap dirty coal stuff really hard to be. You know, we don't even approve that stuff. And so we the you know we lean into natural gas cleaner than coal by all means. Not nearly as clean as nuclear.
But China has a huge leg up on power generation, partly because they just use coal. They don't mind. They don't mind putting in a dirty power plant and they can throw it together fast and just get more power. And it just, you know, so much of the future is leaning in that direction of we need power to to for economic growth and AI and electric cars and all these other things. We're behind the power curve for sure on electricity generation. So these two things need to happen like we absolutely need to push more electricity capacity, smarter grid, better grid, our grids really vulnerable and weak and like it's prone to breaking. So that's not good. So upgrade our whole grid more power capacity. You know, we're these these data centers are going to get built. I, because we have, you know, a state based system where one state can do one thing.
One state can ban ban these things outright in another state right next door can say, no, we're open for business. We're going to have solutions. I think the local communities are going to start to realize they have some power, right? They got leverage. And so if you know, another place that that outlaw our bands, a data center, another community says cool, we'll let you build here on these conditions. You know, guarantee low power costs for our citizens, certain tax incentives for the population like you're going to pay tax you're going to do this, not tax breaks for the data center company. They can run around water use. They can, they can make it so that if there is stress on the system, your data centers disconnected before that house is disconnected, right? So like all these things are possible.
And it's just a matter of the local governments finding the right path and using their leverage. I think though, you know, it's a long way to go and it'll segue into my next question a second, but I think that like, you know, some of these, some of these answers, they sound really great on on paper, right? Like, do I believe, I mean, does any citizen believe that, you know, we're going to make sure that you can watch TV tonight and shut down commerce. That's tied to this data center. I mean, it's absurd to believe that, right? The ones that are holding out. Don't understand, I think, eminent domain laws, or eventually governments can come in, including the federal government. And I think some of the issues we can potentially have going forward, Matt, and you and I are always been kind of followers of this is depending on what party or what group is in power, the environmental protection agency last time I check was a federal agency. And so states can institute some rules and the EPA can come in and say, or the opposite, right? The opposite can happen as well and can really cause quite a quagmire.
So let's on the next question, let's stand what potentially could be the more positive outcomes. I know there's negatives as well, but when we look at like, what an economy can gain and does the community get a fair share, the naysayers basically say, Hey, these things don't take a lot to maintain. There's not a lot of jobs created there. Except for there is a lot of jobs created by construction. It would be like saying, don't build that neighborhood because the people building the neighborhood. The neighborhood doesn't create jobs long term, but it does for three, four, five years. And if you look at how they're building data centers in regions, it is going to create a lot of jobs. It is for electricians, for steel workers, for people producing this. For people producing the steel for construction. And then if you have all those things going on, for restaurants, for services, for apartment complex or home sales in that area. Because again, these are not projects that you go or build in a data center. Okay. I got two weeks worth of work for you. Right. I mean, these are big projects that create really long term and seemingly high paying jobs. What say you?
Yeah, I think it's, you know, the construction crew is not going to be there five years from now necessarily, but there's certainly a economic boom during that construction phase. And then, you know, there are some jobs granted that you know, these things aren't packed full of people. There's some jobs for maintenance. But just the ongoing tax base of having this asset that you can tax, whether, you know, whether it's property or revenue or whatever it is. It's. That's up for community to decide. You know, what's what's the harm to our community. And again, try to mitigate the harms that would be especially water and power, like those are not great noise. So like work on mitigating those before anything gets built. And then, yep, got building sitting over there doing its thing and we're taxing it. It's, you know, these things are they're going to land somewhere.
And so, you know, every community's got to make its own decision, but I don't see. I don't, it's, it's a really popular thing for either side of politics to rail against. It's a good boogie man. It's not a race we're going to lose to the Chinese. Because we don't want to. We don't. Yeah. And people are wondering out there, you know, what kind of constraints could help AI from overbuilding and I'd like your view on that. But I'll tell you, part of my view is we're so far away from overbuilding right now. And, and Elon Musk, you know, prove that he built a bunch of stuff going. Somebody will use it. And now he's making $2 billion a month because people are using it, right? And I think when people look at places like China, you know, I've never been, but I've been told by people that have been there, not just read about it like they really did overbuild cities. They're like, we'll need them in the future. And they built these buildings that have nobody in them. That's not the same as a data center man.
Like that's that's about people living in those and you can look at population and say, when would I need those where people going to move from and you can overbuild on that. But there's nothing my opinion and this is opinion, but I think it's an educated one is we are so far away from the time where we would be overbuilding these. Because again, the appetite. We haven't even touched on the appetite for these things like you and I and our whole firm. We use a lot of AI. But that's not normal when I talk to other people out there like other people out there. I talked to go, yeah, I use it. I use ChatGPT. Or I use Gemini. I use Claude. It's like, no, I mean, we're using it using it on a bunch of different things, right? You and Cherry are using it on planning tools. You'll use it on investment tools in the future. All these things are going on. So what are your thoughts about do we even need to think about constraints for overbuilding or we just like so far away from that. That's that we have bigger problems to solve before that. Yeah, the here the analogy a lot of people like to use is how much fiber optic cable was put into the ground in the late 90s around the dot com bubble.
Right. So there was a term dark fiber, which is just people were companies were bearing fiber in anticipation that it would be used dot com bubble burst. Fiber set dormant and really underutilized at least. Well, our streaming infrastructure is built off that fiber. So it was early. It was used today. There is no dark compute. There is no data center that's sitting, you know, is is is full of chips and power and not being used. We have a backlog of demand. Could you know the economics of a company's not work out. And so nobody, you know, in the end, this isn't a viable business that people end up paying for. Possibly not very likely, but you know in the realm of possibility.
But until that occurs, there is going to be a steady increase in demand. Demand is going to outpace build. It just is where we're like. We are not going to physically build this stuff and the power facilities fast enough to keep up with the demand for compute. So until, you know, I don't think that we're going to over build. And it takes so long to build these things that at the point where you start to see excess compute in the system. You can just turn off projects, right? You just don't. You might, you might be partially built and you just stop building it or you just, you know, you were going to go under contract on that next one that would be open in five years and you turn it off. The stuff that's being built today that's going to finish over the next five years, it's going to be spoken for the compute capacity is going to be spoken for. All right, so take us home on a final question.
And again, let's keep it. This is, I think, a good positive way to end. So what do you think responsible development could look like when you look in the stuff that you've read, I've read everything you're hearing out there. Like, what does it look like that everybody would be like, hey, that would be responsible like, you know, there's always give and take to kind of move forward in life, but where the, you know, where they would probably be a smoother path for the people doing it. What does it look like to you? Yeah, so kind of my take on this came from an article you sent me that was talking about community benefit agreements where the community gets a say, right? So instead of just a yes or no, we're going to block it. Or we're going to let it happen with zero constraints. Do something like this that applies certain standards to how this thing is built and what benefit the community gets out of it after it's running. And as long as that's, you know, fair.
The, the company still wants to build it there's still going to be profitable for the company to build and operate this facility. And the community gets a fair shake out of it. Right. So I think that's how they can protect themselves. They can protect themselves against, you know, their electrical prices going up, water being overused noise location, and so they can do some stuff that says, all right, you can build, but do it to these standards. And also, we want something of benefit from this whole thing. Probably could do the reverse. That might look like, yeah, go ahead. Probably could do the reverse of what used to happen. It used to be when you went and built something because it created jobs and beautification and all those things like a city or a state or whatever would give a tax abatement. Hey, totally. We're going to let you stay there tax free for the first five years or you'll see football stadium is built like no tax for a decade or whatever it is. In this case, maybe it's the reverse right. Maybe you're looking at those comes because.
Like look, the economics of these things is they make a lot of money in the future with very little cost. It's the it's all the upfront cost. So if I were in a committee, for example, I would actually be looking and saying, okay, if that data center wants to be there. And their tax should be $500,000 a year on it. It's $3.5 million a year to the school fund. Whatever it is. I think if you did stuff like that, more people would be like, hey, they're going in somewhere. And what I said earlier about eminent domain. Start up in people like saying, okay, let's to your point, let's get something out of this. Right. As opposed to they just force it. Because they legally can. Again, people don't understand the eminent domain laws. They can come in and go, this is a necessary thing. For the community and blah, blah, blah, blah, and put it up. Right. And I'm going to tell everybody out there that they put it up next to the mansions. Okay. So think about that in your future too. Matt, I knew you were about to say something to go and take us to the end here.
Yeah. No, I think that's that is the way forward. Just like you said, reverse incentive. Right. We're not giving you a tax incentive. You're giving our community guaranteed minimums guaranteed minimums on tax guaranteed minimums on like power, electricity subsidization so that consumers can feel it at home and say, Oh, cool. Like, my power bill actually went down $100 a month because that company's subsidizing me. Yeah. Like, these are, these are possible. So, getting the, getting the environmental impact. Right. Water, power, aesthetics, and noise and all those things, and then get extracting some value for the community. It's a pretty good solution. But it's, it's all these communities on their, you know, having to figure this out on their own. They are at a disadvantage going up against these, you know, these hyperscalers who know what they're doing, right? So, so.
I think we'll see some examples and that example of a community doing this well and building a good deal for themselves will serve as a model for others to kind of replicate. Love it, Matt. Appreciate it. As always, great place for us to close. As usual, click and subscribe. We'd love to see you pass this on other people. We're trying to be informative to not just our clients, but everybody out there. So, on behalf of all everybody at Trillium Wealth, thank you and we'll hopefully see you next week.
Keep listening.
What the Chip Selloff Really Means for Investors
August 11, 2026 · 24 min
Episode notes →TechnologyA New Fed Chair Nominee and AI's Disruption of the Software Sector
February 12, 2026 · 32 min
Episode notes →TechnologyIs AI in a Bubble? Separating Hype from Investment Reality
December 4, 2025 · 33 min
Episode notes →Start with a conversation.
Tell us what is on your mind. We will match you with an advisor near you, or meet by video.