Massive data centers are being planned in Wisconsin to help power the global boom in artificial intelligence. The data centers will add demand for electricity and water — after years of declining usage of those resources in Wisconsin.
How might that impact Wisconsin’s investments in electricity and water infrastructure? To find out more, WUWM’s Maayan Silver spoke with Tyler Byrnes, senior research associate for the Wisconsin Policy Forum. It released a report on data centers and utilities.
This interview has been edited and condensed for length and clarity.
Maayan Silver: Can you explain why it matters that tech companies are investing billions into the construction of data centers in Wisconsin after years of declining use of electricity and water in the state?
Tyler Byrnes: So, data centers do use a lot of electricity and quite a bit of water. And over the course of the last 20 years, we've seen declining demand for electricity — dropped about 9% in total electricity use, and a couple of percent in terms of one-time electricity use. And, so, because data centers use so much electricity, there isn't really capacity in the system to supply the electricity that these data centers need, because we've been sort of expecting electricity to continue to decline. The ability to generate more electricity has [diminished] as things like coal plants age and become more expensive to operate and more expensive to rehabilitate.
And so the electricity demanded by these centers is going to cause the need for increased generation capacity, so the ability to generate more electricity, both over the course of the entire year, but then also, more importantly, at one specific time. And so, electricity generation tends to sort of peak when everybody's using their air conditioner, everybody's at work, And, it's that peak demand that these data centers are going to be more impactful on because they use electricity pretty constantly. So that demand's always going to be there.
Okay, so that's the electricity picture. What can you tell us about data centers and water?
On the water side, the story's a little bit different. So, you know, statewide we sell billions of gallons of water, water utilities, the Madison Water Utility, Milwaukee Water Works, the people you write your check to for your water bill, they sell hundreds of billions of gallons of water each year. And so these data centers use a lot of water, tens of millions of gallons, and that sounds like a lot, but Racine's a really good example.
In the context of Racine, the new Microsoft data center, when it's fully built out, is expected to use about 8 million gallons of water a year. While in 1997, the Racine Water Works, who provides water for that data center, sold 2.1 billion gallons of water. And, so, 2.1 billion gallons versus 8 million, it's just a different scale. And so because these data centers, they do use a lot of water, but there's a lot of excess capacity in the water supply system in many places in the state. Now, that may be different in particular communities. That may be something that people need to look into. But generally speaking, statewide, there's more capacity in the water generating system to meet the demand of these data centers than there is in the electricity generating system.
What does that mean for the electricity generating system in Wisconsin?
So that means that we're going to need to add, at least as projected by utilities, they're going to need to add additional capacity to generate electricity. What that means is probably new power plants…And there's other factors that are driving increased demand, just population growth, electrification of cars that are also sort of contributing to that. But the data centers seem to be the driving factor.
In addition to potentially adding generating capacity, it's likely that there's going to need to be some additional transmission, so power lines and things like that, to get the electricity from where it's generated to these data centers. And, you know, the size and scope of that transmission capacity depends on the location, right? If there's a center located a long way from, a power plant and there's no existing power lines to get there, that's going to be more expensive than one that's located at a site that has in the past had a lot of electricity use.
And so, you know, the amount of money that needs to be invested in the transmission, you know, again, varies by community, where these things are located, geographic factors, those sorts of things. But the concern is that there's going to be all these added costs, and there's concern about who's going to pay for them.
Can you explain who pays for building more power plants and why this matters?
Yeah, so in Wisconsin, electrical utilities and water utilities are regulated by the Public Service Commission. What that means is that when your utility needs to change rates, they need to ask for permission from the Public Service Commission. And there's a review process. And the idea is that customers should pay for the electricity or the water that they're using.
WE Energies has come up with a proposal for a different rate for these very large users that's intended to ensure that they're paying for the cost associated with what they're using. And, so, the transmission lines, the generated capacity that's associated with this increased demand, the intention of these rates is to make sure that they're paying those costs.
In addition, there's a couple bills in the state's Legislature that are being considered that will also, you know, the intention of those bills is to make sure that those costs are paid by those users. Now, it's difficult to make, to design policy, if state legislatures could do everything they want and get it right every time. We live in a different world maybe, but the intention of both these rates and these bills are to make sure that the users are paying for those added costs.
Why is it important to keep sufficient data on how these centers are putting pressure on the state's energy and water infrastructure?
Yeah, I think, the reaction to the specific sites speaks to the public's uncertainty about these new developments. They feel the pressure of increasing costs for electricity, and they're concerned what the impact of building data centers is going to be on their electric bills. And so some added transparency in that area might demonstrate to folks that, yes, actually these costs are being paid. You know, it's when people sort of assert something, it's nice to have some hard data behind that assertion to be able to say, "OK, yeah, I can see it." And again, at the Policy Forum, we're sort of generally pro-transparency. And so, I think that's one of the reasons we pointed that out in our report.
Is there good enough data right now? If not, who would change that going forward?
You know, there's not very detailed estimates of how much one data center uses in terms of electricity or water. There was some reporting done by someone on individual data centers that are operated by Meta, and that's kind of what I used for some of the estimates in this report. And there does seem to be pretty solid estimates of the aggregate amount of electricity used by these data centers.
You know, the Berkeley National Energy Lab produced a report that had both sort of current data and then projections for the future, but that didn't have individual data center reports. And so, it is difficult to find concrete data on both exactly how much electricity fees we'll use over the course of the year or at one time, and exactly how much water they'll use. And so we're dealing with uncertain estimates as part of this conversation.
You wrote in your report that there might be some water quality concerns associated with using so much water for data centers, but that those concerns also exist for many other large users, like a paper plant or a cheese factory. Why might those concerns be limited when it comes to data centers?
The plan for data centers and the new sort of cooling mechanisms are designed to be closed-loop. And, so, there's not a pipe that's discharging that water into surface waters or ground waters. Instead, most of the water is intended to be recirculated kind of like a radiator. And so, there's going to be water quality concerns anytime you're building a large facility, if you're creating more impervious surfaces.
And eventually, this water will be discharged in some way, more evaporated. And so, you know, there, again, any large water user, if you're using it and discharging it, there's likely to be some impact on surface waters. But, you're not using it to wash off a bunch of toxic chemicals. Most of it is intended to be recirculated. And, so, it's a little bit less concerning.
As you mentioned, data centers will bring more demand for water utilities, why might increasing that water use actually benefit customers?
So when you're operating a waterworks, you have to clean the water, you have to pump the water, you have to store the water, you have to have pipes, you have to have people operating the system. A lot of those costs don't change based on the amount of gallons of water you pump. And especially if you've built your system to service a bunch of large industrial customers which are no longer there. That means that each gallon of water that you sell has to cover more of those fixed costs as sales decline.
So, if you sold one million gallons 10 years ago, and your costs are about the same, but now you're selling 500,000 gallons, each gallon of water has to cover twice as much of those fixed costs. And so if there's a large water user that's buying more, now you have more gallons of water to spread those fixed costs over. And so, potentially that will help cover those fixed costs a little bit more. And we've seen water systems where, they had industrial customers and those industrial customers left and they had to raise rates so high that they couldn't really, the users of the water couldn't really afford to pay for the system.
That is at least part of the story of what happened in Flint, Michigan, going on 10, 15 years ago now. And so, these water systems were built to sell large amounts of water. And if they're selling smaller amounts of water, the customers that are supporting the system have to pay more.
Is there an underlying takeaway that people should come away with based on this report?
I think, you know, going forward, it makes sense to be careful about making these long-term investments in the state's electricity grid and the state's sort of business environment. One thing to keep in mind is that, if we'd made projections 10 years ago about what electricity use is going to be now, we'd have said, "Well, it's going to continue dropping." And so, you know, you're making decisions based on that.
Right now, it looks like in the future, over the next 10 years, demand is going to go up because of these data centers. But if that demand doesn't materialize and we make all these investments, then there will be bigger questions about who's going to pay for these investments in new generating capacity if that demand doesn't materialize. And so we always need to be careful about making big projections for the future because that's very difficult. And I think that's something that as folks are making decisions about this, we need to consider what happens if that demand doesn't materialize?
And that's something we point out in the report. And I think, again, a five-year projection, a 10-year projection, 15-year projection, that just gets more and more difficult. We don't know what the future's going to look like. And that uncertainty is something I think we should consider.
