Data centers are large, often nondescript buildings that house industrial-scale computer servers. Those servers connect us to social media, content streaming, cloud storage and more.
Now, to help power artificial intelligence, even larger data centers are being built and proposed in Wisconsin. With them comes concerns about the impact on our environment, power grid and utility bills.
Kyle Cranmer is the director of the Data Science Institute at UW-Madison. He joined WUWM’s Eddie Morales to give a data guy’s take on the AI boom and why companies like Microsoft, Vantage and Meta are bringing data centers to Wisconsin.
This conversation has been edited for length and clarity.
Eddie Morales: Could you explain what data centers are? What sorts of things do they help power today?
Kyle Cranmer: Data centers are basically very large buildings that look like warehouses, but they're filled with computers and internet networking equipment and things like that inside of them. They power essentially almost all internet and digital services that you ever might encounter. So, for instance, if you hear that something is "in the cloud," it's actually very much on the ground and in one of these data centers. So, that includes search engines like Google, your e-mail, messaging, cloud-based tools that you might use, for instance, to back up your photos. It includes streaming services, all of the social media platforms, essentially everything that you encounter on the internet. Increasingly and importantly, it also includes all of the AI services — things like ChatGPT, Google's Gemini and some of these tools that are now being used to create AI images and videos. All of that, for the most part, is being housed and served from these data centers.
If you could provide sort of a timeline for data centers, how have they changed in size and scope? And how have they changed based on what they do and what resources they require?
Data centers have been around in some form or another since the beginning of the internet. As the internet has grown, the need for those kinds of things have grown. I think you saw a real change when something like Google first started. They needed to be able to save all that data and respond to all of those search queries very quickly. And then as people started to back up things in the cloud, they needed to be able to store more and more of that data and have backups for it. So, you started seeing a proliferation of data centers. When Facebook took off, you started having lots of people sharing images and videos and things like that. So, that was really one of the places where the need for data centers started to grow. And then with the transition of, for example, Netflix going from mailing DVDs to streaming, the data centers grew and grew. Now, AI is the new kid on the bloc that's housed in data centers, but is also very resource intensive, so you need more computing capacity. And AI is just growing so fast that people are very bullish on building out more and more data centers because they see that demand coming.
Data centers have been in the news a lot lately. Why is that?
I think data centers are at the intersection of many issues that have people's attention, including a lot of themes that people are simultaneously excited and worried about. Artificial intelligence is one of them. The data center buildout is surging largely because of increased demand for AI. Obviously, companies are very excited about AI. You hear it on the news and in advertisements constantly. But also, people are worried about this new technology and what it means, and there's a whole host of questions. In addition to AI, data centers are also tied up with issues around around sustainability. Data centers use a lot of energy and water, which feels at odds with many of the environmental and climate goals that many people have been pushing for the last decade.
And then there's the economy. On the one hand, data centers represent economic growth and jobs, but people are also worried about their electric bills going up and the affordability crisis. There's excitement about the potential gains in efficiency and the positive impacts of AI on the economy, but people are also worried about a changing workforce and that AI might take jobs away.
The other thing is that, as they're building data centers, it's leading to a lot of community engagement and debate. Sometimes that lead to drama, and these different local discussions are getting a lot of coverage in the news. Sometimes, that's framed with a big tech-versus-community, David-versus-Goliath feel to it. But really, it's a very complicated topic that hits on issues of tax incentives, the impact on the tax base, environmental issues and land use. So it's a really complicated topic, and that's why there's so much debate.
What are arguments for and against data centers, and is there a middle ground or a compromise?
In terms of arguments for data centers, one of the simplest things is just that there's a lot of demand for our digital lives — which certainly includes AI, but also goes beyond it. Our lives are increasingly digital and there's streaming of videos, so there's demand, basically. The other argument for building these out is because there's stories of creating jobs — high paying jobs, in some cases — and meaningful contributions to the local GDP and tax base. There's also an issue that's kind of tricky: for people who want to see renewable energy, maybe the demand for AI and the energy around it will be the thing that incentivizes the development of new renewable energy sources. So those are generally the arguments in favor.
For arguments against, I think one of the most visible is the environmental costs — whether it's concern about data centers being an eyesore, or because of the high electricity usage and water usage, which are big issues. There's also the question: are the high-paying jobs really worth it? Are you creating enough of them? Are they going to persist? Is it going to be a sustainable economic incentive for these communities? And then, you start getting into issues about taxes. Many states are competing for tax incentives to bring data centers to their area. Is that cutting too deeply into the tax revenue? So those are some of the arguments for and against.
Can you put these these energy costs and resources into perspective? Is there a comparison that you'd be able to use to help people better understand that?
Yeah, I mean, the energy use that people are talking about is very, very large. You're talking about doubling the amount of stress on the electrical grid. And you're talking about water use, which depends heavily on what kinds of technologies are used to cool these data centers. It's not like the water goes away, but you're pulling water from somewhere. You're using that to cool off these data centers, and then it basically evaporates into the air — sort of like when you see other large heating and cooling units around hospitals, for instance. The amount of water that's going through the system can be similar to what a small city might do, if you use the so-called "evaporative cooling" techniques. For instance, some of the big data centers that Microsoft is pushing for use what are called "closed-loop" cooling systems. The demands on water are much smaller, but you can run into other issues and it doesn't completely go away. The short answer is the amount of energy use and the amount of water use is very substantial. You're talking about doubling the amount of electricity, which is almost unfathomable.
You mentioned some of the arguments for and against. What are some examples, if you can cite any, of a sort of middle ground? What's like a best case scenario for making sure that communities aren't impacted negatively by the presence of a data center?
I think that there are many really interesting models that are being developed that are essentially about aligning incentives and trying to make sure that there's an overall positive impact on communities. You have a very complicated set of issues that you're trying to balance — issues around tax incentives. What's going to happen to the price of electricity? How is that going to hit consumers? If you're bringing in a large data center, could they be making other kind of commitments to that local community? For instance, Vantage Data Centers has some interesting projects that they're doing in Wisconsin with Milwaukee County, and it's related to the kind of Lighthouse Campus in Port Washington. That includes partnerships with UnitedWay, partnerships with food pantries and various things like that. Exactly how that plays out in terms of economic returns that these data centers are providing or sharing with communities is where the rubber hits the road, I think.
Would you consider Wisconsin to be a hotbed for data centers?
Yes. I don't think it's just Wisconsin, but the Midwest in general has many good features for data centers. These include things like: what's the price of electricity? What does the water supply look like? If you do have a a lot of water needs, sticking it in a desert is probably not the best idea. So, the Midwest is doing pretty well with that. There's adjacency to the Great Lakes, etc. But they're still hitting issues in terms of aquifers and things, so there's still plenty of things to worry about. The other reason why the Midwest is sort of a hotbed is that when you put up these data centers, you also need access to a lot of infrastructure. That includes things like transmission lines for electricity, things like fiber optic cables to connect this thing to the internet. You want to try to co-locate these data centers where you see that kind of infrastructure. And then there's price of land and things like that. All of those issues are relevant, and so the Midwest is attractive for many reasons.
Some people are concerned about the evolving tech — the presence of these data centers not being needed and suddenly changing and becoming obsolete. So, will data centers become obsolete?
That's a good question. I think that's something to be worried about. My short answer is probably not. If the data centers only served AI and you're worried about an AI bubble, then I think the worry that data centers might become obsolete is much higher. But the data centers are very multi-purpose, and they're serving all sorts of aspects of our digital lives. So, I think the investments in the data centers are going to have a long life, and they won't really be obsolete. But, the technology that goes inside of the data centers is maybe a different story. The actual computers themselves definitely will depreciate. Some of them can be very expensive with fancy chips, GPUs, etc. Those definitely will become obsolete in a few years. And so, the equipment that's inside the data centers likely will become obsolete, but the data centers themselves, probably not. Whether we're building more than we need is a different question.
_
