I'm William Brangham and this is "Horizons."

Artificial intelligence data centers are popping up across America.

They're spurring innovation, but also fierce backlash because of their need for huge amounts of water and power.

Is the AI revolution worth the price communities are paying?

What we know about the impact of data centers, coming up next.

♪ Narrator: Support for "Horizons" has been provided by Steve and Marilyn Kerman and the Gordon and Betty Moore Foundation.

Additional support is provided by Friends of the News Hour.

♪ This program was made possible by contributions to your PBS station from viewers like you.

Thank you.

From the David M. Rubenstein Studio at WETA in Washington, here is William Brangham.

Welcome to "Horizons."

We're going to do a deep dive today into the controversies around the explosion of artificial intelligence data centers.

If one is or has been built in your community, you likely know what I'm already talking about.

Hundreds of billions of dollars are being spent right now constructing massive facilities to house and run the computers that are needed for the spiking demand for artificial intelligence.

These centers are being built across the country with a particular concentration in Virginia, Texas, California, and Illinois.

The biggest ones, so-called hyperscale plants like Meta's Hyperion plant in Louisiana, could have a footprint nearly as big as Manhattan.

The leading tech giants, Meta, Google, Amazon, and OpenAI, among others, are in a breakneck race to get their centers up and running, each hoping to seize the advantage and become a dominant supplier of AI.

Hey, hey, ho, ho, data centers have got to go!

Brangham: But as those centers are being built, communities across the country are in open revolt, concerned about the large amounts of water and energy that these centers require.

And it's a bipartisan backlash, with citizen activists pushing back at the speed of construction, the concern over pollution and rising utility bills, and the often opaque deals and tax breaks given to companies by their local officials.

According to one poll, seven in 10 Americans would oppose a data center being constructed in their community, and the majority would support a national moratorium on building more.

Given all of this, we wanted to examine the claims swirling around data centers to try and get a little clarity about their real impact.

And for that, we turn to Michael Webber.

He's a chair at both the LBJ School of Public Affairs and the Department of Mechanical Engineering, both at the University of Texas at Austin.

He served as an officer and consultant for several tech and energy companies, including some in the AI business.

And he's the author of the book "Power Trip, "The Story of Energy," which was also made into a multi-part series on PBS.

Michael Webber, so good to have you on the program.

I want to get into all these pros and cons and accusations and defenses about these data centers.

But first off, just establish a baseline for us.

What are these data centers?

What are they for?

Who's building them?

That's a great question.

So data centers, first of all, have been around for a while.

They're not new, but the pace of their construction and their size is really capturing the attention because the last few years they've gotten a lot bigger and we're starting to use the tools in a more obvious way.

By tools, I mean, like these AI tools.

You might've used Claude or ChatGPT or something like that.

And these tools are calling these data centers, which are these buildings that exist oftentimes outside of cities.

And they've got big warehouses, big tilt wall construction.

And so they're just large facilities, maybe it looks like a distribution center or something, but inside is filled with computer racks, just go on and on and on, these rows and rows of racks of computers that are stacked up and down.

And those data centers fill these computers are answering our queries as we use these tools.

We ask it like, "Make me a logo" or "Look something up for me" or "Help me analyze this data."

That question we type into our computer is going over fiber optics towards that data center, calling all those computers to do a calculation and then giving this back some sort of reply.

So those data centers are the place where those today, where those computations are made.

I see, okay.

I mentioned earlier that this 7 in 10 poll, people supporting a moratorium on building data centers.

I want to put up this recent Quinnipiac poll, which showed that I believe it was 65% of people would oppose a data center in their community.

And when asked why, 72% electricity costs, 64% said water use, 41% said noise.

So I want to go through each of those individually.

Let's start up with electricity.

Data centers do gobble up a lot of energy.

That is quite clear.

A lot of people argue data centers mean in my community that my utility rates are going to go up.

What's the truth to that?

I mean, that's the concern and that's a fair concern.

People are already nervous about electricity rates.

Rates have been going up for years because of having to harden the system for hurricanes and wind storms or replace the system after wildfires in the Mountain West.

So costs have already been going up for years.

And now we have data centers who want a lot of electricity.

So people are rightly concerned that rates will go up.

And I think there's a lot of political or policy concern that retirees and grandmothers will have to pay higher rates to make really rich companies even richer.

So there's a lot of concern and sensitivity about electric rates.

There has always been concern about electricity rates.

So that's not a new thing, but all the recent inflation concerns really show up in our electric bills.

So there's a lot of concern.

And the reasons why that's a valid concern is because these data centers consume a lot of electricity.

And so if you have a big new consumer come in who's going to take a lot of electricity and maybe they can outbid you for it, they might bid the prices up.

So that's a valid concern.

Those same data centers are owned or operated or used by really rich companies.

So maybe we can convince them to invest in making the grid better, which would drive down the rates for everybody.

And the challenge is, we don't know which way it's going to go right now.

Are the data centers and the hyperscalers and these data companies, are they going to invest to make electricity more affordable for all of us?

Or are they going to compete with us for electricity and drive our rates up?

That's the concern I think that they're stepping into with a lot of these local communities.

I know that some of these data centers, when they're being built, are saying, "Look, "because of these concerns "and maybe because of local efficiency reasons, "we will build our own onsite "power plants that we will put "turbines and generators onsite to supply our own electricity."

Does that alleviate this problem?

It solves one problem, but introduces other problems.

So if you're going to build what's called onsite power or behind the meter power generation, the meter being where the building occurs with the local utility, they might say, "Look, we'll build our own power plant.

"So we're not going to charge you "or cause your rates to go up or anything like that."

That might solve the problem of competing with all of us at peak times.

And maybe we should just take a moment to say that peak times tends to be afternoons in August in the summer when everyone has their air conditioners on.

We actually have an abundance of electricity almost the entire year, but there are a few hours each year where we don't have an abundance.

And if they have onsite power generation, maybe that data center can disconnect from the grid or partially disconnect and use their own power plant.

So they're not competing with us at peak times.

And that's desirable in a variety of ways.

The data center companies for the most part are actually trying to solve a different problem.

They're trying to solve a speed to power problem where they want to get connected to electricity as soon as possible because they feel like they're in a race.

And when they talk to the local utility, the utility might say, "Look, it's going to take us "four or five years to connect you to the grid."

They'll say, "You know what?

"We'll build our own power plant cause that's faster."

So they're often trying to solve a speed to power issue.

They can use that same equipment to reduce competition with all of us at peak times in those afternoons in August.

And that solves a couple of problems, which is kind of nice.

But now all of a sudden you got a power plant in your backyard and you might not want the noise or the fumes or the trucks delivering the fuel depending on what the fuel is.

So there are other problems that come with the onsite power generation, but it does perhaps alleviate some pressure on electric rates.

Right, that is another one of the issues that you just touched on, this idea of local pollution.

Because those, if people haven't seen what a big electrical turbine is, it's like a jet engine and that there can be dozens of them, if not more on some of these big plants.

That is something that maybe when you bought your property out in a rural stretch of the country, you weren't anticipating having a couple dozen jet engines firing up all the time in your backyard.

Yeah, a couple of dozen jet engines or hundreds of these diesel generators.

And diesel generators are especially noisy and fumy and all that kind of stuff.

And you can imagine hundreds of them lined up at your fence line, do you want that?

No one wants that.

In fact, diesel generators have been outlawed in California and other places, because they're so dirty.

And this is the challenge, like, "Hey, we'll build "behind the meter or onsite generation, "but is it going to be an onsite generation "at your fence line that you're comfortable with?"

That's part of the conversation.

Right.

Okay.

The next item on that Quinnipiac poll was people's concerns about water.

Data centers, the NOCs use a ton of water, why?

What are they using the water for?

If you picture a data center as like a big white box, maybe it's got, power comes in, electricity comes in, and then two things go out from the chips.

Photons go out, so you're taking electrons to photons.

The chip is converting the electricity into photons, consider photons data or information packets, and generating heat, so the chip gets really hot in there.

A lot of these chips and these computer servers interacts in these data centers.

So electricity comes in, photons and heat goes out.

And the heat's a real problem.

You have to manage the heat so you don't destroy the equipment or damage the facility or anything like that.

And so water is used for water cooling, for the most part, to cool these systems.

If you're inside a data center, there's actually a lot of air cooling.

They're blowing air over the computers and the racks and the chips, but more and more as the chips become more powerful, you need water cooling.

You can't cool them fast enough with just air, so you have water cooling inside the facility.

You also have to cool that water cooling.

And that water cooling loop, imagine like a radiator in your car, which has a closed loop of just radiator fluid going around to cool your engine.

It's that radiator's air cooled as you drive on a highway.

If you take that water cooling loop inside your radiator, you have the same kind of thing in your data center.

Sometimes you use a lake or a river or water to cool that water loop.

And that's part of the tension.

Like, "Wait, how much water are you going to use "to cool your facility?

"Is it a one-time charge where you're loading up the water "in your closed loop cooling and then you're done?

"Or are you going to use water every day "or every hour of every day "to cool the water loops for the data center?"

So it's really around cooling the chips.

And this is where there is a lot of opportunity for innovation and research.

But if you go to a community that doesn't have much water, I'm thinking like West Texas, not much water, a lot of competition for that water, and you say, "We're going to build a data center, "we're going to create all these jobs "and we're going to take all your water," this is a real sensitive point for people because it's irreplaceable.

So there's a lot of pressure and pushback on data centers.

Like, "We need you to use air cooling wherever possible.

"We need water efficiency, thermal efficiency, "reduce your water or do other things."

And we can talk about what some of the other things are, but water is a very sensitive point.

And I would say from Texas, water is a more sensitive point than electricity probably.

We have pretty cheap electricity in Texas, we'd like to keep it that way, but the water feels even more scarce.

So that's a real sore point.

Right, I mean, one of the last programs we just did here on "Horizons" was about the Colorado River running dry and all these states in the American Southwest sucking more and more out of a diminishing resource.

And if you're suddenly putting data centers on top of that, you can see why there's a lot of resistance to that.

Yeah, but the same kind of comment with electricity, the data center companies are very rich.

Maybe they can invest in more water efficiency or water production systems or wetlands restoration or desalination.

They have the money and the desire to invest in water systems.

They might end up improving the water system if we ask them to and they agree and everything else.

So it doesn't just have to be that they're going to take all the water, it can be their part of the solution.

I mean, you're touching again on a point that really does come up a lot of times where local communities look at the contracts that are signed.

Oftentimes the companies will ask for non-disclosure agreements so that the deals themselves that the data center construction companies or the data center operators sign are really opaque.

And people think, "What on earth is going on here?

"Why are they speeding this through?"

Like, are there examples of places where communities have said, "We want you to do this in the way "that doesn't harm our community?"

Are there effective examples that you would point to to say, "Hey, here's an example "where we did one of these right?"

There is actually a lot of good examples.

And I think those are useful for us to know that, okay, we're not trapped in this negative path.

We can look for a better opportunity.

It's not the standard yet.

It tends to be one-off examples.

Meta has a water positive facility they're planning outside El Paso.

Google has done a really clever project with renewables in Minnesota.

Microsoft switched to air cooling outside of San Antonio.

So you can think of the big hyperscalers.

These are brand names we've all worked with or use or know.

There are examples where they've done it in a really good way.

It's not the default, it's not the standard, and so I think that's part of all of us working together to say, "Hey, the minimum standard "should be quite good."

And then let's make it better as time goes on and we get more innovative.

But the challenge is that where the locations are, they're not always within a jurisdiction where there's really clear authority about who regulates what.

In fact, a lot of the development are in some of these unincorporated areas or maybe you have county commissioners or something, but you don't quite have like a city council and a mayor or something.

So there's a lot of ambiguity sometimes about who's in charge or how to respond.

And there's a lot of stealth that happens.

It's not always for malicious, malign behavior.

Sometimes the stealth is they're trying to get the land to build the data center and they don't want their competitors to know.

So there is stealthy behavior as always happens in business.

And that arouses a lot of suspicions because people are like, "What's going on?

"What are you going to build?

Where, why, why didn't I know?

"And do we get to be part of the process "to help you make sure it's going to be "sort of harmonious with society?"

So there are a lot of examples of bad actors that have done it in a sort of too quiet, too malicious way, maybe or dishonest way.

And then some good examples.

And I would say, well, let's push for the good examples.

We know it's possible.

Let's make that normal.

Let's not make that the one-off exception.

Right.

One of the other concerns that we have seen this in a lot of local news reporting is the noise of these facilities.

I mean, they are just quite enormous.

And sometimes, if they are in proximity to where people live, that is just a cacophony that people do not want.

What do we know about how loud these things are?

They're loud.

But there are a couple of different sources of the noise to pay attention to.

One is if you're inside the data center.

When I walk through data centers and I've been to many, you're often given ear protection, like earplugs or ear mitts or earmuffs or something of some sort to cover your ears because it's so noisy inside the data center from all the airflow.

They have to push so much air through the facility to cool the chips and computers.

It's quite noisy.

So you hear the constant hum of the fans moving the air.

That noise inside, you don't really hear outside though.

There's sound insulation for the most part.

Outside, you have a lot of generators.

You mentioned those diesel generators, maybe those like aeroderivative jet engines that are bolted to the ground to give you power on site.

Those diesel generators or gas generators or gas turbines, if those are all noisy, you'll hear that as well.

And then there's all the truck noise and other things.

There's people coming and going.

Certainly during construction, there's a lot of noise.

And so if you move to the rural countryside to get away from the city noise, and all of a sudden you've got a bunch of trucks and generators and fans and cooling towers, it's noisier, you can hear it.

That can be solved usually by distance, by getting more land and having a little more space between you and the fence line, that would help if you have the space to do it.

And that's one reason why they don't build in urban areas is the noise is, well, the land's expensive, but the noise is also a problem.

But there's several sources for the noise.

How much of this, Michael, do you think is a sort of mismatch between, we're doing all of this massive construction for the promise of the AI revolution, which right now, yes, we can use ChatGPT and all of that, and I can figure out a recipe or get an answer to some question that my son is asking me about.

But the real revolution is still in the future and the present day madness is in people's backyards.

Without a doubt.

I think we're always suspicious of new stuff anyway.

That's like kind of a human condition.

But we've got this new stuff showing up.

It's very visible, very loud.

It might be expensive for electricity.

It might take our water.

So we've got all these concerns and we don't even know what it's good for.

Maybe it's better cat videos, this kind of thing.

And that's the challenge is the benefits of AI haven't really shown up in visible ways yet, but we should get better cancer treatments, we should be able to accelerate fusion reactor design.

We should be able to optimize the grid.

Like there are all these really good things we can get out of AI, but those haven't shown up in a very obvious, visible way yet.

So we feel the impacts today, the benefits come later.

That mismatch in time is a huge challenge because the benefits are kind of like, "Oh, "maybe I get a better search engine."

There are benefits from the economic development of the jobs, but not all of us have a job at those companies or get those economic benefits.

So there's a mismatch in who suffers, people in the rural countryside who now have a noisy or loud or whatever it is, data center, very visible, ugly ones, for example, versus who benefits, which are people later on, or maybe some shareholders.

That mismatch in who benefits and who feels the impact and when we get the benefits and when we feel it, that's a huge problem.

The best example today we're getting from AI that people might understand is these self-driving cars like Waymo, which are dramatically safer than human-driven cars.

They're like 80% reduction in accidents and fatalities, other things.

So that's all enabled by autonomy and artificial intelligence.

That is a great example of AI making traffic safer.

And that's one of the benefits we'll get.

So that's one we kind of see now at small scale in some cities, but we really want all those other big benefits and then it will transform the world.

But meanwhile, we're suffering the impacts of the fumes, the water, the electricity, the land, the noise, the light, the heat, everything else.

And I think too, I want to put this poll up.

This is a Pew Research poll about people's views of AI.

According to Pew, half of Americans are, quote, "more concerned than excited "about the increased use of AI in daily life."

And the same poll, a slightly smaller percentage think it will make it harder to think creatively, that it'll worsen our relationships, that it'll have a negative effect on the economy and on education.

Do you think that that, again, is also part of the equation here, that people seem to think, "I'm going to get laid off by AI.

"It's going to make my kids dumber.

"Everyone's cheating on their homework with it.

"Why are we building these massive data centers "to support all these things that I think of as negative?"

Without a doubt, all these human condition concerns or ethical concerns, like what is AI?

Why is it good?

How is it going to change my relationship with my fellow neighbor or my family and everything else?

These are real concerns.

We see it at the universities.

We already have to figure out how to deal with it, with cheating and everything else.

So there are some downside risks, without a doubt.

And I think that's true with every new technology.

We said some of the same things with the arrival of the internet when I was in graduate school.

And Wikipedia, was it cheating to use Wikipedia?

People would argue just the arrival of computers decades before was cheating.

You should have been able to do math with an abacus or your own by hand or a slide rule or something else.

So we've had new tools come along.

We had the same kind of fights every time.

Socrates was concerned when writing was developed.

He thought if we used writing instead of committing things to memory, it would make us weak as learners.

So they were afraid of the new technology, writing.

And so this is kind of a constant thing.

We're always afraid of the new thing because we don't really understand the benefits yet.

And there are downside risks for sure of all this.

I think that like any tool that could be used for good or bad, you can use a hammer for good or bad.

It's a tool.

You can use AI for good and bad.

And for the most part, humans will use it for good.

And some people will use it for bad.

That's a real ethical concern.

But if you're worried about it's going to lay me off and make my kids stupid and take all our water and make our electricity more expensive, it's like, yeah, this is a big set of concerns.

And I think that's on us to figure out on us or everyone who's developing AI to start, "Okay, is this a tool we can live with?

"If so, do we need rules?

"What are those rules going to be to ensure "it's safe for our children and everybody else?"

So I think that's part of the conversation and a huge part of the concern.

And kind of rightly so, I think we'll get through it because we'll start to see some good use cases, but it's really hard right now.

Do you think the companies, again, you and... we both have been saying how they are all in a headlong rush to get to be first to market, to get these centers up to speed.

Do you think that they appreciate and see this burgeoning opposition all over the country and they recognize, one, they have to address it more and maybe make the case more publicly to people as to here's why we're imposing on your communities.

And two, here's what the possible benefits are.

Do you think that they got that message and that they are doing that?

I think they got the message and they're starting now.

I also think they're very late.

They did not start with that.

So I feel like they have moved ahead longer.

This is in a race.

It feels like an arms race that there's only going to be a few winners and they want to be one of the few winners.

So speed matters more than niceties or getting along well with your neighbors.

And now they're finding out, actually getting along well with your neighbors is part of the condition for having speed because your neighbors can slow you down.

So I think they're figuring it out kind of late.

And this reminds me of the hydraulic fracturing or shale revolution that happened about 15 to 20 years ago where a lot of these oil and gas companies looking to produce shale moved into these rural communities with this new technology that would unlock a lot of domestic oil and gas from these shale formations.

And they thought they'd be treated as heroes liberating us from the yoke of Middle Eastern oil and gas.

And instead people got mad because of all the noise and trucks and everything else.

The same kind of story.

And they figured out a few years in like, "Oh, we need to go talk to the neighbors "and really sort this out "because they see us as a part of the problem "rather than part of the solution."

AI companies see that now too.

They thought they'd be heroes.

"We're bringing in new tools, we're going to make you rich, "we're going to employ all these people, "we're going to beat the bad guys."

And now they're finding out people perceive these companies as the bad guys.

So they're having to play some catch up.

And that's pretty typical.

That often happens when there's a sort of a land rush or gold rush, so to speak, where they move quickly, make mistakes, then figure out, "Oh, we need to go back "and manage this better."

So they've got the message, they're working on it.

I don't think they've solved it yet.

And this reminds me of the shale revolution.

I've seen this before for oil and gas.

There are, in this conversation about data centers, you inevitably will hear Elon Musk, Google, others talk about this, that, okay, maybe there are limits to what we can do on the land with regards to these data centers and we ought to go to outer space and build them up there.

Explain the rationale why that might be feasible and why they would want to do that.

Outer space or subsea, offshore and below water, right?

So it strikes me as an engineer is just so difficult.

Offshore is so much harder than onshore because you got to deal with the waves and the wind and the saltwater, the corrosion, the water tightness, everything else.

And then that's easy compared to space.

Space is really hard.

Like these are really hard places to make things work.

To be space rated means something different than to be suitable for fuse on land or offshore.

So the idea that you might go offshore and below water or out into outer space and out of the atmosphere is driven by some sort of necessity, I think is what's happening, where they feel like it's just too slow to build the grid the way we need it.

It's too hard to get the power we need.

There are too many neighbors who complain.

Let's go where there aren't neighbors who can complain, offshore or outer space.

And so you can get a couple of things.

Fewer people complain because you're out of sight, out of mind.

You get your cooling for free.

Deep space is quite cold, the ocean's quite cold.

So it gets a lot of cooling.

So that reduces your energy needs.

But also you might be able to get solar power in space more easily than building a huge solar farm on the ground.

Now, I'm skeptical.

The engineering of that all seems really, really hard.

But if you believe that the cost to go to space will drop, the cost for solar panels will drop, that the cost to expand the grid on ground will double, then it starts to look like space might make sense.

If you think neighbors are always getting in your way, going offshore for the cooling water and to get away from the neighbors might make sense as well.

That strikes me as really hard, frankly.

But really smart people who are developing this, who are thinking about it, they're all looking at it.

So I might be wrong on this, but I would say it's going to be easier to build it on earth, on ground than in space.

But SpaceX and other companies are trying to drive down the lift costs to make it cheaper to go to space.

Maybe it will make sense.

Just in the last 30 seconds we have or so, you do believe though, you're optimistic, that we have enough agency to do it right here.

Let's do it right.

I mean, we can make these data centers power flexible, water positive, carbon negative, and really harmonious with the land.

Let's just do it.

Let's demand it of our partners.

I say, let's do it that way.

Michael Webber, it is always great to hear from you, but especially on such a controversial and confused and convoluted topic.

You're such a clear voice on this.

Thank you so much.

Michael Webber at the University of Texas.

Great to talk to you.

Thank you.

Thank you.

And that is it for this episode of "Horizons."

You can watch us on YouTube or listen wherever you get your podcasts.

If you like what you hear, please subscribe and give us a rating.

That single signal will help more people see this program.

Thank you so much for watching.

We'll see you next week.

Narrator: Support for "Horizons" has been provided by Steve and Marilyn Kerman and the Gordon and Betty Moore Foundation.

Additional support is provided by Friends of the News Hour.

♪ This program was made possible by contributions to your PBS station from viewers like you.

Thank you.

♪ You're watching PBS.