This program is brought to you by the members of WCNY.

Thank you.

With Pomeroy Markers, history is based on the facts and vetted by professional historians.

Over 3,000 installed nationwide.

From Maine to Florida, to California, to Minnesota, to Texas and Alaska.

Since 2006, our markers have helped people celebrate community history.

Marker grant program information available at wgpfoundation.org.

[ MUSIC ] [ MUSIC ] KATKO: Welcome, America, to "Balancing Act," the show that aims to tame the political circus of two-party politics.

I'm John Katko.

This week, data centers.

Are they a boon or a burden?

In the center ring, we're joined by data center expert Jonathan Koomey.

On the trapeze, Elena Schlossberg of the Coalition to Protect Prince William County, and Dan Diorio of the Data Center Coalition, take swings at the issue.

Plus, I'll give you my take.

And then Did You Know, when it comes to data, big things come in small packages.

But first, let's walk the tight rope.

Every time you stream a movie, use online banking, ask ChatGPT a question, or store photos in the cloud, somewhere, a data center is helping make it happen.

But the cloud isn't a cloud at all, it's computing power and data stored on remote servers.

These servers are a type of computer.

Many of them are housed in data centers and accessed over the Internet, instead of from your own computer or smartphone.

The idea isn't new.

It goes back to the giant computing rooms of the 1940s.

Then came the Internet in the 1990s, followed by cloud computing, streaming, and smartphones.

Now, artificial intelligence has put data center growth into overdrive, and these facilities are crucial.

They support banking, health care, communications, government, national security, plus an AI race in which China and other countries are also heavily investing.

Supporters warn that without enough computing capacity, America risks falling behind.

But the cloud, it turns out, takes up a lot of space here on the ground.

Hyperscale data centers can be massive, windowless, industrial buildings, surrounded by cooling equipment, generators, and electrical substations.

For people living nearby, that can mean a very different look and feel for their communities.

There's also the noise.

Fans, chillers, and other equipment can produce a steady hum 24 hours a day.

Then, there is water.

Some cooling systems can use substantial amounts of water, which is critical for the data centers to operate properly by keeping their computers cool.

The bigger universal challenge, however, may be electricity.

Data centers used a stunning 4.4% of all U.S.

electricity in 2023.

The U.S.

Department of Energy projected that could reach between 6.7 and 12% of our energy resources by 2028.

And they need that power around the clock 24/7 365.

Wind and solar can be part of the mix, but reliable 24-hour demand is also renewing interest, and alternative sources of power, including, but not limited to, natural gas, geothermal, and nuclear power.

This includes restarting closed nuclear plants, like Three Mile Island, in Pennsylvania.

Some policymakers, including President Trump, are even asking whether new data centers should have to provide their own power or pay the full cost of adding it to the grid.

For communities, there are benefits.

Data centers can bring billions in investment and construction jobs and tax revenue.

But they may create relatively few permanent jobs once they're built, because the buildings are just loaded with computers.

And the backlash is growing.

This summer, New York State became the first to impose a statewide, one-year moratorium on new, hyperscale data centers.

They claim they need the time to develop new rules to protect ratepayers, communities, and the environment.

One thing that does seem to compute is that America will need more data centers in the future.

But where should we put them?

How should we power them?

And how much should the communities that host them be asked to give up?

Let's process those questions as we ask an expert in that data center raCe.

Joining me now is Dr.

Jonathan Koomey, a leading expert on data centers, computing efficiency, and energy use.

He's also the author of research, so influential, there's even a law named after him.

Welcome, Jonathan, and tell me about your law.

KOOMEY: It's called Koomey's law, some people call it that.

We did research on the historical improvements in energy efficiency of computing over many decades.

And what we found was that the computing efficiency, energy efficiency doubled every 1.6 years.

And that means over time, starting in the mid 1940s, computers just kept getting more efficient at a ferocious pace.

And ultimately, that trend has allowed us to have iPads and cell phones and other small electronic devices that operate on batteries and allow us to be very mobile.

KATKO: With respect to data centers, how many are there in the U.S.

and how are we doing against our main arms race rival in this area, China?

KOOMEY: There are thousands of data centers with varying sizes.

Some are extremely big, like the power uses that of a mid-sized city or a big industrial plant, but some of them are pretty small.

So there are probably, several thousand in the US as a whole.

And data centers are not a new thing; we've had them since the late mid to late 1990s.

It's only in the last 6 or 7 years that people have started to think about AI data centers, which are generally bigger than more conventional data centers that are mostly familiar with.

KATKO: So why are the AI data centers so much bigger, and it seems like the subject of such consternation now?

KOOMEY: Well, it's that there are economies of scale.

If you have a very big facility, it means you can spread your costs, your fixed costs, over a larger number of computations.

So, the cost per computation is much lower when you have a bigger data center.

KATKO: So what are some of the concerns associated with these mega data centers?

KOOMEY: Data centers use electricity.

Some of them use water, and anytime you have a very large facility, of course, the worry is that the amount of electricity they use or the amount of water they use could be quite substantial, especially if it's placed in a small community, but every data center is different.

Unfortunately, people take the most extreme cases, and they apply the energy use or the water use of an extreme case, and they say, well, the data center in my community is probably going to use that much, and that is almost certainly wrong.

KATKO: So let's talk about some of those articulated concerns.

What are these data centers generating that makes people in these communities so upset?

KOOMEY: All data centers use electricity.

Every time you send a query to an AI model or you type in a Google search, that information is then transmitted over the network.

It goes to one of these large facilities, and the computer analyzes it and then sends information back to you.

In order to do that, of course, you have to use electricity.

And the question is, what is the value of the query to you?

And it turns out, a lot of these activities, a lot of the services that computers and data centers deliver are very valuable to businesses and to people.

So that's ultimately the question that we as a society need to answer, which is, is it worth it?

And in a lot of cases, computing services are so valuable that it's absolutely worth it.

But there are real costs, and the question is, you know, there's energy costs, there's water costs in cases where there's a data center that has, power plants on-site.

There are pollution, local pollution effects.

So all of those issues can be a concern if they're not properly handled.

KATKO: So, just so I'm clear, when you walk inside a data center, is it teeming with people, or is it just loaded with machines that tend to get warm and need to be cooled down?

KOOMEY: Yes, you can think of a data center as a big warehouse building filled with computers.

And we need to take the heat out of the computers in order to keep them functioning well.

So that means that there's cooling energy associated with, you know, keeping the data center operating.

There's electricity going into the computers to keep them chugging along and generating the results of those queries that we talked about.

KATKO: So let's try and take your Koomey's law to the next logical question, and that is, if computers keep advancing at such a rapid pace, are we going to need these massive data centers going forward?

Or can we downsize as we get more advanced with our computing?

KOOMEY: There are two things that drive electricity use of data centers.

One thing is the increase in the service demand, the compute demand, the things that, you know, the number of queries that we ask of the computers.

The other thing is efficiency.

If service demand goes up, electricity use goes up.

If efficiency goes up, electricity use goes down.

And the real question is whether service demand is going to keep going up as fast as it has, has in the last, few years.

Basically, the AI companies have been giving away compute at well below cost.

So I think there's real questions about whether they're going to, you know, the service demand is going to keep increasing at recent rates because the AI companies are going to start charging the actual costs of data equipment and data facilities that they've been developing.

Because up till now, it's basically been a lot of free stuff, and people love free stuff, and they use more of it, but then when people start charging the real price of it, I expect that service demand to not increase as fast.

KATKO: With a few minutes we have left, I want to talk about the arms race aspect of this.

I mean, when I was a child, the arms race was a nuclear arms race during the Cold War, and it was terrifying, but it was also the conventional wisdom was that if we allow our adversaries, in particular Russia, to keep building nuclear weapons, and we don't keep up with them, there's no mutually assured destruction there.

So now we have this modern technology that there's been much discussion about.

If we don't keep up with China and the others, is that a bad thing and why?

KOOMEY: Let's first talk about the arms race between the companies, because that's really what's been driving growth in AI demand.

The companies all believe that they're going to get increasing benefits by scaling their models, bigger models, more data, more computers thrown at the problem.

And they all think that they want to get there first.

So they want to be first past the post, and if they do, they're going to win.

That's their whole vision of this.

And there are a lot of questions about whether that's even true, that they're, you know, it's pretty likely that there are other ways to generate those compute services that don't involve scaling, that don't involve the biggest models and the most data.

And so there are, there are ways to deliver those computing services that use 100 times less energy and 100 times less cost.

So, that's the arms race among the companies.

And then among the countries, I think it's really unclear what that question even means.

Because what does winning mean in this context?

KATKO: Well, I guess, not so much winning, is what harm could countries like China do to us, if they, you know, they have a turbocharged AI system fueled by his massive data and we don't?

KOOMEY: So, we need to be focused on what these models can do, and then what potential harms they could instigate.

And it is true that these models are particularly good at, for example, hacking.

They're really good at software.

And so we need to, at a minimum, be prepared to counter other countries using AI to hack our systems.

So we need to become more sophisticated at it.

But right now, the US is leading the charge by a lot.

KATKO: Dr.

Jonathan Koomey, we could go on forever with this, but thank you so much for joining us.

♪ So, are data centers a boon or a burden for the people who live near them?

Here to discuss this are Elena Schlossberg, of the Coalition to Protect Prince William County, Virginia, a primer expansion zone for data centers, and Dan Diorio, of the Data Center Coalition.

Welcome to you both, and Dan, I'll start with you real quick.

The data centers, are they boon or a burden for the communities, you think?

DIORIO: Well, I think they are definitely a boon for communities.

We see dramatically throughout the country that data centers are a significant source of jobs, for skilled tradesmen and women.

both to construct the facilities and maintain them, as well as significant sources of revenue in neighboring Loudoun County data centers have helped to lower taxes two years in a row.

you know, so these are tremendous job and revenue opportunities for communities.

I agree they have to be done right.

KATKO: So, Elena, is there a cost to some of these, so-called boons to the communities?

SCHLOSSBERG: So, let me just explain why and how I got engaged in this issue, on Amazon Web Services campus, secretive.

Our supervisors had also signed NDAs, had triggered a massive transmission line project that was not only going to take my property, but other people's property.

And, to add salt to the wound, we, as ratepayers, cover the cost of their infrastructure.

That was 12 years ago.

What is happening with the data center industry is they are simply taking too much, too much power, too much fossil fuel, too much water, and too much land.

And the burden on everyday residents is reaching catastrophic levels.

We see our utility bills doubling.

We see our quality of life diminishing, diesel generators are running, the DEQ can't stop them, Gas turbines, the impacts are felt not only close to home, but also our entire PJM grid is now at risk of blackouts.

And that is our concerns.

It's not about being anti data or anti job, about the actual impacts that we are dealing with on the ground, both near and far.

KATKO: So, Dan, let's talk about that, because the prevailing comments that I hear, both in New York State and nationwide, is that there is a tremendous burden to the communities, and that sometimes the communities pay a disproportionate share of the costs of the electricity, and other things associated with these data centers.

DIORIO: I think nationwide, when we look, we see actually very little relationship between rising data center, buildout, and increasing electricity costs.

You know, we know in PJM, that data centers were responsible for about 50% of the capacity costs.

But those costs have to be allocated appropriately.

We depend on regulators and utilities and grid operators to help us do that.

Data centers aren't an end user.

We expect our bill -- is our bill in full and that no other ratepayers are paying for that cost?

Data centers are the most efficient water users in the economy.

We see 83% of data centers in Virginia, using as much, if not less, water than a large commercial office building.

So I agree that we ultimately have to do, we have to address these questions.

And I think overwhelmingly, the good actors in the industry are addressing these questions.

KATKO: So, Elena, let's follow up with that.

Your concerns are about the cost, the rising costs, and burns on communities.

How would you suggest that we address those?

SCHLOSSBERG: So, the coalition back in 2014, when we were found, and knew we needed to prove it was Amazon that triggered a public utility infrastructure project, and that they were taking our property via eminent domain by public utility, and that we were covering the cost.

And what we did was we found a unique way to apply a line extension policy to the data center that had triggered the load, which was Amazon.

And unfortunately, while we didn't prevail then, right now, the SCC has adopted a new tariff and rate structure for large load energy users.

The idea that data centers are not the cause of the stress on our utility bills and the grid is simply not worn out by the facts.

PJM acknowledges that it is the data center industry that is creating a reliability risk, not just in Virginia, but recently in July.

There was an overload on specific transmission lines, and for the first time, that surge, when the data centers removed themselves from the grid, because they are overheating transmission line infrastructure, there was a flickering of power from Northern Virginia to Chicago.

So, indeed, the reality is the data center industry is triggering a load demand that is unprecedented.

In the state of Virginia, we need to build 92,000 megawatts, which is 92 gigawatts, which would be basically tantamount to 92 nuclear reactors.

Instead, we are building massive gas plants in communities that don't even have data center development.

Everybody, in some way, is suffering from this unprecedented power demand that cannot be denied by the data center industry, even they acknowledge it.

KATKO: So, Dan, what about that?

What about, some of the comments you hear out there, prevailing comments that the data centers do use an awful lot of power, and we all acknowledge that.

I am sure you do.

The question is, should they be shouldering more of the burden with respect to the cost associated with that?

DIORIO: Well, the power demand is reflective of all of us, using digital infrastructure.

It's reflective of everything that we do every day, whether it be, you know, an interview just like this, or whether it be an online classroom, an online purchase, electronic health care records.

You know, AI is an important growth driver, but it's not the be-all, end-all data centers, and it's less than a quarter of current compute.

and I think very clearly, costs need to be allocated appropriately.

This is an industry that is committed to paying their full cost of service.

And we've been supportive of contracts and tariffs and policies that ensure that that's the case, because no residential ratepayer should bear a portion of the costs that are directly attributable to a data center.

KATKO: So, Elena, what about that?

How do you get data centers to pay their fair share?

What's your proposal?

SCHLOSSBERG: Okay, so let's just be very clear.

The data center industry has triggered an unprecedented, load demand on Virginia.

Our projected forecast was flat, up until 2021.

And then every single year, we have doubled.

So in 2023, in January, our forecast for peak load demand in Virginia was 23 gigawatts.

The next year, Dominion Energy had to forecclose, and that has grown.

How do you make the data center industry pay?

It's very simple.

The transmission infrastructure they trigger must be applied to them and only them.

The transmission like infrastructure must be buried, so you don't harm and take other people's biggest investment.

We hear a lot about the data center industry investment.

What about my investment?

What about the investments of Virginia residents all over?

Apparently, the data center coalition had a big party, and they've now renamed Virginia, Virginia is for servers.

Instead of Virginia, is for lovers.

Well, does that make us the serfs that are supposed to feed this industry by giving up our American dream?

And that's real.

When a transmission line infrastructure comes, and they exercise eminent domain.

You lose out, and that's happening throughout Virginia.

And the data center industry needs to pay for the transmission.

They need to pay for it being buried, and then they need to pay for all the generations that they alone are triggering.

KATKO: Okay, Dan, what do you think about those proposals?

DIORIO: Data centers should pay for their costs.

That's absolutely true.

And I think, ultimately, we know in Virginia, according to the JLARC report, data centers have been paying their full cost of service.

And good policies that ultimately allow for that to be the case moving forward, well, ultimately, I think, you kiow, be beneficial and it's something that the industry has supported.

And so I think we have to ultimately look at the facts in the context there.

Which is that independent studies have shown, data centers are paying their full cost of service, especially in the state of Virginia.

And I agree, you know, we absolutely should be paying for the transmission costs and for generation costs.

And I think the data center industry will continue to remain committed to that.

And we will continue to work with regulators, utilities, and grid operators to ensure that that's the case.

KATKO: Okay, so, Elena, Dan admits whether they're actually doing it or not, he admits that they should be paying their fair shares.

So let's say that you have an agreement on that.

But what about the location of these?

We have about 45 seconds left.

You know, I drove through Northern Virginia many times the last few years, and having data centers in high, high density suburban areas doesn't seem like a good idea.

So what would you do about the location of them?

What would you propose there?

We have about 45 seconds.

SCHLOSSBERG: The only way to resolve this issue is data centers must take less power.

They must take less energy.

Right now, they're operating on gas turbines and communities, like Vantage and Sterling, where people are listening to what sounds like jet engines 24/7.

And those gas turbines are breaking down, and then the diesel generators are running 24/7, which is what happened on a purple day.

You have community suffering now.

The only way to resolve this so that you don't have gas, gas plants being built by communities who don't benefit from data centers.

We don't have communities living next to the diesel generators or the gas turbines, is they must demand less power.

And until that happens, there needs to be a pause, and there needs to be not one more hookup for data centers until you can protect communities, first and foremost.

KATKO: Dan Diorio and Elena Schlossberg.

Thanks so much for a great conversation.

We appreciate it.

It's a tough issue.

♪ Whether we like it or not, data centers are here to stay.

They are crucial to both power AI and advance our economy, but also to keep pace with adversaries such as China, who could and would use AI to do us harm.

It is clearly a technological arms race, and we cannot afford to finish second.

But we cannot ignore the legitimate community concerns that data centers generate.

Thus, like any major achievement in a storied success of our American history, we do it best when all sides work together.

mindful that all must bend a bit to get things done for the greater good.

And that's my take.

of the world's first general-purpose electronic computers filled an entire room?

ENIAC, as it was called, took up about 1,500 square feet.

It weighed 30 tons.

It used some 18,000 vacuum tubes, and consumed about 150 kilowatts of electricity.

And it could perform about 5,000 additions per second.

Today, the specialized AI processor in a smartphone like this one can perform trillions of operations per second.

All inside a device that fits in your hand.

So in 80 years, computing went from a 30-ton machine that filled a room to something you can accidentally leave in an Uber.

Now, that's downsizing.

And now you do know.

That's all for this week, folks.

To send in your comments for the show, or to see "Balancing Act," extras, and exclusives, follow us on social media, or go to BalancingActWithJohnKatko.com.

Thank you for joining us, and remember, in a circus that is politics, there is always a balancing act.

I'm John Katko.

We'll see you next week, America.

NARRATOR: With Pomeroy Markers, history is based on the facts Vetted by professional historians.

Over 3,000 installed nationwide.

From Maine to Florida, to California, to Minnesota, to Texas and Alaska.

Since 2006, our markers have helped people celebrate community history.

Marker grant program information available at wgpfoundation.org.