WEBVTT - AI Data Centers Use a Lot of Energy. You May Be Paying for It

0:00:00.240 --> 0:00:07.200
<v Speaker 1>Bloomberg Audio Studios, Podcasts, radio News.

0:00:07.960 --> 0:00:09.560
<v Speaker 2>Hey, mister Stanley, how you doing great?

0:00:09.600 --> 0:00:10.160
<v Speaker 3>How are you doing?

0:00:10.280 --> 0:00:11.319
<v Speaker 4>I'm good, So I'm Josh.

0:00:11.360 --> 0:00:12.800
<v Speaker 2>We talked on the phone all right.

0:00:13.320 --> 0:00:16.480
<v Speaker 5>Earlier this month, Bloomberg reporter Josh Sall visited a man

0:00:16.560 --> 0:00:20.520
<v Speaker 5>named Kevin Stanley. He's fifty seven and he lives in Baltimore.

0:00:20.720 --> 0:00:22.280
<v Speaker 3>Do you'll need me to turn the lights on more

0:00:22.320 --> 0:00:24.439
<v Speaker 3>because this sounding here can get kind of dark here.

0:00:24.640 --> 0:00:26.960
<v Speaker 5>Josh was in Maryland to report on the effects that

0:00:27.080 --> 0:00:30.880
<v Speaker 5>data centers have on people who live near them. Kevin

0:00:30.920 --> 0:00:33.720
<v Speaker 5>lives about a two hour drive from an area known

0:00:33.760 --> 0:00:37.800
<v Speaker 5>as Data Center Alley, the world's largest concentration of data centers,

0:00:38.000 --> 0:00:41.920
<v Speaker 5>topping other hotspots like Iowa and Oregon, and Josh was

0:00:41.960 --> 0:00:44.400
<v Speaker 5>there to talk to him about one thing, in particular,

0:00:44.720 --> 0:00:48.200
<v Speaker 5>his electricity bill, which has gone up ever since data

0:00:48.200 --> 0:00:49.360
<v Speaker 5>centers came to the region.

0:00:49.960 --> 0:00:52.639
<v Speaker 4>Kevin's a really nice guy, really interesting guy, and he

0:00:52.680 --> 0:00:55.840
<v Speaker 4>talked about how his high power bills have had a

0:00:55.880 --> 0:00:57.960
<v Speaker 4>really rough effect on his life. He has had to

0:00:57.960 --> 0:01:01.000
<v Speaker 4>cut back on buying the groceries he likes, getting haircuts

0:01:01.040 --> 0:01:02.640
<v Speaker 4>as often as he likes. He tries to make his

0:01:02.680 --> 0:01:04.320
<v Speaker 4>diabetes medication stretch out.

0:01:04.360 --> 0:01:07.039
<v Speaker 3>The power bills just keep going up and up, and

0:01:07.319 --> 0:01:12.040
<v Speaker 3>for me, I'm a single person in here, so I'm like, wow, Wow,

0:01:12.240 --> 0:01:13.680
<v Speaker 3>the bill's gone up so much.

0:01:13.959 --> 0:01:17.640
<v Speaker 5>Kevin is blind and he lives on disability payments. He

0:01:17.680 --> 0:01:20.640
<v Speaker 5>says his energy bills are now eighty percent higher than

0:01:20.680 --> 0:01:23.200
<v Speaker 5>they were just about three years ago. And there were

0:01:23.280 --> 0:01:26.360
<v Speaker 5>days when the utility asks customers like him to use

0:01:26.440 --> 0:01:30.240
<v Speaker 5>less power to prevent blackouts or encourages them to use

0:01:30.319 --> 0:01:31.520
<v Speaker 5>less power to save money.

0:01:31.640 --> 0:01:34.800
<v Speaker 3>They have days where they tell us, don't use any electricity,

0:01:34.920 --> 0:01:37.600
<v Speaker 3>don't run air and dig but it's like ninety five degrees,

0:01:38.600 --> 0:01:39.520
<v Speaker 3>I'll dine here.

0:01:40.840 --> 0:01:43.320
<v Speaker 5>It can be difficult to pinpoint the exact root of

0:01:43.360 --> 0:01:46.480
<v Speaker 5>higher electricity bills and strain on the electrical grid, so

0:01:46.640 --> 0:01:49.040
<v Speaker 5>Josh set out to put some numbers to the growing

0:01:49.080 --> 0:01:52.280
<v Speaker 5>pressure data centers are putting on local power supply, and

0:01:52.360 --> 0:01:54.800
<v Speaker 5>he found that at a time of rapid data center

0:01:54.840 --> 0:01:58.640
<v Speaker 5>construction and investment in AI technology that needs them, demand

0:01:58.840 --> 0:02:02.520
<v Speaker 5>is driving up wholes energy costs and those costs are

0:02:02.560 --> 0:02:04.200
<v Speaker 5>being passed on to consumers.

0:02:04.840 --> 0:02:08.400
<v Speaker 2>Josh spoke to some people that you know, started telling

0:02:08.440 --> 0:02:10.000
<v Speaker 2>him that their bills were really high.

0:02:10.320 --> 0:02:14.160
<v Speaker 5>That's Leonardo Nicoletti, a data visualization reporter at Bloomberg who

0:02:14.240 --> 0:02:16.640
<v Speaker 5>traveled with Josh and crunched the numbers.

0:02:16.600 --> 0:02:19.240
<v Speaker 2>When we were thinking about it. They seem to be

0:02:19.320 --> 0:02:21.800
<v Speaker 2>living close to the hotspots of you know, AI data

0:02:21.800 --> 0:02:25.160
<v Speaker 2>centers in the United States. The main finding was that,

0:02:25.440 --> 0:02:29.560
<v Speaker 2>quite strikingly, if you are in an area that is

0:02:29.639 --> 0:02:33.480
<v Speaker 2>located close to data centers or data center activity, you're

0:02:33.639 --> 0:02:37.519
<v Speaker 2>much more likely to experience high price increases.

0:02:38.120 --> 0:02:42.280
<v Speaker 4>The massive increase in the demand for AI and the

0:02:42.360 --> 0:02:44.800
<v Speaker 4>speed with which the data centers get bigger and bigger.

0:02:44.919 --> 0:02:47.000
<v Speaker 4>I think it's been a huge surprise to all of us.

0:02:51.280 --> 0:02:53.200
<v Speaker 5>I'm David Gerrett, and this is the big take from

0:02:53.200 --> 0:02:56.919
<v Speaker 5>Bloomberg News today. On the show, the staggering electricity needs

0:02:56.919 --> 0:03:01.000
<v Speaker 5>of AI data centers and how you could end up

0:03:01.000 --> 0:03:08.440
<v Speaker 5>putting the bill. AI needs a lot of energy to

0:03:08.560 --> 0:03:12.920
<v Speaker 5>work from summarized answers on Google to chat, GPT, grocery lists.

0:03:13.200 --> 0:03:17.120
<v Speaker 5>The rapidly growing technology requires massive amounts of computing power,

0:03:17.360 --> 0:03:20.080
<v Speaker 5>which requires a lot of actual power.

0:03:21.160 --> 0:03:24.080
<v Speaker 2>We actually quantified this in an earlier story that Josh

0:03:24.080 --> 0:03:26.919
<v Speaker 2>and I worked together. We compared the total amount of

0:03:26.960 --> 0:03:31.520
<v Speaker 2>electricity that data centers use yearly to how much electricity

0:03:31.720 --> 0:03:36.200
<v Speaker 2>individual countries use. And we've found that, for example, countries

0:03:36.240 --> 0:03:40.720
<v Speaker 2>like Italy or Australia are using actually less electricity as

0:03:40.760 --> 0:03:44.440
<v Speaker 2>a total than data centers use globally. And these numbers

0:03:44.480 --> 0:03:47.800
<v Speaker 2>are actually higher now because we did this one year ago,

0:03:47.840 --> 0:03:52.600
<v Speaker 2>and electricity demand is growing exponentially from data centers.

0:03:52.640 --> 0:03:55.040
<v Speaker 4>That numbers supposed to keep going up and up. So

0:03:55.280 --> 0:03:58.080
<v Speaker 4>Bloomberg NIF projects that it's supposed to be over four

0:03:58.120 --> 0:04:02.600
<v Speaker 4>percent by twenty thirty five total global electricity consumption. So

0:04:02.640 --> 0:04:04.560
<v Speaker 4>if you took all the data centers at that point

0:04:04.600 --> 0:04:06.880
<v Speaker 4>and made them their own country, at that point, they'd

0:04:06.880 --> 0:04:11.280
<v Speaker 4>be the fourth biggest consumer of electricity after China, the US,

0:04:11.280 --> 0:04:11.760
<v Speaker 4>and India.

0:04:13.280 --> 0:04:15.680
<v Speaker 5>But what Josh and Layo wanted to understand was whether

0:04:15.720 --> 0:04:19.760
<v Speaker 5>the huge demand for electricity was showing up on customers' bills.

0:04:20.120 --> 0:04:22.719
<v Speaker 2>It took us like three months to find the data

0:04:22.760 --> 0:04:26.440
<v Speaker 2>because you know, it's actually really hard to get granular

0:04:26.760 --> 0:04:28.720
<v Speaker 2>data on power prices.

0:04:29.240 --> 0:04:32.320
<v Speaker 5>To get that more granular picture, Layo and another Bloomberg

0:04:32.400 --> 0:04:36.159
<v Speaker 5>data visualization reporter named Demetrius Pogcas found a company called

0:04:36.200 --> 0:04:39.200
<v Speaker 5>grid Status that collects real time data for more than

0:04:39.240 --> 0:04:41.679
<v Speaker 5>twenty five thousand nodes around the US.

0:04:42.240 --> 0:04:45.960
<v Speaker 2>A node is like the location on the grid, so

0:04:46.000 --> 0:04:50.720
<v Speaker 2>that's connected to transmission lines and it measures real time congestion,

0:04:50.960 --> 0:04:55.880
<v Speaker 2>real time fuel prices, and real time like supply costs,

0:04:56.000 --> 0:04:59.200
<v Speaker 2>things like that, right, and then the different factors that

0:04:59.279 --> 0:05:01.560
<v Speaker 2>are measured by that node then make up the price.

0:05:02.400 --> 0:05:04.880
<v Speaker 5>With that data as well as data on the locations

0:05:04.880 --> 0:05:08.040
<v Speaker 5>and capacity of data centers from a company called dc Byte,

0:05:08.320 --> 0:05:11.800
<v Speaker 5>the reporters measured the distance between those nodes and areas

0:05:11.800 --> 0:05:15.479
<v Speaker 5>where there's significant data center activity. They use that to

0:05:15.520 --> 0:05:18.080
<v Speaker 5>create a data set that showed the price changes recorded

0:05:18.080 --> 0:05:20.720
<v Speaker 5>in those areas over time since twenty twenty.

0:05:21.279 --> 0:05:25.800
<v Speaker 2>Essentially, the main finding is that seventy percent of nodes

0:05:25.800 --> 0:05:29.800
<v Speaker 2>that recorded price increases are located within fifty miles of

0:05:30.240 --> 0:05:31.880
<v Speaker 2>significant data center activity.

0:05:32.080 --> 0:05:33.920
<v Speaker 4>And if you look at your power bill, you're going

0:05:33.960 --> 0:05:36.760
<v Speaker 4>to see in most places a distribution charge. That's what

0:05:36.800 --> 0:05:39.160
<v Speaker 4>you're paying for all of the infrastructure, all the wires.

0:05:39.360 --> 0:05:41.240
<v Speaker 4>But you're also going to see a supply charge. And

0:05:41.279 --> 0:05:43.800
<v Speaker 4>the supply charge is what you're paying for that wholesale power.

0:05:43.839 --> 0:05:46.159
<v Speaker 4>So what Lao's research says is by pushing up the

0:05:46.160 --> 0:05:49.360
<v Speaker 4>price of wholesale power. Data centers are putting that upward

0:05:49.400 --> 0:05:51.120
<v Speaker 4>pressure on your power bill.

0:05:51.160 --> 0:05:53.000
<v Speaker 5>So effectively, if I open up that bill, I'm not

0:05:53.040 --> 0:05:56.000
<v Speaker 5>going to see AI data centers as a line item,

0:05:56.400 --> 0:05:59.560
<v Speaker 5>but it's having that effect on what you're describing exactly.

0:06:00.680 --> 0:06:03.599
<v Speaker 5>Josh says there are two key mechanisms that drive up

0:06:03.680 --> 0:06:06.240
<v Speaker 5>electricity costs around data center hotspots.

0:06:06.520 --> 0:06:10.200
<v Speaker 4>The first, to really summarize supply and demand. They use

0:06:10.240 --> 0:06:13.080
<v Speaker 4>a lot of electricity, which can make electricity more expensive

0:06:13.520 --> 0:06:15.600
<v Speaker 4>for everybody, So that's one way that can push up

0:06:15.600 --> 0:06:17.600
<v Speaker 4>customer bills. The other ways that they require a lot

0:06:17.640 --> 0:06:20.640
<v Speaker 4>of new infrastructure, mostly transmission lines, but also new power

0:06:20.720 --> 0:06:23.440
<v Speaker 4>plants and the costs of building that are spread out

0:06:23.440 --> 0:06:26.440
<v Speaker 4>among all customers, including regular people like you.

0:06:27.040 --> 0:06:30.120
<v Speaker 5>Monitoring Analytics is an outside watchdog that keeps an eye

0:06:30.160 --> 0:06:33.600
<v Speaker 5>on the largest US electric grid, PJM, and it says

0:06:33.600 --> 0:06:37.320
<v Speaker 5>that data center development raise costs for customers on pjm's grid,

0:06:37.480 --> 0:06:40.320
<v Speaker 5>which spans from Illinois to Washington, DC, by more than

0:06:40.400 --> 0:06:43.640
<v Speaker 5>nine point three billion dollars over a twelve month period

0:06:43.760 --> 0:06:47.000
<v Speaker 5>starting in June. Today, thirty three percent of all the

0:06:47.040 --> 0:06:50.920
<v Speaker 5>electricity used in Oregon is attributed to data centers. In Virginia,

0:06:51.200 --> 0:06:54.479
<v Speaker 5>it's thirty seven percent. That's according to data from dc

0:06:54.600 --> 0:06:58.480
<v Speaker 5>BYTE and the US Energy Information Agency. Communities in surrounding

0:06:58.520 --> 0:07:01.240
<v Speaker 5>areas are starting to do the map, and they're also

0:07:01.440 --> 0:07:03.599
<v Speaker 5>connecting higher bills to data centers.

0:07:03.880 --> 0:07:06.159
<v Speaker 6>You don't have to see those data centers to see

0:07:06.440 --> 0:07:07.760
<v Speaker 6>the higher electric bills.

0:07:08.000 --> 0:07:11.200
<v Speaker 5>That's David Lapp, a consumer advocate from the Maryland Office

0:07:11.200 --> 0:07:12.080
<v Speaker 5>of People's Council.

0:07:12.520 --> 0:07:15.200
<v Speaker 6>We don't have in Maryland too many data centers, so

0:07:15.280 --> 0:07:17.920
<v Speaker 6>the costs are being driven up by data centers that

0:07:17.960 --> 0:07:20.600
<v Speaker 6>are out of state, largely in many of them in

0:07:20.640 --> 0:07:21.560
<v Speaker 6>northern Virginia.

0:07:22.280 --> 0:07:25.480
<v Speaker 5>In February, the Baltimore City Council led a hearing on

0:07:25.520 --> 0:07:28.280
<v Speaker 5>the issue of rising Baltimore gas and electric bills.

0:07:30.040 --> 0:07:34.880
<v Speaker 4>The current state of these BGE bills and the skyrocketing

0:07:35.120 --> 0:07:38.560
<v Speaker 4>rates is simply not sustainable for our constituents.

0:07:38.960 --> 0:07:40.679
<v Speaker 5>And one of the people who spoke at the hearing

0:07:41.000 --> 0:07:45.360
<v Speaker 5>was Kevin Stanley, the man Josh and Lao interviewed in Baltimore.

0:07:45.840 --> 0:07:49.000
<v Speaker 3>Yes, good evening, everyone, My name is Kevin Stanley.

0:07:49.440 --> 0:07:51.760
<v Speaker 5>Kevin didn't know what was causing his bills to spike.

0:07:52.040 --> 0:07:53.760
<v Speaker 5>He just knew it was a problem.

0:07:54.000 --> 0:07:56.760
<v Speaker 3>And there's no reason that Baltimore is shold have the

0:07:56.840 --> 0:07:59.920
<v Speaker 3>excess of gas electric Race said that.

0:08:00.760 --> 0:08:04.720
<v Speaker 2>When Josh asked Kevin Stanley what he thought about data centers,

0:08:05.320 --> 0:08:06.320
<v Speaker 2>his reply was.

0:08:06.480 --> 0:08:08.080
<v Speaker 3>So they could say, oh, this is going to help

0:08:08.120 --> 0:08:10.960
<v Speaker 3>with AI. But how's that going to help me? How's

0:08:11.000 --> 0:08:12.360
<v Speaker 3>that going to help me pay my bill?

0:08:15.280 --> 0:08:18.040
<v Speaker 5>So how do utilities decide who fronts the cost? And

0:08:18.040 --> 0:08:28.640
<v Speaker 5>how are those communities responding? That's after the break. A

0:08:28.720 --> 0:08:31.640
<v Speaker 5>team of Bloomberg reporters dug into the rising costs of

0:08:31.640 --> 0:08:34.760
<v Speaker 5>electricity to see if the rapid construction of data centers

0:08:34.800 --> 0:08:37.520
<v Speaker 5>in some areas could be to blame, and they found

0:08:37.520 --> 0:08:41.160
<v Speaker 5>a link. Places with heightened data center activity were more

0:08:41.280 --> 0:08:45.240
<v Speaker 5>likely to see higher wholesale energy costs, meaning the presence

0:08:45.280 --> 0:08:48.960
<v Speaker 5>of data centers was raising power prices and eventually pushing

0:08:49.000 --> 0:08:52.720
<v Speaker 5>up customer bills. I asked Bloomberg reporter Josh Saul about

0:08:52.760 --> 0:08:55.760
<v Speaker 5>why that's happening when people usually expect to pay for

0:08:55.840 --> 0:08:59.679
<v Speaker 5>their own utility costs. I think about how I study

0:08:59.720 --> 0:09:02.800
<v Speaker 5>my electricity build to the extent that I do, And

0:09:02.880 --> 0:09:05.320
<v Speaker 5>maybe in the summer I'm using my air conditioner more.

0:09:06.000 --> 0:09:07.760
<v Speaker 5>Maybe in the winter I'm using my heat more. But

0:09:07.840 --> 0:09:10.240
<v Speaker 5>these are all kind of selfish things. I'm determining when

0:09:10.240 --> 0:09:12.960
<v Speaker 5>the bill goes up or down, why should I be

0:09:13.000 --> 0:09:17.040
<v Speaker 5>paying for the expenses of associated with these data centers.

0:09:17.320 --> 0:09:19.080
<v Speaker 4>I think there's two arguments that people make. I mean,

0:09:19.080 --> 0:09:21.800
<v Speaker 4>the first is that AI, whether you like it or not,

0:09:22.040 --> 0:09:24.800
<v Speaker 4>is becoming an increasingly large part of our world. So

0:09:25.000 --> 0:09:28.440
<v Speaker 4>from national security to the argument that we don't want

0:09:28.440 --> 0:09:30.840
<v Speaker 4>other countries to be way better at AI than we are,

0:09:31.200 --> 0:09:33.760
<v Speaker 4>to the fact that now our Google searches and even

0:09:33.800 --> 0:09:38.000
<v Speaker 4>our texts to our friends are guided by or instructed

0:09:38.000 --> 0:09:40.440
<v Speaker 4>by AI. So there's one argument there that we're all

0:09:40.559 --> 0:09:43.319
<v Speaker 4>using AI. And there's also an old utility law. The

0:09:43.360 --> 0:09:45.560
<v Speaker 4>way we do this is that the costs are shared

0:09:45.600 --> 0:09:48.600
<v Speaker 4>among everyone. So if you build a new house or

0:09:48.640 --> 0:09:50.600
<v Speaker 4>hook up to the grid as a new business, you

0:09:50.640 --> 0:09:53.800
<v Speaker 4>don't pay for all of those costs to hook you up.

0:09:53.840 --> 0:09:57.800
<v Speaker 4>It's socialized among everyone on the grid. So utilities are

0:09:58.160 --> 0:10:01.960
<v Speaker 4>following that general principle. Now now with these large new

0:10:02.000 --> 0:10:03.679
<v Speaker 4>customers data centers.

0:10:04.360 --> 0:10:07.040
<v Speaker 5>It's tech companies that are deciding when and where to

0:10:07.080 --> 0:10:10.880
<v Speaker 5>build a new data center. Amazon, Microsoft, and Alphabet's Google

0:10:11.040 --> 0:10:14.920
<v Speaker 5>are the three biggest US cloud providers in twenty twenty four,

0:10:15.040 --> 0:10:18.080
<v Speaker 5>those companies spent more than two hundred billion dollars on

0:10:18.160 --> 0:10:21.600
<v Speaker 5>capital expenditures, most of it to build new data centers.

0:10:22.040 --> 0:10:25.720
<v Speaker 5>What's the relationship like between these tech companies that need

0:10:25.760 --> 0:10:28.480
<v Speaker 5>all of this power and the utilities themselves. Are they

0:10:28.520 --> 0:10:31.160
<v Speaker 5>paying their fair share? Are they paying a disproportionate amount?

0:10:31.440 --> 0:10:32.320
<v Speaker 5>How's that shaking out?

0:10:32.440 --> 0:10:32.560
<v Speaker 2>Well?

0:10:32.640 --> 0:10:35.240
<v Speaker 4>Utilities are really excited because it's a huge new customer

0:10:35.280 --> 0:10:37.400
<v Speaker 4>base for them, and this is coming after decades of

0:10:37.480 --> 0:10:41.160
<v Speaker 4>very flat growth. At the same time, there's some conflicts.

0:10:41.360 --> 0:10:44.319
<v Speaker 4>Tech is known for moving fast and breaking things, that's

0:10:44.360 --> 0:10:47.160
<v Speaker 4>their philosophy. Utilities are known for moving very slow and

0:10:47.200 --> 0:10:49.520
<v Speaker 4>making sure nothing ever breaks, you know, making sure the

0:10:49.600 --> 0:10:52.280
<v Speaker 4>lights stay on. So they're working together here. But one

0:10:52.280 --> 0:10:55.560
<v Speaker 4>thing that we're seeing is that utilities are actively figuring

0:10:55.640 --> 0:10:58.520
<v Speaker 4>out how to make sure that big tech is paying

0:10:58.559 --> 0:11:01.320
<v Speaker 4>its share for these data centers. They're doing things like

0:11:01.360 --> 0:11:05.360
<v Speaker 4>creating new customer classes for data centers to pay differently

0:11:05.440 --> 0:11:09.520
<v Speaker 4>than say, residential or small businesses. They're creating requirements like

0:11:09.600 --> 0:11:13.120
<v Speaker 4>data centers have to pay for certain costs, pay for

0:11:13.240 --> 0:11:15.480
<v Speaker 4>a minimum amount of power over fifteen years, even if

0:11:15.480 --> 0:11:18.720
<v Speaker 4>they use less putting, non collateral different things like that

0:11:18.920 --> 0:11:21.280
<v Speaker 4>to protect their existing customers.

0:11:21.800 --> 0:11:24.760
<v Speaker 5>Josh Leo and their team reached out to Amazon, Microsoft,

0:11:24.760 --> 0:11:28.520
<v Speaker 5>and Alphabet for comment. An Amazon spokesperson said the company

0:11:28.600 --> 0:11:31.720
<v Speaker 5>works closely with utilities and grid operators to plan for

0:11:31.760 --> 0:11:35.880
<v Speaker 5>future growth. Microsoft's vice president of energy told them, quote,

0:11:35.960 --> 0:11:38.640
<v Speaker 5>it's literally our responsibility to make sure that when we

0:11:38.679 --> 0:11:41.360
<v Speaker 5>come to a community, when we get connected to a grid,

0:11:41.640 --> 0:11:44.439
<v Speaker 5>that the cost of the infrastructure that's being dedicated to us,

0:11:44.600 --> 0:11:48.280
<v Speaker 5>that those costs of service get allocated to us. Google

0:11:48.320 --> 0:11:51.280
<v Speaker 5>said it's been working to use less electricity even as

0:11:51.280 --> 0:11:54.199
<v Speaker 5>it expands data centers, and that it supports paying its

0:11:54.200 --> 0:11:58.080
<v Speaker 5>fair share. What did you hear from utilities about this phenomenon?

0:11:58.160 --> 0:12:00.360
<v Speaker 5>What do they say about why builds your going up?

0:12:00.400 --> 0:12:01.640
<v Speaker 5>And what they intend to do about it?

0:12:01.880 --> 0:12:05.240
<v Speaker 4>Utilities mostly point to the idea that data centers are

0:12:05.280 --> 0:12:08.119
<v Speaker 4>paying their fair share. So if a data center requires

0:12:08.200 --> 0:12:11.839
<v Speaker 4>a specific substation for its operations, the tech company will

0:12:11.840 --> 0:12:14.160
<v Speaker 4>pay for that, and they'll point to the reforms that

0:12:14.200 --> 0:12:17.680
<v Speaker 4>they're making in their own rate structure to charge these

0:12:17.800 --> 0:12:20.800
<v Speaker 4>large load customers differently. All of that's true, and all

0:12:20.920 --> 0:12:23.520
<v Speaker 4>that's good, but it doesn't change the fact that a

0:12:23.559 --> 0:12:26.880
<v Speaker 4>lot of costs are still being loaded onto regular customers.

0:12:27.280 --> 0:12:30.680
<v Speaker 5>Dominion Energy, the utility company that serves Northern Virginia's data center,

0:12:30.720 --> 0:12:34.120
<v Speaker 5>Alley told the reporters quote, we believe data centers should

0:12:34.160 --> 0:12:36.480
<v Speaker 5>pay for the full cost of their power. That's how

0:12:36.559 --> 0:12:38.960
<v Speaker 5>we design our rates and it's the standard our regulator

0:12:39.080 --> 0:12:42.240
<v Speaker 5>uses in reviewing them. The CEO of Exelon, which runs

0:12:42.320 --> 0:12:45.520
<v Speaker 5>Baltimore's utility, said the company is pushing for long term

0:12:45.559 --> 0:12:48.479
<v Speaker 5>solutions that are fair and bring peace of mind to customers.

0:12:49.040 --> 0:12:51.640
<v Speaker 4>Talking to one of the utilities, we asked about how

0:12:51.640 --> 0:12:55.480
<v Speaker 4>they pass these costs onto customers, and their experts said,

0:12:55.600 --> 0:12:57.880
<v Speaker 4>we sign long term contracts, so if power spikes up,

0:12:57.920 --> 0:13:00.760
<v Speaker 4>that's for the power sellers to deal with. Customer doesn't

0:13:00.800 --> 0:13:03.800
<v Speaker 4>have to deal with that. However, the next time we

0:13:03.840 --> 0:13:06.239
<v Speaker 4>sign a contract, if there's been a lot of variability

0:13:06.280 --> 0:13:08.880
<v Speaker 4>in cost in power prices, and if the power prices

0:13:08.920 --> 0:13:12.640
<v Speaker 4>have been overall just higher, which happens because of data centers,

0:13:12.760 --> 0:13:15.480
<v Speaker 4>then the next contract that we sign will be for

0:13:15.559 --> 0:13:18.200
<v Speaker 4>higher prices, and then yes, that will get passed on

0:13:18.280 --> 0:13:19.520
<v Speaker 4>directly to our customers.

0:13:19.800 --> 0:13:22.720
<v Speaker 5>Now, some communities are starting to have conversations about what

0:13:22.760 --> 0:13:25.520
<v Speaker 5>it really means to have these big AI data centers

0:13:25.559 --> 0:13:28.640
<v Speaker 5>in their backyard. Are we seeing communities, Are we seeing

0:13:28.679 --> 0:13:30.520
<v Speaker 5>states push back on this in any way?

0:13:31.000 --> 0:13:34.800
<v Speaker 4>Yeah, I'd say there's more dramatic conversation around the power

0:13:34.840 --> 0:13:38.560
<v Speaker 4>grid than anytime in my years covering it. There was

0:13:38.720 --> 0:13:43.200
<v Speaker 4>just a summit in Philadelphia where the governor Pennsylvania, Joshapiro,

0:13:43.640 --> 0:13:46.559
<v Speaker 4>talked about affordability. We talked about these rising bills.

0:13:46.800 --> 0:13:50.480
<v Speaker 1>As governor of this great comme, privilege to represent thirteen

0:13:50.520 --> 0:13:52.880
<v Speaker 1>million pens of land, and I can tell you they

0:13:52.960 --> 0:13:58.360
<v Speaker 1>can't afford never ending price increases because of pgm's policies.

0:13:58.960 --> 0:14:02.240
<v Speaker 4>That's something we're seeing from politicians all around. Sometimes there's

0:14:02.280 --> 0:14:05.040
<v Speaker 4>a little bit of a personality split, where you know,

0:14:05.080 --> 0:14:09.320
<v Speaker 4>local government's excited for the tax income. They want companies

0:14:09.360 --> 0:14:12.040
<v Speaker 4>moving in, they want business activity, they want to fund

0:14:12.040 --> 0:14:13.240
<v Speaker 4>their local school district.

0:14:13.559 --> 0:14:16.520
<v Speaker 5>In the city of Fayetteville, Georgia, for example, the taxes

0:14:16.559 --> 0:14:19.200
<v Speaker 5>paid by a new data center developer, we're expected to

0:14:19.200 --> 0:14:21.960
<v Speaker 5>bring in over one billion dollars in state and local

0:14:22.040 --> 0:14:25.360
<v Speaker 5>taxes the next fifteen years. Taxes going to the county

0:14:25.360 --> 0:14:28.280
<v Speaker 5>Board of Education last year covered the equivalent of some

0:14:28.480 --> 0:14:30.160
<v Speaker 5>half a dozen teachers salaries.

0:14:30.640 --> 0:14:34.240
<v Speaker 4>Then there's a split between that view and people who

0:14:34.320 --> 0:14:37.400
<v Speaker 4>live around there who don't want big new transmission lines

0:14:37.440 --> 0:14:41.600
<v Speaker 4>going through their yards. They don't want big three Walmart

0:14:41.640 --> 0:14:44.400
<v Speaker 4>stacked on top of each other size data centers. They

0:14:44.400 --> 0:14:46.720
<v Speaker 4>don't want all of that around them. So, yeah, there

0:14:46.760 --> 0:14:48.440
<v Speaker 4>can be a lot of conflict in these communities.

0:14:49.880 --> 0:14:52.880
<v Speaker 5>When you look at projections for the growth of AI

0:14:53.040 --> 0:14:56.480
<v Speaker 5>data centers, what do they illustrate and what is that

0:14:56.600 --> 0:14:59.200
<v Speaker 5>likely to mean for the cost of electricity in this country.

0:15:00.000 --> 0:15:02.440
<v Speaker 4>Think it's something that needs to get figured out. Utilities

0:15:02.480 --> 0:15:04.440
<v Speaker 4>are working on it. They can't have their customers and

0:15:04.480 --> 0:15:07.880
<v Speaker 4>their regulators and politicians angry at them. Big tech similarly

0:15:08.160 --> 0:15:11.480
<v Speaker 4>very attuned to what people think of them. That's why

0:15:11.520 --> 0:15:13.640
<v Speaker 4>there's a lot of conversations and a lot of push

0:15:13.720 --> 0:15:16.320
<v Speaker 4>right now to figure out how to charge data centers

0:15:16.360 --> 0:15:19.160
<v Speaker 4>for both the power that they're using and the infrastructure

0:15:19.200 --> 0:15:21.360
<v Speaker 4>that has to be built to supply their data centers.

0:15:22.000 --> 0:15:25.480
<v Speaker 5>In the meantime, Kevin Stanley is doing his best to

0:15:25.520 --> 0:15:27.160
<v Speaker 5>deal with higher electricity bills.

0:15:27.240 --> 0:15:30.160
<v Speaker 3>You know, they talk about upgrading to greed and this

0:15:30.240 --> 0:15:32.800
<v Speaker 3>and that, but we don't see it, especially when you're

0:15:32.800 --> 0:15:38.680
<v Speaker 3>dealing with low income people or working class people. They

0:15:38.800 --> 0:15:41.080
<v Speaker 3>have to feel something. It has to be more tangible.

0:15:41.680 --> 0:15:43.760
<v Speaker 3>You would tell people what is a data center? Say, well,

0:15:43.800 --> 0:15:45.360
<v Speaker 3>what is it? But just surviving.

0:15:52.080 --> 0:15:54.520
<v Speaker 5>This is the Big Take from Bloomberg News. I'm David Gerra.

0:15:54.840 --> 0:15:57.360
<v Speaker 5>To get more from the Big Take and unlimited access

0:15:57.400 --> 0:16:00.840
<v Speaker 5>to all of Bloomberg dot com, subscribe today at Bloomberg

0:16:00.920 --> 0:16:04.840
<v Speaker 5>dot com slash podcast offer. Thanks for listening. We'll be

0:16:04.880 --> 0:16:05.440
<v Speaker 5>back tomorrow.