2026-09-08 - EAB Data Center Subcommittee

00:48How are

10:43It's 5:' I'm going to go ahead and call

10:45to order the September 8th data center

10:49subcommittee meeting of the

10:51environmental advisory board. Welcome

10:53everybody. And with that, I will hand it

10:55over to Bruce for introduction.

10:59Good evening subcommittee members. Uh

11:01today this is our second meeting in

11:03September. We're getting to talk more

11:04about environmental impact especially

11:05that for data centers and other indust

11:07heavy industrial uses. We have two

11:09special speakers for you today. Um the

11:11first presenter uh this is going to be a

11:13little different than what's on your

11:13agenda. I apologize I arranged the

11:15slides slightly different. Uh your first

11:17presenter will be Dr. Noah Kitner. He's

11:19with the University of North Carolina

11:21and the second speaker will Dr. Sarra um

11:23Anu Chalum from the University of North

11:25Carolina as well. So just a little

11:27background about Dr. Kitner. Um he

11:29studies energy systems at multiple

11:31scales from regional to inter uh

11:33international power grids to community

11:35micro grids. Uh his research includes

11:37energy systems planning in different

11:39areas, battery assessment and electric

11:41vehicles and energy storage as well.

11:44He'll be coming up to speak to you today

11:45about energy system sustainability and

11:47human health topics related to data

11:49centers. So Dr. if you don't mind coming

11:51up.

11:57>> Thank you.

11:58>> Yeah. [clears throat]

12:00Okay. Thank you very much uh for having

12:02me and inviting me to speak with you

12:04today about on-site and grid connected

12:06energy for data centers. Um,

12:11as as mentioned, um, I'm a pro, uh,

12:14associate professor in environmental

12:16sciences and engineering at UNCC Chapel

12:18Hill, and I'm also, uh, director of the

12:22energy transition initiative in UNCC's

12:24institute for the environment.

12:26Uh so my research focuses on energy

12:30systems particularly I'm looking at

12:33battery storage systems and other uh uh

12:37technologies that are needed to uh trans

12:40transition the grid toward uh cleaner

12:43and more uh healthier alternatives.

12:47[clears throat]

12:48So uh I've been asked to talk a bit

12:52about on-site generation and different

12:56electricity options for data centers. Uh

12:59so uh today I'll go through uh a little

13:03about the energy system, what's

13:05happening in terms of energy demand and

13:08load growth. Then I'll talk more about

13:12uh grid impacts and why uh data center

13:17operators are more likely to seek

13:19on-site generation than other uh types

13:23of loads. And then we'll think about u

13:26sustainability considerations. So

13:31this is actually a very broad macroscale

13:33view of the whole United States energy

13:36system uh from 2024 from Lawrence

13:39Liverour National Labs uh view of the

13:42energy system. And what you can see is

13:46that on the left side you have all the

13:48different uh primary sources of energy

13:51for the US. In the middle you have uh

13:56electricity generation and it's broken

13:58down. Okay. And it's broken down into

14:03[clears throat]

14:04different sectors going from residential

14:08to uh commercial, industrial and

14:11transportation sector. Uh and

14:15it's all proportionally allocated. So

14:18one thing you might notice is about 2/3

14:22of the energy in the US energy system is

14:26rejected energy in the form of waste

14:28heat. Um and that energy is lost uh

14:33through the system. You also have um

14:38about 32 quadrillion BTUs. So everything

14:42is in uh quads of uh electricity

14:46generation. Part of the challenge uh in

14:49a lot of sectors is that there's a lot

14:51of pressure to increase electricity

14:54generation from different sources and

14:57use that in in different sectors. Uh so

15:00uh some data centers depending on how

15:02they may be classified could be

15:04industrial or uh commercial loads and uh

15:10you can see that electricity feeds a lot

15:12of that uh energy. There's also

15:16natural gas and petroleum used for

15:19heating. So this is a very macroscale

15:21view but I think it's important because

15:23it informs how uh data centers are

15:27thinking about where uh energy will come

15:30from and it shows that there may be some

15:33constraints uh related to where um

15:37electricity is is coming from. So uh

15:40we've seen that advances in artificial

15:43intelligence in particular has led to

15:46massive computational needs and uh

15:50created uh a lot of uh forecasts to show

15:54that uh the US is about to enter a

15:57period of of time where the demand for

16:00energy is rising after the past uh 10

16:05years or so where it hasn't um been

16:07growing as much uh this may

16:13require new planning and and regulatory

16:16frameworks to think about uh aspects of

16:20reliable electricity. So, everyone wants

16:23reliable electricity uh for their homes

16:27and businesses and so uh in order to

16:31maintain a reliable grid. Now, there's

16:33new pressures not just from increasing

16:36demand but also uh climate change and

16:39and weather events and other issues that

16:42are uh affecting the uh frequency of of

16:47power outages and uh on the grid. So um

16:53that's where uh different data centers

16:57are looking toward uh solutions to get

17:00electricity. And that's um where they're

17:04looking at potentially colllocating

17:08generation on site or providing some

17:11grid services like flexibility. It um

17:14I'll explain more about what I mean by

17:16grid flexibility. Uh but those are not

17:20really well understood by research in

17:22terms of how they'll affect uh the

17:26electric grid and rate payers since it's

17:29a

17:30quite a complex uh system and uh there's

17:36not not much experience. Um but there

17:38have been several studies to look at

17:41where electricity consumption at a

17:44national scale is going based on um new

17:48data centers. And under different

17:51forecast scenarios

17:53um there are predictions that it's going

17:56to drive most of the demand growth uh

17:59more than potentially more than 50% of

18:01new growth through the next uh five

18:04years through 2030. And uh that's much

18:09faster more than twice the rate of the

18:11past decade. So this is um very fast

18:15growth and um

18:19grid buildout is much slower than data

18:22center development. So uh there's kind

18:25of a confluence of issues related to the

18:28data centers that are being uh proposed

18:31because a lot of them need a lot of

18:33electricity and other energy services to

18:36be able to operate but um the

18:41developers are uh

18:45not necessarily in

18:48line with all of the utilities because

18:51there's a lot of different

18:53demands on the grid right now. And so uh

18:56some of the issues that have created

18:59bottlenecks uh for using electricity on

19:01the grid include interconnection cues

19:04that are pretty long. So you have to

19:06wait a long time to be able to actually

19:08connect your uh generator uh to the grid

19:11or get permission to uh connect your

19:15large load to the grid which requires

19:17some period of study for reliability and

19:20other environmental issues. There's also

19:23uh some supply chain issues and

19:25different equipment and uh permitting

19:27delays are already there. So with all of

19:30those background issues happening in the

19:32energy sector as data centers are

19:35looking at building new generation,

19:39they're encountering a lot of challenges

19:41around the uh around the country and in

19:45North Carolina too in terms of being

19:48able to figure out how they're going to

19:50get power. That's led a lot of data

19:52centers to pursue looking at uh what are

19:55the on-site generation options available

19:58especially because the cost of

20:00distributed energy resources has

20:02declined and also there's an

20:04availability of other options for them

20:08uh in terms of uh diesel generators and

20:12uh and natural gas turbines which I'll

20:15I'll talk about some of those uh issues.

20:19So with a lot of limited capacity in the

20:23electric grid, the uh

20:27utilities reser planning reserve margins

20:29are decreasing. It's uh threatening the

20:33reliability of the grid and increasing

20:36some costs to operate the grid.

20:39With all those factors in mind, someone

20:42who's operating a data center might want

20:44reliable electricity. And so they're

20:46looking at what can they do to

20:48supplement their own energy demands

20:52uh

20:53to add this generation. So these are

20:55some of the reasons why I think that uh

20:58data centers in particular are adding so

21:01much

21:03electricity generation and other uh

21:06backup energy resources on site either

21:09for um heating or cooling. There's also

21:13I think it's important to distinguish

21:15between different types of data centers

21:17because data centers get get um

21:20classified differently and uh some of

21:23the really energy most energy hungry

21:27data centers are these uh hyperscalers

21:30or AI optimized data centers which are

21:34um used primarily for AI computational

21:38analysis which may be differentiated

21:41from the data centers that are providing

21:43more um backup cloud storage or uh

21:48crypto mining or other uh uses. So uh

21:53there hasn't been

21:55too much work yet to figure out exactly

21:58what those uh different data center

22:01energy demands will look like. But there

22:04have been some studies to um uh begin

22:07examining what happens uh what what are

22:11the risks to the power grid when these

22:14uh data centers arrive. And most of the

22:17risks could be categorized into a few

22:20areas ranging from short-term to

22:22long-term risks. So I'd say the short

22:24term you have voltage uh stability and

22:28other uh grid stability issues that

22:31happens on a time scale of seconds to

22:33minutes to hours. Um making sure that

22:36the frequency on the grid is maintained

22:40at at the same level and that's um more

22:42of a um grid stability issue. Uh there's

22:48also some medium and long-term risks

22:51like uh ensuring that the available

22:55electricity supply is sufficient and

22:58there's uh adequate resources for uh

23:02planning these. So um

23:06the

23:08traditional frameworks to look at

23:11reliability haven't really managed uh to

23:17look at customers that are adding uh

23:21dozens of of megawatts or potentially

23:23hundreds of megawws

23:25uh very quickly on on the grid. And

23:28that's uh causing a lot of uh concern

23:30about how do you actually make sure that

23:33the system stays reliable and functional

23:36uh given those um high loads or where

23:40where will that electricity comes from?

23:42Also just from the um background level

23:46of the electric grid and and system that

23:48we're dealing with right now is that um

23:51there's a long lead time to building new

23:55uh transmission infrastructure. So, um,

23:59one of the reasons that on-site

24:02generation and backup power has gotten

24:05even more, uh, cost effective and data

24:09centers are looking at this more

24:11seriously is that it takes about 10

24:14years to build or potentially more to

24:16build new transmission lines. And so

24:19transmission lines are really not being

24:22evaluated as a serious option to site

24:25new data centers because the the

24:28challenge is it's really costly and

24:30there's already congestion on existing

24:32transmission lines in the grid. So as a

24:35result of really expensive transmission,

24:38uh data center operators are looking at

24:42what can they do to either store or

24:45generate electricity on site to

24:47supplement that. And um there's

24:50different options that they have but

24:52really um motivated by this long time uh

24:58to build transmission infrastructure and

25:00even other um technologies are taking a

25:05long time to build or uh because of

25:08supply chain issues for natural gas

25:10turbines or other generators. But we're

25:13seeing um

25:16plants look at a few different options

25:18for backup or supplemental power. A lot

25:22of the data centers are looking at um

25:25diesel or natural gas, but those are

25:27sight constrained geographically.

25:29Battery energy storage may be an option

25:32for some data centers because um the

25:36planning and and permitting time

25:38typically can be faster than for uh a

25:44gas turbine.

25:46There's also uh challenges if you were

25:49going I mean solar uh or wind may be uh

25:54faster options to to be able to build

25:57but uh part of the challenge is that

25:59they provide variable electricity. So uh

26:02it's not going to necessarily uh provide

26:06uh 247 backup electricity or or um

26:10full-time electricity. it'll supplement

26:12some of the load and could be coupled

26:15with um battery storage or it could be

26:17coupled with diesel or gas turbines. So

26:21different uh types of data centers are

26:23looking at combinations of these options

26:26based on the geography of where they're

26:28looking to build the the site and um

26:32what is is realistically available. So

26:36when we're talking about reliability

26:37challenges, most data center operators

26:40want 59s of reliability. That means uh

26:4499.999%

26:46of the time they should have uh reliable

26:51power. That equates um to um really uh

26:57high level of of reliability. um even

27:01something like 99%

27:04reliability alone amounts to several

27:08hours of outages per year. So that's why

27:13they're really looking at um almost

27:17consistent access to electricity. And in

27:20order to achieve even more nines of

27:23reliability, that's where backup power

27:26options become uh

27:30what what data centers are looking at to

27:32be able to um meet these and that's

27:35where I think there are different

27:37challenges from the uh electricity side

27:39and looking environmentally at what are

27:41the impacts of these systems because um

27:44your your options for backup power

27:47especially when um the grid electricity

27:51is not available usually have

27:53traditionally uh focused on diesel

27:56generators or natural gas turbines. The

28:00limitation with natural gas is that

28:03there's not pipelines everywhere and um

28:08the storage of natural gas could be

28:12difficult on site. Uh so while diesel

28:16might look more expensive as a fuel on

28:18paper, a lot of uh companies will look

28:22at diesel as an option because the

28:25generator itself is around the same uh

28:28cost but the fuel can be stored easily

28:32more easily on site. It may require

28:35being trucked in uh on roads. So there

28:38may need to be roads but the same

28:40pipeline infrastructure is not as

28:42necessary. It's a more energy dense fuel

28:45than natural gas. So logistically

28:48because uh diesel can take up less

28:51space. Uh that that's one reason why

28:55diesel is being uh projected. There's

28:58definitely challenges environmentally

29:00with using diesel and um we'll hear more

29:04about some of the particulate matter

29:06impacts and other um air pollution

29:09impacts of using diesel. Typically,

29:12diesel might be used for backup power or

29:15potentially during periods where grid

29:18electricity is really expensive or more

29:20expensive and it becomes less expensive

29:22to operate your backup generation. For

29:25diesel, that that seems rare, but for

29:27natural gas turbines, that could be um

29:31pretty that that could uh the power

29:34prices are fairly volatile. And so there

29:36could be periods where um it makes sense

29:39to run natural gas. uh onsite which is

29:43considered kind of behind the meter and

29:45and these are all what I would consider

29:48behind the meter options because from

29:50the utilities perspective they wouldn't

29:53necessarily

29:55uh see or interact with those generation

29:58options. The other uh contenders for

30:03behind the meter on-site generation

30:05include battery storage, which could

30:08reduce um peak loads and is a an

30:13opportunity to store electricity and do

30:18some kind of price arbitrage

30:21from the perspective of the data center.

30:23uh to

30:25store electricity when it's pretty

30:27inexpensive and then use it later when

30:30electricity might be expensive or not

30:32really available on the grid. It's also

30:36probably a way that we'll see uh data

30:39center operators try to circumvent

30:42different requirements from new tariffs

30:44that are being designed like in

30:45Virginia. So for instance, Virginia is

30:48going to start charging uh different

30:50data centers a fee based on their uh

30:54peak capacity and their peak grid

30:57connection request. If you install um

31:02battery storage or even some combination

31:04of other uh on-site generation, it could

31:07be diesel or natural gas. If you install

31:10those onsite, then you could potentially

31:14uh reduce your grid connection request.

31:17And so there are some economic

31:19incentives and reasons why data centers

31:22may uh pursue uh colllocation of adding

31:26some kind of gas turbine or or battery

31:29storage. This could also be supplemented

31:31by something like solar electricity or

31:34or wind. But the challenge with solar

31:37and wind for the data center are that

31:39they might take up more uh space or land

31:43uh to be able to add um

31:47generation. So if uh if they have

31:50available land then they may look at um

31:53using solar as well because it could

31:55reduce or offset some of their um peak

31:58consumption.

32:00I think those are some of the key

32:02motivations for why uh data centers may

32:06be looking at on-site generation. Uh

32:10so

32:13um [clears throat] there have been some

32:15examples of different uh types of

32:18innovations or policies to try to uh

32:23expand uh capacity and protect the grid

32:26reliability uh in PJM Pennsylvania, New

32:30Jersey, Maryland uh grid area. So

32:32they've been looking at ways to try to

32:35speed up interconnections or change the

32:38tariffs that large loads from data

32:41centers face. So that um that's

32:43something that's been happening in

32:45Virginia. And uh then there's been rules

32:48around bringing your own capacity or

32:50generation. So these types of uh setups

32:55are not unique to North Carolina

32:57necessarily, but are happening in other

33:00other states and locations.

33:02And uh there are um

33:08a lot of the the rules are are trying to

33:12figure out who pays for these grid

33:14upgrades or transmission line systems

33:16that um if if the data center is

33:19connected to the grid, they're going to

33:22draw from. But the current tariffs and

33:26uh

33:28frameworks for uh compensating the

33:31utilities are not designed really for

33:35data centers and utilities face big

33:37risks because if they build a lot of

33:40transmission or other infrastructure to

33:42support the data center and then there's

33:45a risk of a stranded asset. If the data

33:47center then decides to stop operations

33:49or um leave, then they've invested in a

33:52lot of um new electricity infrastructure

33:56uh for areas that may not um need it. So

34:01um

34:03there have been some studies to look at

34:05well what if data centers could uh

34:08benefit the grid and serve as an asset

34:12for demand response and to some extent

34:15that's been uh proposed but the the

34:18challenge is that you would have to get

34:21uh utilities and the data centers in

34:23agreement of how much time the data

34:27center is actually going to be

34:28operational and and during which times s

34:31uh data centers would allow uh the

34:35utility to

34:36buy their uh demand or or be able to

34:41ramp them up or down depending on the

34:44grid needs. So um on a day-to-day basis

34:48balancing the grid is really important

34:50and contributes to a large portion of

34:53the operational cost of making

34:55electricity affordable. But then um the

35:00data center is uh

35:04it's it's not clear the extent to which

35:07um the data center will be able to

35:09curtail uh its its load during periods

35:13of uh a demand response setup. So uh

35:18data centers would be able to

35:20participate in these programs more

35:21effectively if they have battery storage

35:24or other systems on site. Um

35:28but that's not always the case. So um

35:36uh essentially like if you have some uh

35:40flexibility which flexibility I would

35:42define as the ability to u ramp up or

35:46down your electricity

35:49uh usage in a very quick manner like

35:53within seconds to minutes. Like if

35:55you're able to do that because you have

35:57on-site battery storage or some other

36:01on-site generation like a gas turbine,

36:05then uh you could benefit the utility in

36:09a way where um the peak demand might be

36:13flexible. And so there could be some uh

36:16mutually beneficial arrangements, but

36:18there also could be some uh situations

36:20where uh it it doesn't really benefit

36:25the utility or the utility gets these um

36:28peak periods. So um

36:32the that that's uh so there are um types

36:37of data data center flexibility options

36:40that that are available and in addition

36:43to uh thinking about grid reliability.

36:47Uh on-site storage or on-site gas or

36:50diesel generation is uh often examined

36:54as a strategy uh to be able to do that.

36:57uh it's faster to build and connect and

37:01um

37:02and so and you don't have to deal with

37:06uh adding much more transmission lines

37:09which can be really costly and uh is

37:13requires coordination across multiple

37:15jurisdictions and stakeholders uh

37:17potentially going across state lines.

37:20Um,

37:22in Texas, a lot of new data centers are

37:25adding battery energy storage for this

37:27reason, for flexibility. And so it's

37:30likely that other data centers will look

37:34at that um because especially

37:37um in states now like in Virginia where

37:40they've changed their tariff structure

37:42where it really incentivizes adding um

37:47battery storage to the data center or

37:49else um the the data centers will have a

37:53lot of um increased costs. So those are

37:57um some of the issues that utilities are

38:01facing right now. Um

38:07so I'll talk a little bit about

38:09Virginia's tariffs and then I'll talk

38:11about the some of the sustainability

38:13considerations. Um in uh Virginia

38:17they've recently or as of so January 1st

38:212027 they'll start implementing new

38:24types of tariffs and uh the main changes

38:29are that instead of having an on- peak

38:32uh demand charge for transmission

38:34systems they'll look at the actual name

38:37plate capacity of uh what's installed at

38:41the uh data center and uh create a uh a

38:46fee based on uh a percentage of the

38:51minimum generation demand billing. and

38:54um

38:56they'll

38:59so there's a considerable upfront

39:01investment that's um going to be

39:05required for new uh data centers there

39:08and it's a new service class whereas in

39:11um North Carolina or other other states

39:14there's not um really a separate tariff

39:18class for for data centers. So that's um

39:22uh an example of things that are

39:24happening. Uh in some of our research,

39:26we've been looking at like what that

39:28actually means for uh data centers and

39:32power generation of what they would

39:34likely do. And so we've simulated some

39:38of the uh typical load profiles for a

39:43hypers scale data center uh which

39:47might peak uh in in the middle of of the

39:51day. But um this is uh based on some

39:55assumptions related to their pretty

39:58consistent utilization rate of 80 80%

40:01and um

40:05and a load of like 500 megawws. So 500

40:09megawatts is a really big uh system and

40:13uh there are different scales of of data

40:15centers. Uh some of the systems being

40:18proposed in North Carolina might um be

40:20much smaller than 500 megawws. Uh but

40:23they still would look at on-site power.

40:27And uh

40:29the other the other option that's not or

40:32the other options that have been

40:33proposed for data centers around the

40:35country but haven't really been utilized

40:39um yet include hydrogen and nuclear. So

40:43hydrogen could be used through a fuel

40:46cell that's on site. Fuel cells actually

40:49can operate similar to a generator and

40:52uh use uh natural gas fuel or hydrogen.

40:57Um hydrogen can be produced in different

41:01ways. So there um there are different

41:05types of hydrogen but um a lot of the

41:09hydrogen economy in the US is more

41:11concentrated in states like California

41:14or or Texas and and not as it's not as

41:18uh common in North Carolina for

41:22instance. And then um small modular

41:24reactors have been proposed as options

41:27for uh colllocation with data centers

41:30but um there's not many operational

41:34projects. Uh one is that they've been

41:37expensive. There's a there's a project

41:39that may be developed in Ohio uh with

41:43New Scale, which is a company that has

41:45developed small modular reactors, but

41:48they're not um

41:51widely seen in existing data centers

41:54that um have already been built across

41:58the US. The reason why uh small modular

42:01reactors have been proposed is that they

42:04could co co-produce both uh electricity

42:07and heat uh which can provide some

42:12cooling energy depending on the systems

42:14that are being used for the data center.

42:17But um

42:21I think there there's a lot of

42:23permitting and other planning challenges

42:25around adding new um small modular

42:30reactors uh for collocation with data

42:33centers. So even though uh they they may

42:36be an option in the future, they remain

42:39fairly expensive uh and probably

42:43uh

42:45unlikely to uh unlikely to be developed

42:50in a in quickly and an economically uh

42:54quick manner. So we're seeing that um

42:58that's where battery storage has has

43:00played a big role. But um there are a

43:03lot of uh issues also with thinking

43:06about diesel or uh natural gas as

43:10on-site generators because they might be

43:12subjected they might not be subjected to

43:14the same uh regulations as typical

43:17facilities that are connected to the uh

43:20power grid depending on their size and

43:23capacity. And so that can lead to um

43:28utilizing them during different times

43:29that violate or that is uh higher than

43:33what is regulated by the clean air act.

43:36And uh some of the concerning emissions

43:40that uh Sarov will talk about more uh

43:43include NOx uh particulate matter PM2.5

43:47carbon monoxide and VOCC's. Also, um

43:51greenhouse gas emissions are are another

43:53issue because if you look at um the

43:56greenhouse gas emissions from uh diesel

43:59versus natural gas versus uh

44:03solar, hydropower or other

44:07wind uh power sources. Um,

44:11diesel, I mean, um, diesel is by far the

44:15highest and, um, averages around 700 to

44:20800, uh, grams of CO2 per kilowatt hour,

44:24whereas natural gas is about, um, uh, 50

44:29or 60% of the those emissions um, from

44:33diesel. uh but uh solar wind even when

44:38backed up by battery storage are much

44:41lower on the order of 20 to 50 grams of

44:45CO2 when you account for all the

44:47manufacturing and construction

44:49requirements. Uh natural gas may be

44:52between 300 to 500 grams of CO2 per

44:56kilowatt hour. So that those are uh

44:58important to think about. Um and then

45:00there's issues of uh the the land

45:04footprint and noise. So um the uh diesel

45:09can often range from uh a small diesel

45:13generator that's only 50 kows goes from

45:1670 to 85 dB. But once they get beyond a

45:20megawatt or 1500 kilowatts in capacity,

45:23they can easily exceed 100 dB. uh gas

45:27turbines that are also typically larger

45:30than a megawatt. Uh we'll go into the

45:33110 to 155 decel range. Uh and battery

45:38energy storage ranges from 0 to 50 dB

45:42depending on the size and cooling fan

45:44components. So, um, if you're looking at

45:47noise as another factor to consider for

45:50on-site generation, I think, um,

45:54there's, uh, quite a big range there.

45:57Uh, so I I wanted to, uh, talk to you

46:00about these different options and then,

46:03uh,

46:06and I gave you an overview of why I

46:08think on-site generation is being uh,

46:11examined more clearly. there are uh

46:15different options for data centers uh to

46:19pursue and they they lead to different

46:22impacts related to their um

46:24environmental concerns and also um

46:26noise. So I'm happy to take any

46:28questions you may have or answer any or

46:31follow up with any discussion. Thank

46:33you.

46:34>> You you mentioned early on about natural

46:37gas storage on site. Uh, how would you

46:41store natural gas on site?

46:44>> Oh, I think that would be very difficult

46:46to do. Um, and that's one of the I I

46:48mean, uh, there's um

46:53that that's why a lot of companies are

46:56looking at diesel as an alternative to

46:58natural.

46:58>> I mean, because the distribution

47:00companies, gas distribution companies

47:02have LG tanks where they

47:03>> Yeah.

47:04>> pull it down. And I don't think that

47:06would be real feasible for

47:08smaller operation.

47:10>> Yeah, I completely agree that and that's

47:12why a lot of um the data centers that

47:15I've seen have looked at diesel even

47:17though if you look at like an energetic

47:19basis, diesel is often way more

47:20expensive as a fuel. So you might think,

47:22oh, why would they even choose diesel

47:24instead of natural?

47:25>> They've got a a gas line going into the

47:28facility, then that would make it

47:30>> that would that would make it work. But

47:31yeah, if you don't if you don't already

47:32have the line, then you have to um

47:34account for the line. Yeah.

47:36>> And as far as battery storage, we're

47:39looking at that is shaving peak loads

47:42for the most part,

47:44not for any kind of primary use

47:47>> mostly. But um but I mean it it's hard.

47:50It depends on what you mean by primary

47:52use. like it's not it's not going to um

47:54completely power the data center on its

47:57own, but I would imagine that if you

47:59have a battery on site, it'll probably

48:01get used every day.

48:03>> Um and so it's going to be charging and

48:07discharging almost daily and uh cycling

48:11through. So that will affect the peak

48:12load, but it will also provide some

48:14supplemental

48:16um reduction during non- peak times as

48:19well. Uh that's compared to a diesel

48:22generator where you might have the

48:25diesel gener generator available on site

48:29but uh it may only get used on peak days

48:34to manage the peak load or it may only

48:36get used in times when there are

48:38reliability concerns. There might be an

48:40outage and for backup power. So the

48:44battery battery may be used for backup

48:47power and maintaining the grid

48:49stability,

48:50but if it's installed on site, it would

48:53likely be cycled almost daily, whereas

48:56the diesel generator probably wouldn't.

48:59>> Okay.

49:02And one one last thing you also

49:04mentioned solar and wind as as possible

49:07options from

49:09>> from a practical standpoint with today's

49:11technology. Is that all that feasible

49:14for unless you have a lot of land mass?

49:18Um so uh there there it is highly

49:22feasible for today not to power the

49:24whole system but a lot of data centers

49:27are looking at adding solar or wind

49:29depending on their geography because

49:31they can shave off their peak demands or

49:34they can use it as a supplemental

49:36source. So, probably not for primary,

49:38but I mean, if you could cover 50 to 75%

49:41of your electricity needs by adding

49:44solar on-site or nearby, then a lot of

49:48places are looking at that because it um

49:51compares with then they have to purchase

49:54less electricity from the utility

49:56company.

49:56>> I'm I'm I'm I've got a myopic view

50:00because I'm looking at at the Apex area.

50:02>> Sure. So, I'm I'm trying to see whether

50:05that those two options would be really

50:08feasible.

50:10>> Um, I mean, yeah, within APE, I mean,

50:13within Apex, I think you could see some

50:15solar on like the the roof of the design

50:18of a possible data center. I think that

50:21would be um a possibility. It would

50:23probably not primarily power the data

50:26center. It would it would be a

50:28supplemental. see how it could primarily

50:30power it but yeah it would

50:32>> provide some supplemental energy.

50:34>> Yeah. Yeah. And it would probably limit

50:36their peak capacity requirement from the

50:39uh from the electric utility.

50:46So with this behind the meter

50:48generation, so like sort of a utility

50:51like Duke, you know, more so in days

50:53gone by, they had climate goals where

50:56they were trying to reduce carbon and

50:57switch sort of their mix to more carbon

51:00neutral.

51:01>> I mean, it's still a state law.

51:03>> Yeah. But when you get to behind the

51:05meter generation like this, would they

51:08sort of be exempt from any of those

51:10goals and it would just be up to the

51:12individual private company?

51:15Well, the the thing about um behind the

51:18meter that well, so um the thing about

51:21behind the meter is if you were big

51:23enough and for a large load like a data

51:26center, it's different than like if you

51:28wanted solar on your house because you

51:30could you could basically um

51:34add a lot of these generation sources

51:37behind the meter because they're not

51:39connecting with or interacting with the

51:42main utility. And if they have the um

51:46switches to basically

51:48choose when they're using uh what's

51:51called self consumption or basically

51:53generating on site, then they could do

51:57that when it's economically feasible and

52:00then buy electricity from the utility

52:02when it's uh during other times. And um

52:08that that is it is possible to do that

52:11without having to negotiate so much with

52:14the utility because they could um

52:18operate those generators and systems

52:21themselves. And so you could

52:25collectively get a bunch of these behind

52:27the meter generators who were just using

52:31just exclusively natural gas and diesel

52:34for power generation.

52:36>> Yeah.

52:36>> And Okay.

52:38>> And then that's behind the meter because

52:41um they're not interacting as much with

52:44the utility. It's harder to probably do

52:45that with natural gas because you would

52:48need the pipeline and dist I mean there

52:52are other

52:54factors but for diesel generation it's

52:56definitely possible to um operate the

53:00diesel generator to uh provide

53:02electricity that then is is not seen

53:05behind the meter basically just refers

53:07to not not like invisible to the

53:11utility. Um and so they're not selling

53:14that power or heat and then they're they

53:19are not um responsible for it,

53:22>> right? So they they themselves even

53:24though they're doing the same thing as

53:27like Duke Energy, they they don't might

53:30not have any climate goals or anything

53:32like that. Definitely they're not

53:34selferving. Yeah, there's there may not

53:36be the same uh yeah, they can circumvent

53:39the the rules of like the carbon plan or

53:42other um emission requirements.

53:46>> Thank you.

53:49>> So, I wanted to say first off um thank

53:51you very much because this actually

53:52helped a lot. The one that was

53:54originally proposed for this area was

53:57300 megawws and there was a discussion

53:59of diesel. So, I'm imagining you trucks

54:02coming in every day. Um, I also

54:05appreciate, you know, that you spoke on

54:07the noise and the, um,

54:10what it would do to the power grid. Um,

54:14I guess my biggest question is what have

54:17you seen with the strains on the power

54:19grids of what they've done to an area

54:22like brownouts or blackouts or things

54:24like that? Um, and

54:29as quickly as this is changing, is it

54:32possible that you could see a developer

54:35coming in putting something down like

54:36this and then within a few years it

54:40rapidly changing that that they need to

54:43up and move somewhere else.

54:47Those are really good questions. I mean

54:50I think uh because of the complex nature

54:53of the power grid it's hard to attribute

54:55some brownouts specifically to the

54:57presence of data centers but if we look

54:59at like historical events that have

55:02happened recently including like winter

55:04storm Yuri in Texas in 2021 there's

55:08definitely pressures on load growth

55:10without and also unc

55:13coordinated risk mitigation for planning

55:15that uh temperatures might go uh very

55:19low during those periods which caused a

55:22rapid rise in demand for electricity at

55:24a time when then there wasn't

55:26electricity and it created this ripple

55:28effect where there were outages for

55:30days. Um that definitely play like this

55:36the constrained supply

55:39can definitely play a role and not and

55:42having these um smaller reserve margins

55:45going forward is a risk that um

55:48increased

55:50large loads on the grid could reduce the

55:52planning reserve margin that's

55:54available. um that could happen in North

55:57Carolina even with other

56:00events with hurricanes or um uh

56:04extreme weather. Um but also so I think

56:07it's there's um compound risks happening

56:10and it's so it's um it's not just the

56:13weather but it's also the reduced

56:17planning reserve margin which is like a

56:19long-term planning horizon uh

56:24during different times. So that that's

56:25definitely a challenge. Um and then uh

56:30yeah there are risks of being st I mean

56:32uh that I think that's what uh Virginia

56:35was trying to do in terms of redesigning

56:38their tariff is um force a lot of the

56:43data centers to pay more of the costs

56:46upfront for a longer period of time. And

56:51so that was thought that that could

56:53potentially reduce the risk that they

56:56could leave in two or three years or you

56:59know go bankrupt and that or you know

57:01just stop and then there was a lot of

57:03and I think from the utility perspective

57:05the reason that would be a problem is

57:08that if you upgraded a lot of um

57:11substations, transformers, other

57:13electrical equipment in that area that's

57:15a major investment and uh if If

57:20[clears throat] you did it in an area

57:21where it's o much more oversized

57:23compared to the load, then um someone

57:27else is going to have to pay for it.

57:31>> Thank you.

57:40>> Thank you so much. Appreciate it.

57:52Thank you, Dr. Kitner. All right, our

57:54next speaker is Dr. Sarab Arunachalam.

57:56He'll be speaking today. His background

57:58is in air quality emissions and modeling

58:00and public health. He's over 20 years of

58:02experience in emissions and air quality

58:03modeling across multiple scales. And

58:05he'll be providing us with a lot of

58:07great information today.

58:11>> Good evening.

58:15Can you hear me? Okay, it's good. Thank

58:17you. Thank you for the invitation,

58:19Bruce, for this um environmental

58:21advisory board and um thanks to see this

58:24board stood up now since the various

58:26town council meetings over the last year

58:28or so. I'm very happy to be here today

58:31and to talk about the environmental

58:32impacts of data centers. Hopefully this

58:34will be a nice complement and picking up

58:36from where Noah um last stopped with

58:39this uh focus on energy and how energy

58:41can be um created for these data centers

58:44and where the environmental impacts may

58:46be. With that, I'm a professor at UNCC.

58:48I direct um a center for environmental

58:50modeling for policy development as well

58:52as the um institute for the environment

58:53where I'm a deputy director um looking

58:55at various aspects of um environmental

58:57impacts uh specifically focusing on both

59:00the environment and subsequent public

59:02health impacts. Here's a picture from uh

59:05what is called as the data alley uh data

59:08center alley. Uh this is Virginia. You

59:10don't have to think too far to guess

59:12where this is taken from. This is one of

59:14the pictures in Luden County in in

59:15Northern Virginia. And you can see a

59:18bunch of data centers uh buildings in um

59:22very very close proximity to residential

59:24neighborhood on the lower right half of

59:26the screen. And then if you zoom into

59:28one of these buildings, you can see um

59:31uh both generators as well as the um

59:33cooling [clears throat] towers on top of

59:34the building. This is how a data center

59:37campus looks like. And I know we were

59:39all looking at the potential um for one

59:42of these to come very close to where we

59:44are today in Apex. So this is what this

59:46could look like um if they were

59:48approved, processed and so on. So what

59:50I'll do over the next few minutes is to

59:52give you sort of a three or fourpart

59:54presentation. first to give a very broad

59:56um national landscape of how this

59:58particular um area or field looks like

1:00:01in terms of data centers and where we

1:00:02are going and then I'll talk about the

1:00:04environmental impacts of what data

1:00:06centers mean and then uh zoom in to our

1:00:10own town of Apex to see what could

1:00:12happen if the proposed and now withdrawn

1:00:15data center was here. Okay.

1:00:18Um just um large big picture from the

1:00:21current national or federal

1:00:23administration. This is from the White

1:00:24House. Uh if you go to ai.gov, you can

1:00:27see this. There is an action plan, a a

1:00:29action plan by the current

1:00:30administration where the broad big

1:00:32picture statement and the focus is the

1:00:34US is in a race, a race to achieve

1:00:37global dominance in artificial

1:00:39intelligence. Whoever has the largest AI

1:00:41ecosystem will set global AI standards

1:00:44and reap broad economic and military

1:00:46benefits. With that, this is focused on

1:00:48three different pillars. The first one

1:00:50is accelerate AI innovation. Second is

1:00:52build AI infrastructure and the third is

1:00:54lead in AI international diplomacy and

1:00:57security. So this is sort of the message

1:00:58that we have from the current

1:00:59administration at the top down in terms

1:01:01of where this country should be going.

1:01:03Okay with that um just a broad picture

1:01:06on terms of where the global electricity

1:01:08growth is. Uh we saw some numbers from

1:01:10Noah earlier. I'll present a different

1:01:12perspective on this but in general you

1:01:14see some very important numbers come out

1:01:16of this. If you look at um the energy

1:01:18demand from data centers this uh fifth

1:01:21bar from the top and one in orange

1:01:22you'll see that there is a demand by

1:01:24about 500 terowatt hours over the next

1:01:28few years and compared to rest of the

1:01:30sectors you can see heavy industry all

1:01:31the way at the top 1500 so about a third

1:01:33of the total demand um of heavy industry

1:01:36is going to be from data centers and

1:01:38this collectively if you look at all the

1:01:40different rows uh the bars that I have

1:01:42that's about 10% of the total demand

1:01:44that we expect globally that's you

1:01:46expect that to be a very very large

1:01:48portion of what is expected to grow uh

1:01:50in terms of demand over the next few

1:01:52years. By few years, this is from 2020

1:01:54for 2030. It's a six-year projection um

1:01:57that came out recently. Um if you look

1:01:59at the data energy growth, um both the

1:02:01US and China really dominate. They are

1:02:04predicted to account for nearly 80% of

1:02:05the global growth in electricity

1:02:06consumption by data centers in the next

1:02:08few years. Okay, you can see the very

1:02:10large demand um in US at the bottom,

1:02:12then China and the rest of um Europe,

1:02:14Asia, etc. uh rest of Asia excluding

1:02:16China you can see where they stand but

1:02:18clearly US and China dominate this by

1:02:21about 80% or so with all of that being

1:02:24said so where is really the energy

1:02:26footprint if you look at AI uh some of

1:02:28us may be using it some of us may stand

1:02:30u a little bit far away from use of AI

1:02:32like myself um but the point is um if

1:02:35you look at different aspects of what

1:02:37most of us may do uh depending on what

1:02:39exactly you're doing you may have a

1:02:41different use of how much energy is

1:02:43being consumed by the particular

1:02:45activity. If you look at this, you have

1:02:47at the bottom a text classification

1:02:48where you're doing something on a

1:02:49textbased search. Um, so if different AI

1:02:52tools have different consumption models

1:02:54based upon the specifics of what is

1:02:56being used and so you see a range of the

1:02:58lowest on text classification, you see

1:03:00somewhere between about 0.5 to a little

1:03:02over one or so. That's the demand in

1:03:05terms of energy consumed to perform a

1:03:07text search about thousand times. So

1:03:08that gives you a range. As you keep

1:03:10going up this particular chart to

1:03:12answering a question or automatic speech

1:03:13recognition or image captioning or image

1:03:15generation which is the worst of all if

1:03:17you use AI to generate an image that's

1:03:19the worst of all you can see how the

1:03:21energy demand associated with that can

1:03:22be several hundreds of um what hours

1:03:26okay so that gives you a sense of what

1:03:28exactly we are doing when you do a

1:03:29search whether it's a textbased search

1:03:31or asking a question or having a

1:03:32conversation with an AI tool so these um

1:03:35numbers give you a spread of these

1:03:37numbers in terms of what the energy

1:03:38demand is M with that um couple of years

1:03:41ago there was a proposed national

1:03:43legislation I think this was under the

1:03:44previous administration um called the AI

1:03:47environmental impacts act of 2024 this

1:03:49was a senate bill introduced by Senator

1:03:51Marky of Massachusetts where he had

1:03:53three primary objectives part of this

1:03:55bill u first it directs the EPA to carry

1:03:58out a study on AI's environmental

1:03:59impacts number one second directs the

1:04:01national institute of standards and

1:04:02technology or NIST to develop standards

1:04:04to measure and report the full range of

1:04:06AI's environmental impacts and the Third

1:04:09is to create a voluntary framework for

1:04:10air developers to report on

1:04:13environmental impacts. Okay, this was a

1:04:15proposed legislation. Of course, it's

1:04:16yet to be passed as you may imagine as

1:04:18of now.

1:04:20So what are the numbers? So now I'll

1:04:22talk um in terms of what the numbers are

1:04:24in terms of what different sources

1:04:25there. There is no central repository of

1:04:27this. The different groups have put

1:04:28together different set of numbers. This

1:04:30is one take on where the numbers are,

1:04:32but I think these are pretty solid in

1:04:34terms of uh where these stand on a

1:04:36relative basis. So you have the first

1:04:38table, the first half of the screen

1:04:40showing the total numbers for the US.

1:04:42Yeah, there about 3,44 data centers,

1:04:45about 1,300 are operating as we speak

1:04:47and about 2,000 being planned. Then I

1:04:49picked up numbers for two different

1:04:51states. One the state of North Carolina

1:04:53where leave where we live in. So that's

1:04:54of interest to us. But also looking at

1:04:56our neighbor to the north, Virginia

1:04:58given that we hear a lot about Virginia

1:04:59and what is going on there. Just kind of

1:05:01give you a perspective and a sense of

1:05:02comparison between North Carolina and

1:05:04Virginia. So you see where North

1:05:06Carolina stands about 65 data centers in

1:05:08the state 13 are operating and about 52

1:05:10being planned. So you do see that quite

1:05:12a lot are being planned in our own

1:05:14state. How does it compare to Virginia?

1:05:16The next row you see very very large

1:05:18numbers state of Virginia right 371

1:05:20versus 458 being planned and um to some

1:05:23degree I've seen numbers where they say

1:05:25the energy demand for the state of

1:05:27Virginia's data centers are larger than

1:05:29many other countries of the world. Okay.

1:05:31Then the second half of the table or the

1:05:33second table looks at the uh number of

1:05:35unique developers in terms of um number

1:05:37of um businesses as well as the

1:05:40operating as well as plan capacity and

1:05:42that gives you a sense of you know the

1:05:43projected growth for the US 61,000 um

1:05:46operating versus 380,000 plants about

1:05:48you know little over um five times

1:05:50growth there in the state of um our own

1:05:53state we go from about 500 to about

1:05:554,000 okay that's another eight times

1:05:57projected growth in terms of what is

1:05:59being planned and then Virginia from

1:06:0117,000 to 37,000 just over 2x in

1:06:04Virginia and I have some examples of um

1:06:07uh the largest that is operating both in

1:06:09the US or and our own states of North

1:06:12Carolina then Virginia up north as well

1:06:14as the largest being planned within the

1:06:15US NC just to give you a sense of how

1:06:18large some of these are in terms of

1:06:20current operating as well as what is

1:06:22being planned down the road

1:06:25okay with that so with that broad

1:06:26landscape now we'll look at what does

1:06:28this mean in terms of data centers and

1:06:30then their impacts on the environment

1:06:31and health. Um I'll go through a um

1:06:35couple of slides with a lot of different

1:06:36bullets. Um given the time that we have,

1:06:38I could be giving an entire hours worth

1:06:40of talk on each bullet on the screen.

1:06:42But what I chose to do today was to give

1:06:45you a broad picture of the various

1:06:46impacts, but then I'll touch up on a

1:06:48couple of those a little bit in detail

1:06:49given the time that we have otherwise

1:06:51you're going to be here for a very very

1:06:52very long day and week or more. Okay. Um

1:06:55the first one of course is the um

1:06:57increased energy demand and the strain

1:06:59on the grid. Um we've talked about a

1:07:01little bit uh with Noah's um excellent

1:07:03briefing before me and you saw some of

1:07:05that earlier. Then um we did talk about

1:07:08um uh noise but again data centers just

1:07:10for the sake of the number of generators

1:07:12the cooling that are running they cause

1:07:14a lot of um noise pollution. Um there's

1:07:17been a lot of articles if you Google

1:07:19data noise pollution you'll see bunch of

1:07:21live videos and audio clips from people

1:07:23who live next to data centers. Um lot of

1:07:25these do exceed the standard 85 decibel

1:07:28threshold which are considered harmful.

1:07:30And if you think about airports and

1:07:32airplane operation around airports, you

1:07:34know um FAA has these um noise visible

1:07:37maps which shows what are the footprint

1:07:39of aircraft around airports when the

1:07:41planes take off and land. What the

1:07:42footprint is in terms of noise pollution

1:07:44of aircraft and then they have used

1:07:46those map to see what are the

1:07:47residential area areas which are right

1:07:49under the planes taking off and landing

1:07:51and what levels of decibb they are

1:07:53experiencing and then there are some

1:07:55mitigation option that FA has to support

1:07:58these communities who live next ports.

1:07:59The point here is data centers cause

1:08:01harmful noise pollution and often these

1:08:03exceed these threshold which are

1:08:05considered harmful for human health. The

1:08:07other aspect of is just given the heat

1:08:09that these generate they enhance thermal

1:08:11pollution and then they cause what are

1:08:13called as heat islands which are

1:08:14basically you know enclose small areas

1:08:17of pocket depending on the size of the

1:08:19um uh energy that is being generated and

1:08:21they have a much larger capacity to

1:08:24store heat which means gets very hot

1:08:26right there. So you have these um data

1:08:28centers when these are very large and

1:08:30have a lot of generators they cause

1:08:31talent. There are a couple of studies

1:08:32that came out recently which documented

1:08:34these and the impacts of this. The next

1:08:36one is a biggie um every one of them

1:08:38biggie but the next one is particularly

1:08:39important for us to think about um

1:08:41impact u data centers impact water

1:08:43availability as well as water quality.

1:08:45Okay, availability. A typical hypers

1:08:47scale data center can use a couple of

1:08:49millions. Sometimes I've seen numbers as

1:08:51high as 5 million gallons of water to be

1:08:53used per day for cooling these towers

1:08:55and and of course it depends on the size

1:08:57of the data center and how much

1:08:58generators they have and so on and so

1:08:59forth. But hypers scale data centers can

1:09:01consume a lot lot lot of water. Okay.

1:09:04Then in terms of quality they also

1:09:06affect water quality. Uh one of the

1:09:07specific pollutants we talked about is

1:09:09depending on what is being used in the

1:09:10data center and the amount of effluence

1:09:12they have. You can have PAS. That's a

1:09:14very very um uh important chemical to

1:09:16think about. Uh these are also called as

1:09:18forever chemicals. These are called

1:09:20forever only because they stay in the

1:09:23atmosphere in the water forever. They

1:09:25don't deteriorate. They don't

1:09:26disintegrate and they cause all kinds of

1:09:28health issues. Tons of studies. Um in

1:09:30our own state in Wilmington, there was a

1:09:32chem plant that had quite a bit of

1:09:34issues. You may have seen um in news out

1:09:36there. So we have those. So based upon

1:09:38the usage of chemicals and materials in

1:09:40a data center facility, you will have

1:09:42airfluent chili pas and that will affect

1:09:44the water quality of downstream. So if

1:09:46you are in apex for example the rivers

1:09:48that go from here, they go to fateville

1:09:50if I remember right. So fate will be

1:09:51talking about hey what's going on with

1:09:53water coming from this part of the

1:09:54state. Okay, there a lot of other toxins

1:09:56uh nutrient loads depending on the air

1:09:58pollution that gets deposited on the

1:10:00water based upon the water cooling you

1:10:02may have acid deposition in the water

1:10:03bodies which leads to other um

1:10:05detrimental impacts. Data centers also

1:10:08cause light pollution uh depending on

1:10:09the height of the building and where

1:10:11they're located. They have these uh

1:10:13buildings which have lights and of

1:10:14course they'll cause light pollution

1:10:15there. Fire risk. This is an important

1:10:17one. We need to think about what are the

1:10:19potential risks due to fire. So when

1:10:21these large data centers with very large

1:10:23generators um catch fire due to an

1:10:25accident or other then you have a

1:10:27serious fire risk which um small towns

1:10:29like Apex I I like to call it small town

1:10:32right I don't think we are large yet um

1:10:34um they have a pretty serious um issue

1:10:36in terms of what needs to be done and

1:10:38the capacity needs to be um managed up

1:10:40front to see how to manage this

1:10:42particular fire should that ever happen.

1:10:44Then um as part of data centers you also

1:10:47have this electricity um transmission

1:10:49high voltage transmission lines which

1:10:50increase the electromagnetic force

1:10:52exposure right so your EMF exposure from

1:10:54high voltage transmission lines. So

1:10:56depending on where these data centers

1:10:57are you may have this HV transmission

1:11:00lines which increase the EMF force um

1:11:02and hence the corresponding exposure to

1:11:04people who live next to that. Continuing

1:11:07on the same list just more of these

1:11:09think about e-waste. There's a lot of uh

1:11:11potential electronic waste comes out of

1:11:12these. we need to think about u proper

1:11:15um uh disposal of these and what may

1:11:17happen in the near- term as well as long

1:11:20term to take care of this e-waste. The

1:11:22next aspect is depending on the various

1:11:24um aspects of what goes into building

1:11:26these uh center generators you may have

1:11:29a lot of um minerals that are being

1:11:31used. So one has to one has to have lot

1:11:33of uh minerals that are exploited um

1:11:36explode and so on. So this will enhance

1:11:38critical mineral exploitation from the

1:11:40earth's resources. The next one is land

1:11:43use degradation conversion. So depending

1:11:45on the kind of land use it may have, we

1:11:47talk about zoning. The town uh looks at

1:11:49zoning issues a lot. So you convert a

1:11:51specific parcel of land from one to the

1:11:53other. And when you do that, you have um

1:11:55various aspects of what happens to the

1:11:56land itself from perhaps an agricultural

1:11:59piece of land um pastures and so on to

1:12:02having more concrete. Now when you have

1:12:03more concrete, you create imperous land

1:12:05which means there are issues with water

1:12:06flow and so on. Okay? So then you have

1:12:08flood risk should there ever be a flood

1:12:10in that part of the state. Then of

1:12:11course when you have land use conversion

1:12:13you displace communities and then you

1:12:14also displace wildlife and natural

1:12:16habitats around the specific location or

1:12:18the parcel of land you're looking at.

1:12:20And the last one is public health

1:12:21impacts. I talked about a lot of

1:12:23environmental issues before this but

1:12:24then you also have significant public

1:12:25health impacts specifically from the

1:12:27emissions that come out of the data

1:12:28centers in terms of what is being

1:12:30combusted if they're combusting anything

1:12:32and the associated with air pollution

1:12:33exposure and then you also have um

1:12:35endocrine disruption based upon the

1:12:37specific chemicals that are being used

1:12:38there. Okay. So this is a broad spectrum

1:12:41of um uh the environmental concerns and

1:12:43one needs to look at um all of them in a

1:12:44very uh detailed manner. Now in terms of

1:12:48air pollution or even thinking about

1:12:49different aspects of the environmental

1:12:50concerns I talked in the last two slides

1:12:53um when you have a generator or a bunch

1:12:55of generators at a specific data center

1:12:58there are multiple variables one need to

1:12:59think about in terms of what really

1:13:00happens at the data center and I listed

1:13:02a few here. The first one is fuel type.

1:13:04Okay we we I think this is a very

1:13:05obvious one. Most of us know this and

1:13:07Noah touched upon this um this a little

1:13:09bit earlier in his talk. So the type of

1:13:11fuel that is being used by data center

1:13:13is something to think about. So diesel

1:13:14is the one that's most commonly used. Um

1:13:17then comes natural gas which get a

1:13:18little bit cleaner but still it is

1:13:20fossil fuel at the end of the day. Uh

1:13:21then fuel cell which um bring out

1:13:23hydrogen. Then um terms of renewables

1:13:26there is something called as

1:13:27hydrogenated vegetable oil. Okay some um

1:13:29places use them. Uh then there could be

1:13:31a whole lot of other renewables. So one

1:13:33need to think about what is the specific

1:13:35type of fuel that is being used for data

1:13:37center and hence what are the associated

1:13:39environmental health impacts. Then the

1:13:41next variable to think about is what is

1:13:42the capacity right both in terms of the

1:13:44total capacity for example the proposed

1:13:47new data center which was withdrawn the

1:13:49capacity was 300 megawatt and then it is

1:13:52not a single unit of 300 megawatt they

1:13:54talked about how there could be multiple

1:13:56smaller generators in this example I

1:13:58think they proposed 100 or so of three

1:14:00megawatt capacity so you have multiple

1:14:02generators each of some capacity 1 2 3

1:14:04megawatt and so on then you have a whole

1:14:06bunch of those 100 200 300 depending on

1:14:08what is the size the third one is how

1:14:10you configure the specific generator on

1:14:13site. Right? So in this example, there

1:14:15was discussion of having four buildings

1:14:16if I remember right and four buildings

1:14:18will have somehow 100 generators

1:14:19distributed across the four buildings.

1:14:21And then when you have four buildings,

1:14:22100 generators, you have to think about

1:14:24what is the specific configuration each

1:14:26stack because that leads to different

1:14:27dispersion parameters. So what is the

1:14:29stack height? What is the stack

1:14:30diameter? What is the exit gas velocity?

1:14:32What is the exit temperature and so on.

1:14:34So those determine the dispersion and

1:14:35hence the associated air pollution that

1:14:37come from the stack. So again think

1:14:40about what is the specific configuration

1:14:41for each of these um generators. The

1:14:44next one is duration of the run. Okay

1:14:46depending on the configuration and the

1:14:48need and how they integrate with the

1:14:49grid or not you have to think about how

1:14:52long these generators run and the

1:14:55associated air pollution exposure is a

1:14:56function of whether these are continuous

1:14:58or they do start stop and so on and they

1:15:00all have different impacts. Then

1:15:02depending on any or all of the uh four

1:15:06bullets above the last one is what are

1:15:08the pollutants emitted based upon the

1:15:09fuel type. Um you have no knock socks

1:15:12PM2.5. I'll talk about PM2.5 in a minute

1:15:14a little bit more. That's fine

1:15:15particulate matter of size 2 and a half

1:15:17microns which includes multiple

1:15:19pollutants and the most important one is

1:15:20black carbon or soot then CO CO2 and so

1:15:23on. So when you talk about air quality

1:15:26impacts um I touched upon briefly in the

1:15:28last slide. Um there are multiple

1:15:29pollutants that affect human health. Um

1:15:32NOx, SO2, CO, PM2.2 and so on. There is

1:15:35also CO2 uh carbon dioxide. It's a

1:15:37climate pollutant or greenhouse gas

1:15:38pollutant. And I I won't discuss um

1:15:41greenhouse gas impacts a whole lot in

1:15:43this discussion only because I want to

1:15:44focus on one aspect a lot more which is

1:15:46air pollutant that affect human health.

1:15:48But CO2 is a climate pollutant. It has

1:15:49long-term impacts and that's not

1:15:51something to be discarded easily. Okay.

1:15:54Then the duration of genital runs and

1:15:55hence exposure have different impacts.

1:15:57You have both short-term which is acute

1:15:59impact based upon exposure to something

1:16:01really bad over one or two hours or

1:16:03long-term chronic impacts where you

1:16:05expose something over days or a year

1:16:07depending on the specific pollutant. The

1:16:09US EPA regulates six different air

1:16:12pollutants through what are called as

1:16:13national ambient air quality standards

1:16:15where they have very rigorous science

1:16:17which supports how human health is

1:16:18affected based upon exposure to air

1:16:20pollutant and based upon multiple

1:16:22epidemiological studies, risk assessment

1:16:24studies and so on. They have come up

1:16:25with uh what are called as knacks or

1:16:27standards that protect human health. Six

1:16:30are regulated by the EPA and the states

1:16:31in turn have to um uh come up with plans

1:16:34to make sure they meet these air

1:16:36pollutant standards for different

1:16:37pollutants. For example, NO2 one gas

1:16:40pollutant has both a short form or an

1:16:42acute 1 hour standard as well as a

1:16:45chronic or an annual national ambient

1:16:47air quality standard. Similarly, PM2.5

1:16:49has both a daily standard, which is what

1:16:51you're exposed over a period of a day,

1:16:5324 hours, as well as over an annual or

1:16:55long-term period um annual next.

1:16:59So, what is PM2.5? You heard about this

1:17:01quite a bit um both earlier by Noah and

1:17:03myself and maybe in previous town

1:17:04council meetings. So, PM2.5 is one of

1:17:07the most complex air pollutant you can

1:17:09think of. Okay. In terms of ename health

1:17:11risk, this is the number one killer of

1:17:13people in the world with about 8 million

1:17:16people dying prematurely every year

1:17:18because of exposure to PM2.5. That's a

1:17:20number just broadly speaking um uh

1:17:22globally within the US we lose about

1:17:25100,000 people every year approximately

1:17:27every year due to exposure to this

1:17:29particular political PM2.5. Okay. Now

1:17:31what is PM2.5? Look at the picture on

1:17:33the right. If you look at this brownish,

1:17:35you know, irregular shape polygon, that

1:17:37is a picture of a beach sand. Typical

1:17:39size of beach sand, which is about 80

1:17:41microns or um in size. Then the one next

1:17:44to it is a typical human hair strand, a

1:17:48single strand of human hair. That's

1:17:49about 50 to 70 microns. That's a

1:17:51thickness of a single human hair. And

1:17:53then within a single human hair strand,

1:17:54you can think of putting six different

1:17:56balls. Those are blue color balls on the

1:17:59extreme right there. And those are

1:18:01particles called as PM10 or fine

1:18:03particulate matter of size 10 microns.

1:18:06So these were the ones that people are

1:18:07thinking about 20 25 years ago as big

1:18:10particles that affect air pollution. But

1:18:11right now that is not much of an issue

1:18:13right now. But if you take one PM10 ball

1:18:15the blue ball within one PM10 you can

1:18:18fit four small red balls. That's your

1:18:21PM2.5. So that's the smallest part you

1:18:23can think of in terms of this particular

1:18:25airplane impact. So each orange or red

1:18:29ball that you see is your PM2.5. That's

1:18:31the size that you can think of. Um this

1:18:34is a well-known adverse air pollutant

1:18:36for human health and it's being

1:18:38regulated both in the US and by World

1:18:40Health Organization. The next level of

1:18:42um PM 2.5 or PM is what is called ultra

1:18:46fine particles. These are particles even

1:18:48smaller less than 100 microns. Right? Um

1:18:51so we won't talk about them for now. So

1:18:52we'll just focus on PM 2.5. But I will

1:18:55say there's emerging evidence where

1:18:56ultrafine particles also cause adverse

1:18:58human health. Okay, for sake of this

1:19:00discussion, PM2.5 is both a solid as

1:19:04well as a liquid particle in the air.

1:19:05They measure about 2 and a half

1:19:06micrometers or microns uh in diameter.

1:19:09These includes particles both that are

1:19:11directly emitted from any combustion,

1:19:12right? You have a combustion process,

1:19:13these are directly emitted as well as

1:19:15they get resuspended from the road dust.

1:19:17Right? When you have a car go by, you

1:19:19have dust there and then you have the

1:19:20resuspension goes on. that's also um uh

1:19:22directly um emitted. Then you also have

1:19:25PM that is formed from conversion of

1:19:27gases such as sulfur dioxide, nitrogen

1:19:30dioxide, ammonia, vol organic component

1:19:31so on. These go through atmospheric

1:19:33process physical as well as chemical

1:19:35conversion and they form another

1:19:37component of PM2.5 and we put all these

1:19:40together and that's your total PM which

1:19:42is both directly emitted particle as

1:19:44well as something that gets converted

1:19:45from gases through atmospheric process

1:19:47and they come in many sizes and shapes

1:19:49and composed of hundreds of chemicals

1:19:51because just lot of different particles

1:19:52right but given the small size the

1:19:55important part here is they can

1:19:56penetrate deep into the body mostly

1:19:58through inhalation right um and then

1:20:00based upon the size they go into the

1:20:02lungs and then as they get smaller they

1:20:05can easily break what is called the

1:20:06bloodb brain barrier. Once they go to

1:20:08the brain you have all kinds of effects

1:20:10on the human body. Okay, that's your

1:20:11PM2.5 sort of um primer. So what are the

1:20:15impacts of PM2.5? They have both acute

1:20:17as well as chronic exposure effects.

1:20:20Acute you have respiratory irritation,

1:20:21asthma and COPD um uh flare-ups,

1:20:24cardiovascular stress and so on. So

1:20:26basically they increase the risk of

1:20:27non-fatal heart attacks and irregular

1:20:28heartbeats. Um from chronic or long-term

1:20:31exposure which is over a year you get

1:20:33both heart and brain damage lung

1:20:34conditions gets um worse cancer risk and

1:20:37then systemic illness such as diabetes

1:20:40adverse birth outcomes and so on. So

1:20:41basically the emerging literature says

1:20:43there are more and more adverse health

1:20:45impacts that people are facing based

1:20:47upon exposure to PM 2.5 um including um

1:20:51the brain related issues such as

1:20:53dementia and sometimes decreased

1:20:55capacity to learn in children who live

1:20:57next to schools okay when you have high

1:20:59traffic um uh um emissions associated

1:21:01with the schools and then you have

1:21:03children their capacity to learn goes

1:21:05down and the documented studies that

1:21:07look into that so who is at risk to PM2

1:21:09exposure you have children and the

1:21:10elderly people with pre-existing health

1:21:12conditions like asthma or heart disease.

1:21:14Healthy individuals are also at risk

1:21:16especially with long-term exposure.

1:21:18Okay, so that's the population you want

1:21:19to think about. Then a little bit on NO2

1:21:22another gas pollution I want to think

1:21:23about only because this is also

1:21:24important from a diesel combustion point

1:21:26of view from a data center as well as

1:21:28from natural gas. Short-term effects um

1:21:30airway irritation, asthma flare up, lung

1:21:32function drops when you get more

1:21:33exposure to um you know to a short time.

1:21:36Uh long-term effects, you have new

1:21:38incense of asthma. healthy individuals

1:21:40start to get asthma, lung damage,

1:21:42chronic disease, um ongoing exposure to

1:21:44chronic lung disease, heart problem and

1:21:45so on. The single most important effect

1:21:48of NO2 is asthma. Okay, both in children

1:21:50and adults. Uh people who don't have

1:21:52asthma get asthma um incidences. People

1:21:55who do asthma have asthma exacerbation

1:21:57based upon exposure to NO2. Okay. So,

1:21:59children and older adults are most

1:22:00suscept susceptible to these uh

1:22:02pollutants and people with asthma like I

1:22:04said they face severe reactions even at

1:22:06low pollutant levels.

1:22:08So one study which is being

1:22:10peer-reviewed um I like to always focus

1:22:12on peer-reviewed studies but there's a

1:22:14study that's being peer-reviewed right

1:22:15now which said that based upon growth of

1:22:17data centers in the US they will cause

1:22:20about 600 asthma symptom cases new

1:22:22incident of asthma as well as about 1300

1:22:25premature deaths in 2028 and the total

1:22:27public health costs can exceed $20

1:22:29billion um based upon um uh value they

1:22:33put to human life as well as the medical

1:22:34costs associated with adverse health

1:22:36impacts. Um this particular impact this

1:22:39particular number seems to be an

1:22:40increase over 200% from 2023 when they

1:22:42last did this when the stationary while

1:22:45the stationary fuel combustion related

1:22:46health cost will increase only 17%.

1:22:48Think about that. So stationary fuel

1:22:51source combustion that's like the power

1:22:52plants and other industrial sources they

1:22:54increase health risk only by 17%. But

1:22:57the growth in data centers would be

1:22:58200%. It's about you know 12 times

1:23:00higher number than the 17% from um

1:23:03stationary source fuel combustion. Okay

1:23:05this is not uniform. I have a map coming

1:23:07up on the next slide which basically

1:23:08says well that's a broad big picture

1:23:09number across the country. The highest

1:23:11county level cost is seven times the

1:23:13national average means the effects are

1:23:16not uniform across the country. There

1:23:18are some counties some people are more

1:23:20vulnerable to this than others based

1:23:21upon either their health conditions or

1:23:23where they live based upon other social

1:23:24demographic and so on. And this highest

1:23:26county level cost is 200 times the

1:23:29lowest country call. So it's not uniform

1:23:31and the point I want to make is these

1:23:32are broad national numbers but there's

1:23:34quite a bit of spread that one needs to

1:23:36think about in this set of numbers. So

1:23:38this is the breakdown of that. So 23 on

1:23:40the left in the in the bar charts and

1:23:42then the uh green and orangish color

1:23:44show how the low and high range can grow

1:23:47based upon this projected increase in

1:23:49emissions from data centers. The middle

1:23:51um map of the US shows the per household

1:23:54health cost. You can see the

1:23:55northeastern US roughly the uh what is

1:23:57called the Ohio River Valley. We have

1:23:59bunch of power plants there. But also

1:24:00based upon this particular study, they

1:24:02showed the per household health cost can

1:24:04be pretty high in that part of the

1:24:05country. And the top 10 counties are on

1:24:07listed on the right there in terms of

1:24:09where you'll have the most impacts. And

1:24:11look at the um 10 counties there. Seven

1:24:15or eight or those are in West Virginia.

1:24:17Okay. Why West Virginia? Anybody wants

1:24:20to guess?

1:24:21>> Cold.

1:24:22>> Coal. But also the demographics of

1:24:25people there in general you have a lot

1:24:26poorer people who have more you know

1:24:29baseline health insurance not being good

1:24:30when they're exposed to more air

1:24:32pollution you have more health insurance

1:24:33there right so that's that's why popping

1:24:35up like that so so the per household

1:24:37health cost basically the point is to

1:24:39show that vulnerable populations may

1:24:41have more exposures and hence more

1:24:42health risk right and that this this

1:24:44analysis clearly shows that so the per

1:24:46household cost varies from about 300 um

1:24:49uh from 245 to 332 for the top 10

1:24:51counties but that's the spread that you

1:24:53But this could be about 200 times worse

1:24:55than the lowest county.

1:24:59Okay. Now, let's switch to Apex. Okay.

1:25:01And for this, um, I'll start with a map

1:25:04that shows the Southern Environment Law

1:25:05Center to see what they saw as the

1:25:08existing proposed data centers for the

1:25:10southeastern US, basically the region

1:25:11that they focus on. That's the map on

1:25:13the left. On the right is a little bit

1:25:14zoom into our own part of the state. Um

1:25:16you see the new hill digital campus was

1:25:19proposed and then withdrawn in the lower

1:25:21left corner of the right hand map. You

1:25:23see the other data centers in the rest

1:25:24of this um region.

1:25:27So what I did um in the last couple of

1:25:30weeks was to do some additional

1:25:32modeling. Uh I'll show some results. Um

1:25:34I have a bunch of charts coming up next.

1:25:36But what I did was to think about how

1:25:38the data center may operate. Um you

1:25:40heard quite a bit from Noah earlier talk

1:25:42about base load and backup Z and so on.

1:25:44So basically um my version of this is to

1:25:46talk about base load option one where

1:25:48the data centers draw the power from the

1:25:50grid and in this case Duke energy if if

1:25:52that's a facility because that's what we

1:25:54heard that the developer had an

1:25:56agreement with Duke at some point to

1:25:58have additional generation to meet the

1:26:00demand and with this base option one u

1:26:03power being drawn from the grid you

1:26:04address both routine as well as a peak

1:26:06demand during extreme weather. Some of

1:26:09us may recall the early part of this

1:26:11year when there was a cold snap and

1:26:13anyone recalls a message on your TV

1:26:16phones and so on about conserve energy

1:26:18conserve energy right that was a cold

1:26:20snap we had that came from u the

1:26:22utilities right and the state so when

1:26:25you have uh a data center like this

1:26:26running with the base um load being

1:26:28drawn from the grid you may have

1:26:29additional demand going on okay peak

1:26:32demand in both extreme weather hot and

1:26:33cold u base load option two is where we

1:26:36have the so-called behind the meter We

1:26:38talked a little bit earlier with Noah.

1:26:40So where the proposed data center

1:26:42generates the power onsite behind the

1:26:44meter but in that case they have to

1:26:46develop new infrastructure on site.

1:26:48Okay, which doesn't exist as of now. And

1:26:50the third one is where we have the

1:26:51backup generation onsite which is really

1:26:54backup the base floor coming either from

1:26:55options one or two. The backup is what I

1:26:58think is of concern to a lot of people.

1:26:59Not to mention that the first two

1:27:01options are not of concern.

1:27:03Base floor option one. Okay. This will

1:27:05lead to excess emissions and potential

1:27:07violation of the national ambient air

1:27:09quality standards in the region. Okay.

1:27:11So basically when you have uh Duke

1:27:13committing to take on more um generation

1:27:15of power it has to come at some cost

1:27:17right what are they burning so they have

1:27:19to do something to bring on uh more

1:27:21consumption um and then that comes at

1:27:23some cost because right now no Duke

1:27:26facility other than shareh nuclear and I

1:27:28don't think nuclear is renewable either

1:27:30right so the point is um this will lead

1:27:32to excess emissions and potential

1:27:33violation of the standards and I want to

1:27:35talk a little bit about what this means

1:27:37for violation of air quality standards

1:27:39when a region is designated to be in

1:27:40violation of the national ambient air

1:27:42quality standards of pollutant. Okay. So

1:27:43you have a region like let's say the

1:27:45Raleigh Durham metro area if this is in

1:27:47violation of the standards for air

1:27:48pollution for a given pollutant this

1:27:50region will go through substantial

1:27:52adverse impacts from both an economic

1:27:54and public health point of view. Okay.

1:27:56Until measures are in place to reduce

1:27:58emissions from cultable sources where

1:28:00the state works in partnership with the

1:28:02facilities and the industries um and

1:28:03other sources in the region to see what

1:28:06needs to happen to bring the area into

1:28:08attainment of the standard. So that's

1:28:10when the state works with the local

1:28:11agencies to come up with what is called

1:28:12as a state implementation plan. So when

1:28:14you have excess emissions in the region

1:28:16based upon what Duke or another facility

1:28:18may be doing, you have excess emissions

1:28:19which may lead to violation of the

1:28:21standard based upon what they are doing.

1:28:23So many years ago um ozone was a key

1:28:25pollution of concern for North

1:28:26Carolina's urban areas including the

1:28:28triangle, the tri and the Charlotte.

1:28:29These three areas had significant ozone

1:28:31problems some time back. And then in the

1:28:34year 2002 a historic legislation was

1:28:36passed what is called as the clean smoke

1:28:38stacks act that was under partnership

1:28:40from the state utility commission uh and

1:28:42EMC environmental management commission

1:28:45I think uh they work with the uh

1:28:47facility at that time Duke energy and

1:28:49Carolina power at that time to pass a

1:28:51historic legislation where they had a

1:28:53mandate um to reduce NOx emissions

1:28:56significantly from the 14 coal burning

1:28:58power plant in the state uh both Knox

1:29:00and socks were reduced uh this was in

1:29:022002 too. And then based upon those

1:29:05projected emissions reductions, you

1:29:07know, we did a bunch of analysis and

1:29:08modeling. Um I was um um part of the

1:29:11analysis way back then and at that time

1:29:14we projected that the state will um

1:29:16attain the air quality standards for

1:29:17ozone in the year 2012. And guess what?

1:29:20In the year 2012, by Duke Energy

1:29:24reducing its power plants and putting a

1:29:25lot of controls on them over time for

1:29:27reduced oxen socks, the state's air did

1:29:29become clean by the year 2012. There's

1:29:32another study which also looked at the

1:29:33impacts of PM2.5 sulfate reductions and

1:29:36they also quantified that there were

1:29:38substantial health benefits in the

1:29:39state's residents due to reduction of um

1:29:42uh knocks and socks from these power

1:29:44plants and at that time they um

1:29:46estimated that about 1,700 premature

1:29:48debts are being awarded annually due to

1:29:50clean air in the state. Okay. So as a

1:29:52function of all of that um I'm happy to

1:29:54tell that the state today is an

1:29:55attainment of all national I mean air

1:29:57quality standard. This is from the

1:29:58NCDAQ. Um we are having clean air

1:30:01everywhere for all criteria pollutants.

1:30:03But again this is based upon a standard

1:30:05there's an emerging evidence that you

1:30:07know no nonzero value is good for us.

1:30:11Okay no non-zero value is good for us.

1:30:14So even though it's a standard which

1:30:15says you have adverse health impacts

1:30:17above a threshold even lower levels of

1:30:19concentration of pollutant you do have

1:30:20adverse health impacts.

1:30:23>> Now base floor option two. So this is

1:30:25where uh we have on-site generation

1:30:27using fossil fuels such as natural gas

1:30:28or other. But then when you have that

1:30:30you also have to think about what are

1:30:32the local environmental concerns as a

1:30:34function of on-site generation. Okay. Um

1:30:36there was a proposed um natural gas

1:30:39power plant just over 3,000 megawatt

1:30:42natural gas power plant uh in um one of

1:30:45the counties in Virginia just across the

1:30:46border from this um from North Carolina

1:30:49in the north. And this is Pennsylvania

1:30:51County. And based upon this proposal,

1:30:53they were going to generate um uh power

1:30:55by burning natural gas for this

1:30:57capacity. And they estimated about 326

1:30:59tons of excess PM emissions. And then

1:31:01based upon a study that was commissioned

1:31:03by the Southern Environment Law Center,

1:31:04they basically found that if you look at

1:31:06the map on the right, there'll be about

1:31:081.28 million people downwind of this

1:31:11facility who will have additional

1:31:12exposures. um about half of those

1:31:14impacts were in the state of Virginia

1:31:16and the rest of them find their way

1:31:18across the border into North Carolina

1:31:20and a very small percentage of South

1:31:21Carolina as well. And then they

1:31:22estimated that the total health impacts

1:31:24of this is about $31 million over one

1:31:26year or 625 million by the year 2040.

1:31:30Okay, that's what they estimated. The

1:31:31point of this is to show that even when

1:31:33you have on-site generation and when you

1:31:34use fossil fuel, you will have health

1:31:36impacts both in the near field and far

1:31:38field depending on the strength of the

1:31:40facility as well as how far the winds

1:31:41go. Okay.

1:31:43And just in terms of what these numbers

1:31:45mean, 326 tons um and $31 million. Think

1:31:48of every ton of PM emissions having

1:31:51about $100,000 in health costs. Just

1:31:54ballpark, right? One ton of PM emissions

1:31:57has about $100,000 in health costs. And

1:31:58you can do the scaling depending on how

1:32:00you want to look at the numbers.

1:32:03All right. New case studies. This is

1:32:04where uh uh I I looked at this

1:32:06particular um proposed facility and now

1:32:08withdrawn and then did a bunch of

1:32:10modeling to give um a sense of what

1:32:13these impacts may be for um the town and

1:32:16neighbors um here. So this withdrawn

1:32:19project um probably everybody in this

1:32:20room knows what this facility was all

1:32:22about. Uh located on 190 acres which is

1:32:24zone for industrial employment um

1:32:26located between US1 and old US1 just

1:32:28west of Sharon Harris. um they proposed

1:32:30uh to construct four 200,000 square foot

1:32:33buildings and it'll have diesel backup

1:32:34generator. That's what they proposed

1:32:36equipped with multiple emissions

1:32:37controls to meet this EPA tier 4

1:32:39standards. Okay. And these controls were

1:32:41supposed to um reduce emitted

1:32:43pollutants. That's what they proposed.

1:32:46So what I did was use an EPA dispersion

1:32:48model. This is called air mod. It's a

1:32:50regulatory model that EPA has. This what

1:32:52is used for local scale dispersion. Um

1:32:54it's a standard uh model that EPA has

1:32:56developed over the years. I've been

1:32:58involved with some of the development

1:32:59and application with EPA. Uh but

1:33:01basically what we did was to um estimate

1:33:03the emission dispersion from backup

1:33:04generation using air mod which calculate

1:33:06the dispersion of pollutants for

1:33:08research and regulatory purposes and the

1:33:10outputs are vared based upon the knack

1:33:12uh the national ambient air quality

1:33:13standard which is based upon the best

1:33:16scientific understanding of the adverse

1:33:17impacts of air pollution and determines

1:33:19whether an area is within attainment.

1:33:20Okay. So to do the modeling uh compile

1:33:24information based upon existing case

1:33:25studies at other data centers to see

1:33:27what are the type of interest they use

1:33:29what are the configuration they use and

1:33:30then um use a few different

1:33:32configurations um and then scale these

1:33:34emissions based upon the proposed um

1:33:36capacity here to align with the

1:33:38previously proposed new digital campus.

1:33:41So um I I ran three different um

1:33:43simulations here um two using diesel and

1:33:46the third using natural gas. I just call

1:33:48them as configurations one, two and

1:33:49three. Basically um each of the three

1:33:52configurations have very different stack

1:33:53configurations, different um number of u

1:33:56facilities, different number of

1:33:57generators, different uh stack diameter,

1:33:59stack height and so on. basically uh um

1:34:01picked another published study which is

1:34:03used for permitting elsewhere and use it

1:34:05to see what will happen if they use the

1:34:06exact one here because I don't think the

1:34:09developer who applied to the town of

1:34:11Epics gave a lot of these details right

1:34:12so we had to go find something else from

1:34:14outside because they didn't go to the

1:34:15permitting phase of this particular um

1:34:17issue um in each case the capacity of

1:34:19each was 3.1 megawatt then this is

1:34:22something I came across and I'm open for

1:34:23discussion here when a data center

1:34:25proposes um capacity of 300 megawatt um

1:34:28I've seen some evidence to show that

1:34:30they actually need to plan for

1:34:31redundancy of 600 megawatt. Okay. So to

1:34:35account for the potential redundancy um

1:34:37did the modeling two different ways. N

1:34:39is where I stick to 300. So you have 97

1:34:41generators each of 3.1. The 2N is

1:34:44basically where you're doubling the

1:34:45capacity and I have 184 generators each

1:34:48um generally 3.0 megawatt. Then um

1:34:51diesel 2 is a second configuration. Then

1:34:53natural gas 3 is a different

1:34:54configuration but using natural gas for

1:34:56um combustion.

1:34:58Um before we go into the specifics of

1:35:00this particular modeling for this

1:35:01facility I want to talk about how

1:35:03dispersion works. This is a notional

1:35:05plot um which shows how dispersion

1:35:07works. Top left shows two different

1:35:10locations of um power plants or points.

1:35:12I call them as points because you have a

1:35:14specific location a specific stack and

1:35:16then it has some stack parameters. It's

1:35:18burning something and then depending on

1:35:20how the winds blow, how the atmosphere

1:35:22works, how the dispersion works. Um you

1:35:25look at then the top left panel how the

1:35:27dispersion goes on a single hourly time

1:35:29scale based on how the mutology or the

1:35:31winds are blowing you see these two

1:35:33plumes like shooting out of those two

1:35:35particular points that's how point

1:35:36source dispersion works right and then

1:35:38on the right instead of two points if

1:35:40you have a low-level area source which

1:35:43is basically sort of a something spread

1:35:44over a um larger area lot of small

1:35:46sources instead of modeling each source

1:35:48separately just put a large area source

1:35:50over some area and then based upon the

1:35:52height it's usually a lower level source

1:35:54opposed a point source which is elevated

1:35:56right so um with a very similar strength

1:35:58on the right you have a hourly

1:36:00dispersion from a low level source on

1:36:01the right if you look closely you'll see

1:36:03this greenish polygon in the middle of

1:36:05this figure if I can sort that's that's

1:36:07the source definition and from there you

1:36:09have early dispersion then the same two

1:36:12source types point on the left and area

1:36:14on the right are shown as um annual

1:36:17dispersion in the bottom right you see

1:36:19what happens over a period of one year

1:36:20and how the dispersion works after you

1:36:22do the dispersion based on ambient

1:36:24methology and the source is continuously

1:36:26emitting some pollutant right and and

1:36:28see how it works over a period of one

1:36:29year and you see how the plume patterns

1:36:31change over what happens over one year

1:36:33versus 1 hour I want to show this to

1:36:35make a point that depending on remember

1:36:37I earlier talked about how long the

1:36:39facility is running the running

1:36:41continuous start stop 1 hour 2 hour 4

1:36:44hours later and so on so I want to think

1:36:45about how the dispersion works at

1:36:47different time scales and where the

1:36:49plumes may go so on the left if you look

1:36:51at the plume center line the Nox it can

1:36:53be about 1 ppb at the plume center line

1:36:55whereas on an annual time scale is

1:36:57only.1 ppbpb. So you see how the numbers

1:37:00go down depending on what is the

1:37:02temporal metric you look at. On the

1:37:04right you have the same numbers for the

1:37:06area source which is a low-level source

1:37:07which is a irregular shaped green color

1:37:09polygon I showed. So the point again is

1:37:11show that you know dispersion works

1:37:12differently for different source type

1:37:14and it's very important to know how a

1:37:16proposed data center configures their

1:37:17sources.

1:37:19Okay. All right. So here's my study

1:37:22domain. Um uh it's basically the center

1:37:24of this concentric circles is where the

1:37:27proposed new campus was. And if you look

1:37:30closer um um you'll see those four

1:37:32different dots that represent the four

1:37:33buildings. And what I did here was to

1:37:35show concentric circles of 1 kilometer

1:37:37apart starting from the proposed

1:37:39facility and then um so on the left

1:37:41figure is a 10 km radius going out from

1:37:43the proposed facility in New Hill. on

1:37:45the right is a 5 km. Same thing but just

1:37:47five kilometers. And what we did was in

1:37:49this modeling to put receptors in every

1:37:51one of those circles at different

1:37:52distances um from the center and then

1:37:55see how dispersion works from the

1:37:56facility and see where the impacts may

1:37:58be. Now where we are right now if you

1:38:00look at this so on the left panel this

1:38:02is uh that's the proposed facility new

1:38:04hill um which was proposed and withdrawn

1:38:06and we are roughly this building is

1:38:09about here where the exis just about 12

1:38:12km or so from where the facility was.

1:38:14Okay, that's our current location right

1:38:16now where X of the word apex is. Okay,

1:38:19so just to give you the coordinates. So

1:38:20what I'll do next is show a series of

1:38:22colorful maps which show how the

1:38:24pollution works from the proposed

1:38:26facility on the left at 10 km at those

1:38:29concentric 1 km incremental circles.

1:38:33Okay. So now we'll look at NOX

1:38:35dispersion. So I'll focus on two

1:38:36different NOX and PM2.5. And then NOX

1:38:39goes on to form NO2 for which we have

1:38:41the standard nitrogen dioxide. So um

1:38:44these three different figures show how

1:38:46the dispersion worked for the three

1:38:48configurations I talked about. DC1, DC2,

1:38:51DC3. Uh DC 1 and two burn diesel, the

1:38:54NC3 burns natural gas and looking at a

1:38:57um plume based upon running the

1:39:00dispersion for one full year, but then

1:39:02pick out for over a period of one year

1:39:05what is a single worst case hour. Okay,

1:39:09looking at the most extreme case you may

1:39:10have and what is the single most extreme

1:39:12hour to see what may happen based upon

1:39:14the exposure to based upon this

1:39:16particular um dispersion coming from the

1:39:18source. So if you look at DC1 and DC2 um

1:39:20where I'm drawing these red holes you

1:39:22see the max goes from about 196 um uh on

1:39:26the left to about 179 in the middle

1:39:28that's in parts per billion and the

1:39:30standard is 100 parts per billion. So

1:39:32you see all nonzero values which are

1:39:35shown are values above 100. Right? If

1:39:37you look at the legend box in the

1:39:38extreme right, the blue is starts

1:39:40starting at 100 and we go up to about

1:39:41little over 180 uh in red. So you see

1:39:44values which are you know roughly in the

1:39:46southern quadrant of where the source

1:39:47was and these values are at 196 at a

1:39:50maximum or 179 and the national ambient

1:39:53air quality standard for you know KNOX

1:39:55is 100 right so I'm showing always above

1:39:57100 then now if you look at uh what will

1:40:00happen if you double the capacity

1:40:02remember I talked about the redundancy

1:40:03that um some facilities talk about so if

1:40:06you want to have 300 megawatt you

1:40:07actually want to have capacity for 600

1:40:09megawatt so now you have 184 um number

1:40:12of generators each at 3.1. Now the

1:40:14entire 10 km radius you see is now

1:40:16colored um in a nonzero value and

1:40:19they're all over 100. I can see where

1:40:21the um concentrations are. Look at the

1:40:23maxima um 389 on the left and then um

1:40:26359 on the right. So configuration two

1:40:29drops the maxima a little bit but not a

1:40:31whole lot. But in natural gas on the

1:40:33right you don't see a whole lot of

1:40:34action happening based upon this

1:40:36configuration.

1:40:37PM 2.5 that's the pollut we talked about

1:40:39quite a bit or fine particulate matter.

1:40:42um single hourly maxima. Um we have 15

1:40:45on the left with configuration one and

1:40:47then 13 in configuration two. And again

1:40:49you see that while the hot spots you

1:40:52know the reddish orange are in the um

1:40:54southern uh half of the um circle you do

1:40:57see this yellowish green coming u away

1:40:59from the facility towards east and

1:41:01north.

1:41:04Um this is comparing um uh 2x. Now last

1:41:07one was single x or 300 megawatt. was

1:41:09600 megawatt and now the um maxima have

1:41:12gone up even higher than what you saw

1:41:13last time to 29.99 in DC1 and 27.64 in

1:41:18DC2.

1:41:19Now what I did was um since facilities

1:41:23don't run for a full year, they run for

1:41:24a few hours, right? Based upon the

1:41:26backup generation model. So what I did

1:41:28was configure the model to say I'll run

1:41:31for four hours at a stretch, right? and

1:41:33then run for multiple four hours over

1:41:35the entire year and find the maximum of

1:41:37those multiple four hours. So

1:41:39potentially what you're looking at is if

1:41:41the facility would run only for four

1:41:42hours during the year, what is the worst

1:41:44case based upon dispersion conditions,

1:41:46right? So there are limits that

1:41:48different um um towns, counties have in

1:41:51terms of how long these generas can run.

1:41:52Um for example, south coast, one of the

1:41:54basins there in California have a limit

1:41:56of 200 hours as a maximum they can run

1:41:59for a year. Um I don't remember if the

1:42:01town of Apex has one, right? But the

1:42:03state I believe has 50 um hours they can

1:42:05run a backup generator. So we need to

1:42:07see what are the appropriate um numbers

1:42:09we want to use for backup generation.

1:42:11How long they can run. In this case I'm

1:42:13running for 4 hours and see what is the

1:42:14worst four 4hour case. And now the

1:42:16maxima go from uh 119 on the left to

1:42:19109. So still you get values that are

1:42:21exceeding this threshold of 100. Okay.

1:42:25Now PM 2.5 again the same for our uh

1:42:28discussion. Um 9 on the left and then

1:42:308.43 43 um in the middle.

1:42:33Now what I did was okay you have all

1:42:35these uh spatial plots look at different

1:42:36numbers. Now I say okay let me just put

1:42:38them all together and see okay across

1:42:40all the receptors that I modeled okay

1:42:42how does it compare based upon

1:42:43threshold. So this is just if you look

1:42:45at the first one these are called as box

1:42:47and whisker plots. You just look at the

1:42:48distribution of the entire data. The

1:42:50first one DC1 you do see the maxima you

1:42:53know exceeding the red line. Red line is

1:42:54a threshold of 100 um uh in that case uh

1:42:57where they exceed this uh a little bit

1:42:58over. Then the middle one is a

1:43:00configuration two which is dropping the

1:43:02overall distribution slightly smaller.

1:43:03So you get a sense of what is the

1:43:05general range of numbers that each of

1:43:06the configuration do uh to the NO2

1:43:09maximum.

1:43:10Now on the right is where you go with

1:43:12the redundant capacity where you're

1:43:14doubling the capacity from 300 megawatt

1:43:16to 600 megawatt and again you have

1:43:18significant number of locations which

1:43:19are exceeding the threshold. You see the

1:43:21whole number of dots here um going up to

1:43:24194 and then up to 179 in the middle.

1:43:28Now this is a comparison of 1 hour

1:43:30versus the 4 hour um average maxima and

1:43:32again in this case you see 1 hour on the

1:43:34left 98 and 89 the right 71 and 77

1:43:40and then this is the last one for PM 2.5

1:43:42to see how the PM varies. So basically

1:43:43what I did was the point of these

1:43:45analysis to show that depending on the

1:43:48source configuration depending on the

1:43:50time metric you're looking at 1 hour 4

1:43:52hour annual you do get different impacts

1:43:54okay and from health point of view you

1:43:55have to think about what is it really

1:43:56important for what you're looking at and

1:43:58see how this may affect human output

1:44:02so moving on um I want to alert um the

1:44:05board um to um two different activities

1:44:08happening at the state level one is the

1:44:10North Carolina's um artificial

1:44:12intelligence leadership council. This is

1:44:14convened by um um our governor and under

1:44:17the DIT. uh they have a they put out a

1:44:20strategic road map recently and they

1:44:22focus on three different broad

1:44:24objectives to protect uh which focus on

1:44:27safeguarding North Carolina residents

1:44:28responsible AI use in strong risk

1:44:30mitigation practices and then to prepare

1:44:32uh where you strengthen the AI literacy

1:44:34and readiness of students workers and

1:44:35communities and then transform the

1:44:36outlines of how the state can modernize

1:44:39leverage AI but under protect the first

1:44:41goal they have multiple goals in each of

1:44:43these objectives the first goal is to

1:44:45protect residents from AI harms so it is

1:44:47called out in the strategic roadmap map

1:44:49um but we need to think about how we go

1:44:50about the protection part of what this

1:44:52um road map lays out. The second

1:44:54activity at the state level is this

1:44:56energy policy task force um they put an

1:44:57interim report um earlier this year.

1:44:59Some of you may have seen this.

1:45:01Basically what they said is significant

1:45:03load growth is coming to North Carolina

1:45:05associated with opportunity and risks.

1:45:07The figure on the shows um shows the an

1:45:09the figure on the right shows the annual

1:45:11CO2 emissions from electricity

1:45:13generation North Carolina. we see this

1:45:14very nice drop off starting from 2005

1:45:172023 or so but then they say after 23 uh

1:45:21for the next few years you'll have a

1:45:22leveling off of the CO2 emission because

1:45:24that's what the projections for the

1:45:26growth is and then after that it'll

1:45:28start to drop after 2030 or so. So while

1:45:30the statewide increase is um was only 7%

1:45:33earlier now we expect a growth of 16 to

1:45:3660% over the next two decades. Okay,

1:45:37that's what is being um looked at by the

1:45:40task force and the large fraction of

1:45:42this particular load growth is due to

1:45:43data centers and other large loads.

1:45:46Okay, so to wrap up um environmental

1:45:49public health impacts of data centers is

1:45:50substantial. Okay, I think I I've shown

1:45:52quite a bit of evidence for that. Um

1:45:54AI's power demand is new territory for

1:45:56the town, for the state and for the

1:45:58nation. Okay, permitting decisions for

1:46:00DCs need to account for both the

1:46:01environmental and public health costs

1:46:03and these are real costs of society.

1:46:04These are not fictional statistics

1:46:06numbers. These are real costs. People

1:46:08have adverse costs based upon these

1:46:09exposures. And the configuration of data

1:46:11centers can vary widely. I demonstrated

1:46:14that with my tests and moving power

1:46:16generation to meet excess demand from

1:46:17grid to local on-site may have its own

1:46:19un intended adverse impacts. And both

1:46:22the state and the town such as Apex, we

1:46:24need to review the permit application

1:46:26carefully to assess impacts and minimize

1:46:27public health impacts to protect

1:46:29vulnerable population. Okay. One needs

1:46:31to quantify all known adverse effects

1:46:34and have appropriate plans for

1:46:35mitigation.

1:46:37And how do we promote sustainable growth

1:46:39of AI industry while meeting energy

1:46:41demand? Um review the progress being

1:46:42made at the state level. Uh the two um

1:46:45resources I touched upon the energy

1:46:46policy task force and the AI leadership

1:46:48council for their recommendations and

1:46:50other legislative updates. But at the

1:46:52end I think we want to explore how to

1:46:54promote this industry for its alleged

1:46:56benefits to society such as jobs, taxes,

1:46:58etc. without compromising both the

1:47:00environment and the public health of its

1:47:02citizens. With that, here is my take on

1:47:06a data center.

1:47:09>> Okay. Thank you.

1:47:11Happy to answer questions.

1:47:15>> I want to first thank both our speakers.

1:47:17You know, a lot of town councils don't

1:47:19have an environmental advisory board. So

1:47:23um you know while many are in

1:47:25moratoriums right now, they they might

1:47:28want to consider you know

1:47:31having environmental advisory boards and

1:47:33having speakers like you and Noah here

1:47:36um to speak. you know, you and I

1:47:38actually spoke a little bit last year

1:47:40when this one was proposed

1:47:43and, you know, concerns of things from

1:47:47taking environmental studies uh in

1:47:49college, thinking of um Love Canal and

1:47:52and what that does to an area. I mean,

1:47:54and that's very different than obviously

1:47:56a data center, but it does um cause a

1:47:59lot of the same health issues. Um, I was

1:48:02in Marvel, North Carolina this past

1:48:04weekend and that is actually where one

1:48:06of the largest in North Carolina is. Um,

1:48:11I got to drive up to it and hear the

1:48:13noise that it actually produced and

1:48:15actually have a connection to somebody

1:48:17there that's dealing with both cancer

1:48:19and asthma. Um, is it from the data

1:48:24center or is it just uh issues that were

1:48:27already there with this person? that I

1:48:29don't know but you did um raise a

1:48:33question that I have for you. Um is

1:48:36there any scenario under which a hypers

1:48:38scale data center will not have any

1:48:40adverse health effects uh to residents

1:48:43in the town of Apex?

1:48:50>> Not with the current technologies being

1:48:51proposed and used.

1:48:54>> Okay. Uh clearly not. um you you saw

1:48:57that based upon configurations the most

1:49:00popular fuel types are diesel, natural

1:49:02gas that's what there is some evidence

1:49:04of using fuel cell but again I don't

1:49:07think it's um scalable we need to see um

1:49:10what will come out of this um and then

1:49:12depending on um the specific size I

1:49:15don't see it happening right now but I

1:49:18think I'm always counting on human

1:49:20ingenuity

1:49:22which means sometime soon some smart

1:49:25mind somewhere will figure out what is

1:49:27the best way to do this and do exactly

1:49:29what you just asked.

1:49:30>> Okay. But doesn't exist right now.

1:49:34>> Thank you.

1:49:34>> You talked about EPA

1:49:37air quality standards and stuff. Um,

1:49:41>> do you know of any communities who

1:49:44impose stricter standards than what the

1:49:47EPA has or are they limited by what the

1:49:50EPA requires?

1:49:51>> The state of California. Okay.

1:49:54>> They are ahead.

1:49:55VP,

1:49:56>> right? Uh and some towns locally have

1:49:59their own they call them air basins or

1:50:01air districts and they have much

1:50:03stricter um they call state standards or

1:50:06air basin standards which are usually

1:50:07more stringent than EPA and they have

1:50:09been the leader for a large degree the

1:50:11past several decades California right

1:50:14but California is California and EPA is

1:50:16EPA right I

1:50:17>> I was I was thinking more lo local

1:50:19communities like Apex or

1:50:22>> I'm not aware of a local

1:50:25community like Apex having something

1:50:26most important.

1:50:29>> Well, I mean air quality, I mean

1:50:33it's hard to build a wall around the

1:50:35town and keep the air quality,

1:50:38>> right? But but there have been there

1:50:39have been instances where a specific

1:50:42town has come up with stringent

1:50:44standards for some sector of the fossil

1:50:48fuel. For example, building emissions,

1:50:49right? Uh we use heating and cooling for

1:50:51buildings, right? So some cities, towns

1:50:54of the country have come up with

1:50:56standards which are you know have for

1:50:57permitting a building to have some

1:51:00heating cooling. They come with more

1:51:02progressive standards to see how the

1:51:04town can build before you give a permit

1:51:05to a developer to have something.

1:51:07>> These are standards you have to follow

1:51:08and some there are examples I don't have

1:51:10it handy right now. I'm happy to share

1:51:12those examples of how a local town a

1:51:14local municipality becomes something

1:51:16more stringent for developing building

1:51:17standards.

1:51:20>> That that would be helpful.

1:51:24So looking at your maps, it kind of

1:51:26looked like the most effective

1:51:28protection that an individual could do

1:51:32for themselves is just to be as far away

1:51:34as possible from one a facility that

1:51:37emits um as you modeled.

1:51:40>> Yeah.

1:51:42>> Um

1:51:43are there any other industries that are

1:51:46comparable that you sort of see the same

1:51:48pattern in emissions? Because this is

1:51:51sort of like a new thing. Is there

1:51:52anything else that you could say? Well,

1:51:55this is sort of an apple to an apple.

1:51:58Would it be like a coal fired power

1:52:00plant or are you talking like

1:52:02>> Yeah. Yeah. Yeah. Coal fire power

1:52:04plants. They they are um slightly

1:52:06different beast in the sense that these

1:52:07are usually tall one or two stacks,

1:52:10right? I talked about the clean smoke

1:52:12stacks act of 2002 when we had 14 coal

1:52:15burning power plants. But a power plant

1:52:17is a facility which are multiple stacks

1:52:19but not 100 to 300. They have two,

1:52:21three, four, five stacks at the most in

1:52:23the facility. These are really tall,

1:52:25right? Um 100 ft or more and they put

1:52:28out emissions at at that stack based

1:52:30upon what they're combusting and then

1:52:32based upon the height. The reason they

1:52:33are tall is so that they push the

1:52:36pollution up there and then they get

1:52:37dispersed further down. You saw my maps,

1:52:39right? So that's how they get dispersed,

1:52:40right? But at a low level, this is a

1:52:43unique situation with these data

1:52:45centers, right? Because you have

1:52:46multiple small genators all in a very

1:52:49small space,

1:52:52right? Multiple. Think of a 300 megawatt

1:52:55going to have 100 different genative

1:52:56each generating 3 megawatt, right? So

1:52:59this is unique.

1:53:09>> Thank you so much. Thank you. Appreciate

1:53:10it.

1:53:10>> Very informative.

1:53:11>> Very help.

1:53:18>> So,

1:53:20anything else?

1:53:22>> No, just you know, the state of North

1:53:24Carolina, the way our laws and what

1:53:26we're allowed to regulate is passed

1:53:27down, has to be conferred or allowed by

1:53:29the state itself. And so, some examples

1:53:31might be really good to look at to see

1:53:32what other cities around the country are

1:53:34doing, but we not may not be applicable

1:53:36to North Carolina as well.

1:53:38>> It would have to be applicable to us for

1:53:40sure

1:53:41>> if we can do anything.

1:53:45>> So it's sort of becoming clear that uh

1:53:49this sort of whole on-site generation

1:53:52like this is coming up later. This is

1:53:54not a discussion we were having when the

1:53:57new health thing um was proposed.

1:54:01>> It was simply backup or emergency

1:54:03generation

1:54:04they were proposing at the time. So

1:54:07early on you sort of talked about

1:54:10different flavors of data centers and

1:54:13frontiers, but I'm wondering does a

1:54:15component have of that have to be are

1:54:18you do you plan on on-site generation

1:54:21and what does that look like? What what

1:54:23fuels are you talking about here? Right.

1:54:26Um versus just strictly emergency aside

1:54:30from defining what exactly is an

1:54:32emergency.

1:54:33>> Wow. Yeah.

1:54:37I think it's it's it's going to be

1:54:39difficult

1:54:43to try to insist on on

1:54:48more of a renewable

1:54:50uh generation system unless you're

1:54:53talking pure batteries.

1:54:55uh simply because of of

1:54:59the density of of of the apex area, I

1:55:02don't think you're going to be able to

1:55:03find enough

1:55:07land mass uh to come up with something

1:55:10uh

1:55:12to avoid either diesel or or some sort

1:55:15of natural gas, propane, LG

1:55:19types of uh power.

1:55:24But then does it become is it an

1:55:27appropriate thing given the development

1:55:30pattern to even say yeah sure you can

1:55:34have on-site generation or should that

1:55:36be like a just a non-starter

1:55:40>> that may be something you're not able to

1:55:42regulate the state who writes the

1:55:44permits for those things would regulate

1:55:46it. So I think um all this good

1:55:48information, something we could consider

1:55:50and think about maybe more at a town

1:55:51level would be um the quality of

1:55:54standards for the generators themselves,

1:55:56the type of fuel that potentially could

1:55:58be used, the distribution of data

1:56:01centers, whether one or multiple, where

1:56:03they're located, um how they set up

1:56:05their sites, which we've kind of talked

1:56:06about as land use goes, where they're

1:56:08located. Um, and I mean that proposed

1:56:11data center, we're going to keep the

1:56:13generators within the confines of the

1:56:15building. I mean, the buildings were

1:56:17going to be outside of where the

1:56:18generators were located.

1:56:21And that's one thing I did like about

1:56:23that proposal

1:56:26that would that would help with so like

1:56:28they built them internal to the the

1:56:30primary structures that helped with the

1:56:32noise aspect of it.

1:56:33Now, I don't know if you're also

1:56:35creating a more concentration of the

1:56:37fumes going up and that

1:56:39>> probably would. Yeah. I mean, you're

1:56:42creating a a tunnel going up.

1:56:47>> That makes it really important to have a

1:56:50distance threshold from sensitive,

1:56:53>> right,

1:56:53>> places like outdoor recreation, schools,

1:56:56I would imagine, the hospital, um,

1:56:59places like that.

1:57:11I mean, one thing that I've been a

1:57:13little worried about on on these rules,

1:57:16the envir environmental issues and and

1:57:19the air quality and and water quality

1:57:21are are

1:57:24paramount, I think, but I think the town

1:57:27itself has a lot of constraints on what

1:57:30it can and cannot

1:57:32I think I think that's why it's

1:57:33important that we we've been talking

1:57:35about this for so long. We'll continue

1:57:36to talk about this and we've had experts

1:57:37show up to talk about this is because

1:57:39there are a lot of things that we're not

1:57:41going to be able to do from a regulatory

1:57:42standpoint, but we can have the

1:57:44conversation. We can make our elected

1:57:46officials aware of it. We can let our

1:57:48residents know be aware of it. And then

1:57:49there are some of these things that we

1:57:50could still try and be see if we could

1:57:52work into as zoning conditions given

1:57:54that potentially if if council decides

1:57:56to move forward data in whatever form,

1:57:58it'll probably be regulated to a heavy

1:58:00industrial area which does not exist on

1:58:02the on the map right now.

1:58:03>> And so that would be a new resoning. And

1:58:05so we would have we could come up with a

1:58:08list of proposed suggested zoning

1:58:10conditions for data center specifically

1:58:13to help address some of these things

1:58:15where we couldn't maybe regulate that

1:58:16directly through the UDO amendments

1:58:18themselves.

1:58:20But that again, we'll have to we'll have

1:58:21to come up with those ideas.

1:58:23>> Whole new set.

1:58:24>> Yeah.

1:58:26>> I'm not going to be on that task force.

1:58:30>> Well, we as part of this process, we'll

1:58:32draft the UDO amendment and we'll draft

1:58:34suggested zoning conditions. Yeah.

1:58:36>> And that'll go to planning board and

1:58:37council to discuss further

1:58:40>> be subject to public hearings where

1:58:41people can come and speak of whether

1:58:42it's a whether what's been proposed is a

1:58:44good idea or whether it doesn't do

1:58:45enough.

1:58:53Um,

1:58:54>> is there anything from from these two

1:58:56presentations

1:58:57other than I know there's like a kind of

1:58:59a wait now just hearing like any type of

1:59:02any type of new element being introduced

1:59:04causes issues and health issues. Like is

1:59:06there anything specifically that you

1:59:08heard from either of our two speakers

1:59:09that you would like followup information

1:59:11or anything on?

1:59:18I mean maybe as more information

1:59:23comes about since this is such a new

1:59:25subject for everyone that anything they

1:59:28come across they would share with us

1:59:31you know I think that would be helpful

1:59:33>> um he uh Dr. Rob he had mentioned an

1:59:38inreview report uh paper right now. I

1:59:40can share that with you. I found another

1:59:41paper from Harvard. I'll share that with

1:59:43you guys as well and I'll put that on

1:59:44the website that's speaking directly to

1:59:46the impact data centers on public health

1:59:48and things like that. And we'll probably

1:59:49see more of this because it becomes

1:59:51community part of the local national

1:59:53conversation.

1:59:59I I just all of this is very fascinating

2:00:02this evening, but before I came here

2:00:04tonight, I had the television on in the

2:00:06background and and there was a

2:00:07commercial for I don't know, ChatGT or

2:00:09something like that. And you know,

2:00:11they're they're they're talking about,

2:00:12oh, you know, go on it and and talk

2:00:14about what your dog's Halloween

2:00:16costume's going to be and and just, you

2:00:18know, after seeing that and sitting

2:00:20there thinking, why? [laughter]

2:00:29When is our next meeting?

2:00:30>> Yeah, I can pull it up on the screen for

2:00:32everyone who's here.

2:00:43So, um, tomorrow,

2:00:46September 9th, we'll be doing a 12:00

2:00:48EAB resident call on Teams. Um, so

2:00:50that's been the invite has been sent

2:00:52out. If anyone has not gotten that

2:00:53inveter, would like to be included and

2:00:54they have not signed up, they can email

2:00:56me directly. My email is all over the AB

2:00:59data center web page and I'll send make

2:01:01sure that they get an invite for that.

2:01:02We'll be talking about um topics that

2:01:05were brought up the last two meetings,

2:01:06the one we had in the second and today.

2:01:07They'll be able to provide some input

2:01:09and feedback and then our next actual

2:01:11meeting like this one here will be on

2:01:13October 7th in this room and we'll be

2:01:15talking about noise.

2:01:19>> All right. Well, thank you

2:01:22everybody can join us as speakers

2:01:24tonight. And with that we will preserve.