Powering the Future: Data Centers, AI, and Energy with Peter Hans Hirschboeck - E138

In this episode, we sit down with Peter Hans Hirschboeck, founder and managing director of impactECI, to explore the rapidly evolving world of data centers and energy infrastructure. With a background in electrical engineering and years of experience leading energy projects at AWS, Peter shares invaluable insights on how the demand for AI computing power is reshaping the energy landscape.

We discuss:

• The massive growth of data centers and its impact on energy grids
• Why grid bottlenecks and interconnection queues are delaying projects
• How hyperscalers and AI companies are securing power for large-scale deployments
• The increasing role of nuclear, renewables, and behind-the-meter solutions
• How utilities, developers, and tech companies can better collaborate to solve infrastructure challenges

Peter also breaks down the current trends in power procurement, the importance of speed to deployment, and why grid reform is essential to meet future energy demands.

If you’re in energy, data centers, or AI infrastructure, this episode is packed with must-know insights!

Connect with Peter & impactECI here!

Paces helps developers find and evaluate the sites most suitable for renewable development. Interested in a call with James, CEO @ Paces?

Transcript

00:01.44
James McWalter
ah at Hello today, we're speaking with Peter Hans Hirschbeck, founder and managing director at Impact ECI and a valued advisor on data centers for Paces. so Welcome to the podcast, Peter.

00:12.74
Peter
Thanks, James.

00:14.44
James McWalter
Brilliant. I guess to start, we'd love to hear a little bit about kind of your early career and how you ended up working on data centers.

00:20.55
Peter
Oh, that's a long story. And you know with a lot of people in data center space these days who came through the energy sector, it is a long story. So I studied electrical engineering at the University of Illinois. And you know electrical engineering covers a lot of things. It covers satellites and transistors and waves and all these crazy things.

00:40.73
Peter
i talked to my like, advisor in the department. And she said, well, there'll always be jobs in energy and power. And I was like, okay, I'll give it a try. Because at that time, you know, Motorola was getting sort of eaten up. And a lot of jobs were going to, you know, overseas or wherever. um So I checked out power, and I really liked it. And so after school, I worked at a utility out in San Diego, San Diego Gas and Electric, ended up at Navigant, which is now Guidehouse doing a lot of strategy work. And so it started out doing

01:12.87
Peter
power plant work, and then it was wind plant work, and then it was dirt, and it just kept going more and more towards sort of the infrastructure and climate sector, I guess, and kept doing that for a decade, and then then had an opportunity to join AWS, and they're what they call the energy project management team where, um you know, responsible for building massive electrical facilities and making sure the utility was able to connect, you know, hundreds of megawatts at a time.

01:42.65
Peter
um And yeah, i got that that was really the first exposure I had to data centers is working in AWS.

01:50.30
James McWalter
So going to, so it was like around, I think 2019, right? At the beginning of 2019.

01:54.10
Peter
That's right.

01:54.49
James McWalter
I guess like, what what was like this, like what was status quo data center world at that time?

01:58.88
Peter
Yeah, it's crazy. because even you know Data center world seems to be changing every six to nine months. So going back to 2019, I have to scratch my brain to remember what it was like. But we were looking at some of the largest campuses we would even consider were 100, 150 megawatts um across you know four or five buildings. And you know a lot of times you would get grid connections on the sort of distribution system of the grid without even having to build a substation. It was a different time.

02:30.06
James McWalter
So the, um, and I guess like, you know, obviously Amazon, AWS, you know, hyperscaler, very impressive company. Um, you know, I guess what was the kind of like, as you were joining them compared to, you know, some of the other companies, some very well established companies, but yes, smaller entities. Like what was the kind of a cultural shift working at at an AWS?

02:50.78
Peter
Um, I think there was a few things. The first was just the massive scale of the projects we were working on. um You know, even as I just described those are small data centers in today's world at the time they were massive and get to remember that, you know,

03:06.73
Peter
even five years ago, there was no load growth in this country electrically because everything that was you know being generated from economic activity was kind of being offset by all the energy efficiency type, you know, natural evolution via sort of non-programmatic energy efficiency. So things like better refrigerators and dryers and and all that, but then also also programmatic energy efficiency where, you know, states and utilities had targets they had to hit. Um, so you were getting like 0% load and data centers sort of changed the game. And we were doing massive projects in an environment. No one was doing them. That's it. That was the first sort of shock. The second was the amount of money and speed that was required, um, to, to sort of grow AWS at, you know, 40% year over year. Um, you know,

03:55.84
Peter
It's funny when I look back that, you know, in the first day, I think we maybe had a gigawatt of operating data centers, which is now one project if you're lucky enough to get it, ah which is kind of crazy.

04:06.13
Peter
And at the time they owned, you know, 50% of the market or more. And, you know, that was a gigawatt. Now that's, you know, one, one big project in Texas.

04:13.98
James McWalter
One project. Yeah.

04:15.79
Peter
ah

04:17.19
James McWalter
you know what I think you're pretty familiar with this, but for the audience, you know our co-founder faces Charles. like He was actively working on data center efficiency at the programmatic level as well in his last couple of years while at Facebook.

04:30.56
James McWalter
um and you know I think a lot of the kind of ah the Even like the the power use curves that folks were printing were like, oh, the power use from data centers on net just never had that like massive exponential right in those years because of the efficiency gains that you outlined.

04:47.48
James McWalter
um Maybe there's some exceptions to that. like you know Ireland has had its own you know amount of large-scale consumption of of the grid over the course of the years. But in the US, s like it actually like like data centers as a proportion of like total power and demand and load was fairly flat basically until 24 months ago.

04:57.77
Peter
Thank you.

05:06.55
Peter
Yeah. And it's interesting too, because I think, you know, we talk about hyper scale data centers, right. And, you know, there's this whole class that, you know, AWS was able to to take, be so successful in, which was.

05:19.82
Peter
It wasn't as if the data centers didn't exist, the computing facilities didn't exist. A lot of them were being done at the enterprise level. So Citibank had their own data center or whoever had their own data center. And it was only when cloud became, it's clear that AWS knows what they're doing in infrastructure. They have these sticky software and cloud.

05:38.98
Peter
platforms and and sets of services that people want to migrate to, that that shift started to occur from you know legacy data centers, on-prem data centers into the cloud. And then that boom really took off. so you know there was and And naturally, there's a lot more compute these data systems in the society we live in. Data is a lot more used than it was five, 10 years ago. But um yeah, there the thing is is that As you move from on-prem to sort of cloud data centers, the efficiencies are much greater because, you know, if I use the example city and maybe that's a good example, maybe it's not, but, you know, they're not in the business of operating data centers, whereas AWS is. And, you know, they're not building, you know, 50 megawatt shells to to house the facilities and they're not doing central cooling plants. And so in general, the efficiency just moved that way as well.

06:29.28
James McWalter
And you one of the things, I think we were both at the same conference a few months ago, where you had some folks on the utility side you know who were outlining the history of this period where we're talking about, 2015 through 2022, as like a period where a lot of the utilities were expecting to have to support a lot of load growth. And when it didn't materialize from the data centers, they were potentially slower to move this time around, right because they got a little bit spooked previously.

06:58.46
James McWalter
I guess like, you know, when you were kind of like building relationships with utilities, I guess like, what was the kind of general sense for utilities at that time, ah to, you know, the potential and like, and, the you know, and then the reaction, I guess, as the actual load growth did not appear to the same level that they potentially expected.

07:14.42
Peter
Well, I, yeah, I certainly had, uh, you know, I have good relationships with utilities. You know, we, we considered them partners, right? Uh, you know, I always had a data center as an electron refinery and you know, the important input is the electricity to turn it from 60 Hertz into light and bits and and all of that good stuff and heat for that matter. But in any event, um, the relationships I built with the the utilities, we were very transparent about our load growth and we would say.

07:42.73
Peter
hey, utility A, you know, we're going to be, built these are the facilities we have in the pipeline. You can see this because we've requested grid connections. Here's when we intend to turn them on. And here is the associated sort of demand and consumption that will be impacting your grid width. And, you know, we would tell them there's, you know, billions of dollars of underwriting behind this. So, you know, this project will happen. At least the high percentages this project will happen. And so,

08:12.18
Peter
you know, when we say it didn't materialize, i always I never actually had that experience. I had the experience of it did materialize, and they were just um discounting us or not prepared. or And I think maybe the more likely reason is, you know, it it was really hard to take that amount of power and hand it over to plan for it, to get the infrastructure built, to get the outages you need to put in something new in service. I feel like it's just you know an industry like tech moves so fast, and an industry like utilities doesn't. And 30-year assets, 40-year assets, that's the norm in utility speaking.

08:52.45
Peter
You know, they didn't have the same type of returns that a tech company had turning, you know, what would you do with 150 megawatts at a utility versus 150 megawatts at a data center? Those two things are drastically different in terms of value.

09:06.24
James McWalter
one One thing I've always been like really curious about is you know people are going to every potential market now right to to try to capture the demand. um But you know there's a reason, I guess, why that Northern Virginia, parts of Ohio, parts of the West Coast right became like the initial hubs.

09:22.12
James McWalter
I guess like why why those places? right like ah Because so much of the reaction from the market is like, oh, you know you can't get power in Virginia anymore.

09:25.96
Peter
Yeah.

09:29.86
Peter
Yeah.

09:30.21
James McWalter
right

09:31.25
Peter
So I'm a power guy, right? Like I understand like 60 and 50 Hertz electrons and I understand, you know, power plants and transmission lines and substations and all of these things.

09:42.18
Peter
Uh, I'm going to go out on the limb and I'm going to try to explain why I think, or why it's been told to me that this was exciting. And you know, some of it has to do with. um, sort of defense applications and communications, uh, like ARPANET. Um, some of it has to do with, you know, there was a guy who owned a dairy farm and he decided to build a data center and everyone else wanted to be around him. And then I think one of the most important reasons that it's been explained to me, and again, this is how I understand it, is that when you run, you know, uh, what I would call a cluster or when you build data centers and you have clusters around it.

10:18.51
Peter
That means that, you know, database A is located there and database B is located there and service A is the as service C is there and. you know we We look at computation or Cloud use as you know James and Peter are talking to each other through this platform. But in reality, what's happening is this platform has a service provider and a sub-service provider that bounce things off of each other. and The latency time when everything is located in a cluster um sort of shortens that latency time and and then the only real source of delay is is to get to the endpoints.

10:55.52
Peter
And that is how I understand it. But to say why Northern Virginia, I think it was because it was sort of the first cluster and everyone wanted to grow around it. Now, again, I'll add the caveat that that may be just the way the world was communicated to me, but that's how I understand it.

11:11.75
James McWalter
that's ah That that's all makes sense. And I'm sure somebody will reach out if we're correct on that. um i think i would love to not yeah We can get back to this. I do think it's very interesting, the kind of interaction between power people, you know like internet you know people who are software people, like how all these folks are interacting, especially in this kind of like you know modern age, the the commercial real estate people. right like Everyone is now like taking these different pieces.

11:38.68
James McWalter
And one of the things we are being exposed to at PACE is, I don't know, even more for you, Peter, is like how all these different entities are like seeing the problem and often maybe miscommunicating what is important about some of these opportunities. um But I guess before we get into that, um you know you ended up wanting to start something yourself. like That's all obviously always like a big step when you started Impact ECI. I guess what drove the initial impetus to to jump off on your own? And you know what was the i guess what what does it impact that ECI did?

12:08.53
Peter
Yeah, um I guess the initial impetus came from the fact that while AWS is an amazing place and you're doing such cool things and you learn so much, it's ah it's a really hard place to work. um It's stressful. There's tough assignments, tough hours at times. And um you know I recognize the skill set that I had were in demand that skill set was in demand by other folks out there, other entities. um you know I wanted to do a little bit more of active climate work as opposed to you know just setting up um

12:44.69
Peter
you know deals or or things like this to actually advise OEMs. And I had come from a consulting background, so it was never a perfect fit with AWS. So, you know, I made the decision after about four years to to step out on my own and yeah, impact ECI was the result. It means ah The ECI means energy, climate, and infrastructure. And so we support organizations that on the energy side are large consumers, large producers, somewhere in the value or supply chains, OEMs and equipment, entities that you know have a financial interest in energy at some point. um And the climate side, it's it's really focused on emitters and OEMs and and that sort of thing. And then what I call infrastructure

13:33.48
Peter
which is really just mega infrastructure program and project management for companies like data center companies or like utilities, things like this.

13:44.56
James McWalter
So the, you know, you then have been like in this industry right now, you know, you saw it up front in personnel with AWS. And then as you've been building out impact DCI, you know, you're seeing lots of different folks right across, you know, different parts of the industry, trying to figure out some of these problems. So I guess like in the last like two and a half years while working on impact DCI, like how has the market shifted, particularly on the load side?

14:09.77
Peter
Oh, let me count the ways. Um, so two and a half years ago, uh, three years ago, somewhere in there getting capacity for a data center project, say, Oh, I've got this plan. I want to go out and get a hundred, 150 megawatts. So yeah. Okay. Um, you know, we've got a 230 KV line here. We're putting in a new substation in here. There's no supply chain issues. Yeah. We can probably have you power in.

14:33.28
Peter
you know, three years. It's like, great, that's a lot longer than I'd hoped for, but three years is fine. Now, and there'd also be parts of the grid that were untapped. And so you could go to a region where maybe there hadn't been a huge data center presence, but then, you know, it was available. ah Arizona was a big example of the mechanics market, where maybe there was a lot of grid capacity that was just available. And and now, especially with the expansion of AI data centers, which can be sort of located more you know distant from major pockets of cities and clusters and things like this. All over the country, a lot of the available capacity on the grid, whether that's substation capacity or transmission line capacity or capacity on the generation network, um there's very few regions in the US that you can get away with just going to a utility provider and asking for

15:29.68
Peter
100, 200, 300 megawatts, you know there are certainly spots. They're more like diamonds in the rough versus the norm though.

15:38.10
James McWalter
And, you know, and I think like, basically, you've also had this incredible, I guess, perspective on the importance of speed to deployment, right? Because, you know, even when three years felt relatively fast, um you know, ah there was like, this large scale ah competition that was happening between Google GCP, Azure and Microsoft AWS, which was the first but then starting to have you know, new entrants there that are competing against them on that cloud business side. But you were you didn't have like the race that we' we're having today around, you know, first to massive AI model deployment and these large scale ah AI training runs. And so, you know, we're kind of like in this perfect storm where the timelines have to compress for the business ah purposes of the large hyperscalers. At the same time, when, you know, queues are on the load side are, you know, ah five, six, seven years.

16:30.59
Peter
Yeah, and you know you you bring it it up. It's exactly the right point. you On the AI side, um you know it's such a big prize to be obtained for whoever gets there first. So they move so fast. And um you know they're they're asking for tons of load now. that They can be a little bit more um distant. So they're not exactly one-to-one competing. you know They're not trying to go into the middle of you know the same markets as as the hyperscale cloud service provider deployments are, but um you know it's ah it's kind of crazy. um Sorry, I've lost track of the question. Can you repeat that again?

17:09.96
James McWalter
I was just saying like the speed, interconnection queues are lengthening at the same time, that's speed.

17:15.06
Peter
yeah Yeah, and so the other thing is is that

17:15.92
James McWalter
yep

17:19.95
Peter
Because of these interconnection timelines, getting to five years, getting to six years, having a you know crazy study process, because if you're a land owner, or developer, or a speculator, you know the wild catter of the data center space, you're what's known as spamming the queue and just sending in applications. You say, I've got this land under control. Let's ask for 300 megawatts. And if everyone does that and the utility has a structure in place to do the study where they model the system with different you know contingencies and loads all over the place. And there's a 10 gigawatt queue of load interconnection. And of course, it's going to be a mess. And so these queues get really long. um And we can talk about grid reform in a minute, but um

18:06.33
Peter
A lot of it is driving these developers of data centers, whether they're hyperscalers or AI companies, to think about alternative connection modalities than just ah what I would call a 2N grid connection, which is two sources from from the grid serving your site so that one goes down, you have the other one. um Maybe they built power plants. You see a lot of the announcements of of the cloud service providers gonna use nuclear for their data centers when when that is available to them. Whether it's restarting reactors or putting data centers behind the meters that existing plants or building next generation nuclear like SMR is up to power the data centers of 2035 or around then. um you know Some are thinking solar and storage as you well know. um And so,

18:55.16
Peter
Yeah, just considering alternatives to the grid, I think is you know what we've seen in the last six to nine months.

19:02.16
James McWalter
Yeah, I think the, I'd love to kind of like talk through some of the the kind of ways folks are reacting to the queue timelines, and then also like how that's now showing up other problems in timeline issues around the procurement of some certain equipment, right?

19:17.52
James McWalter
Because like all of a sudden, if you're you're going to build a power plant, you need line of sight of ah procurement that you didn't, you know, maybe you need more, you know, or you didn't need for a typical kind of grid connection.

19:21.48
Peter
Yeah.

19:28.62
James McWalter
I think one of the bits that we've seen, and you know we first started at PACE is having our first data center developer customers about 15 months ago, 14 months ago. um And they had a set of things that were pretty you know pretty standard. I think a lot of folks are are pretty familiar. You want to be near a population center from a latency perspective. You want to be in one of the tier one markets, you know some of the ones we already mentioned. And you want to have the grid connection occur within three for eventually five, six you know years, and it started extending.

19:57.98
James McWalter
And then what we started seeing over the course of summer, you know we would have customers who would like whisper these cities that are more like the the secret. this is the you know this is This is a particular city.

20:08.04
James McWalter
Have you heard of like this part of Nevada, or you heard of this part of whatever the country?

20:10.28
Peter
Yeah.

20:12.09
James McWalter
I was like, oh, nobody else knows about this. And like by, I guess, like August 24, literally everyone we're talking to, we're all saying the same place. It's like, oh, the secret's out to a certain extent.

20:20.82
Peter
Salt Lake City, have you heard of it?

20:22.09
James McWalter
exactly Exactly. Like all all this kind of thing. Cheyenne Mountain, hot all of a sudden, right? Like, you have all this kind of thing happening. And what we started seeing then was this, at Paces, was folks who were like, okay, we're just not not going to get our power ah to the to the degree we need. um And then the other piece, so let's look at alternative systems like full behind-the-meter gas and and some of the other ah hybrid systems, you know, i happy to go deeper on.

20:44.01
James McWalter
But I thought what was really interesting as well for us was like learning the differences between the load queue and the generation queue. And the the fact that you have so little visibility in the load queue, it's way easier to get into it. The cost and the fees associated with it are trivial compared to the generation side. um And there is january off to recently, the there's been a huge resistance by the utilities to be very public about this information. So yeah, we'd love to hear, I guess,

21:10.33
James McWalter
you know like Grid reform ideas if you have them, but also just like what how you know how kind of dealing with the load queue, how much of it is just like making a phone call, right?

21:13.89
Peter
Yeah.

21:19.39
James McWalter
Trying to find the right guy.

21:21.33
Peter
I'll do that part first, right? Because I have that on the tip of my tongue and then we can get into grid reform, which is, you know, there are very, there's there's not very few entities, but there's many fewer entities that do gen side queue sort of work than the load side. You know, there's thousands of utilities across the US and those are the ones who have the, you know, of service territories that they have exclusive domain to serve their customers on. You know, there's many less on the gen side, whether it's an organized market or liability entity or a federal power agency, a lot fewer of those. And so, um yeah, as you as you mentioned, a lot greater transparency on the gen side. And so these utilities,

22:11.43
Peter
especially in markets that are, you know, less deregulate, I guess, sorry, that's the wrong way, from the wrong word, especially in markets that are sort of less.

22:29.75
Peter
um Yeah, so yes utilities are much less transparent about what load is coming in. A lot of that it has to do with just sort of being client confidential. We don't want to share information about our customers because you know they're consuming load and for a data center, maybe that's a trade secret, or for others, you'd want to be stealth about you know investment until you have the load secure. on On the generation interconnection side, a lot of it is for for public benefit. At the end of the day,

22:59.34
Peter
you know Most of these markets want to ensure that there's sufficient generation and it's at the right price for um you know economic growth and for customers to not get ripped up ripped off and and that sort of thing. They want to make sure the sort of lowest cost energy supply. um Those optimization frameworks aren't really there on on the gens are on the load side. So it's a really interesting dynamic. And you know you see some agencies actually reporting like load interconnection requests.

23:29.23
Peter
And I really hope more of them do it, because you know it's it's a challenge when you hear that one of these utilities or markets has 10 gigawatts of load interconnection requests. And they don't tell you any detail about it. And you probably know, just having done this for a long time, that half of it is BS. It's maybe more than half. But which that that sort of leads us to the grid reform.

23:55.95
Peter
And so I'm lucky enough to do projects all over the world. So, you know, projects in Spain or the UK or in Japan or Brazil. And you see some other countries that are struggling with, you know, this very same thing, which is, hey, we have way more load interconnection requests than we have.

24:15.07
Peter
you know capacity available to serve that load. And we know that a lot of that is you know speculative. How do we make sure that um you know we only honor requests for things that are actually going to get built or that aren't speculative? And so you see a lot of different mechanisms. You see sort of deposit payments or posting of security to be able to say you know it's X dollars or X euros per kilowatt in terms of posting a security or a parent guarantee or something like this. You see a lot of pro-rata shares to say, well, this region we know we're going to grant one gigawatt and everyone is going to have a chance to apply. And if your product project qualifies, you're granted a pro-rata share of that one gigawatt. You know, there's, there's mechanisms where

25:04.74
Peter
You know, it's it's first first ready, first build, right? Where um if you're ready to go, then you got the capacity, sort of qualified projects. There's probably a dozen of these really interesting things going on in grid reform. And I've talked to some friends at utilities in the US and and that's starting to take root here as well, where, you know, there is going to be load interconnection request reform in the US. s

25:33.32
James McWalter
I think for a lot of the folks who are this podcast are folks who are generation developers, and I think they'll find it very surprising that the load queue is even more opaque than, you know, because like, and it is, right? It's like, oh, like generation queue is also a nightmare. It's like, whoa, it's so much better. You do so much more with it. And by definition, you have way fewer entities, right? Because yeah like There's a lot of M&E folks who are connecting plus one megawatt projects on the generation side. But you have a lot of entities you know posting um you know very, very large variants of like load on the load side.

26:05.63
Peter
Yeah, I mean, if you look at the hijinks that were going on in you know PJM related to storage connections a few years back and even now still,

26:05.63
James McWalter
ah

26:13.37
Peter
like ah You're right to say that maybe you're looking at a 20 megawatt project or a 50 megawatt project, but everyone is trying to get load for 500 megawatt data centers and there's five times as many. You're right to say it's it's a bit of a different beast.

26:28.31
James McWalter
Yeah, absolutely. And then I guess like in terms of the kind of alternative model, you know in terms of behind the meter grid connected or line precise, folks are looking at initially were primarily gas. um yeah We were co-writers on this white paper about ah how you could yeah supplement gas to get to the you know to get to the required level of data center redundancy with a mix of ah of batteries and solar.

26:55.36
James McWalter
um I guess like, can you just speak to the importance of redundancy? And then I guess secondarily, you know, one of the challenges I think a lot of folks who've been in the power space a long time are having with this massive data center opportunity is really understanding how difficult it is and how much has to be de-risked for a power project to actually be something a data center developer is interested in.

27:17.04
James McWalter
Because we've seen it paces in this massive disconnect with folks who are like, hey, this this looks project and this looks like a great renewable project or some other project.

27:18.36
Peter
Yeah.

27:25.32
James McWalter
But it is miles away from being something a data center developer or a eventual off-taker that's a hyperscaler would be interested in um because they's just like oh it's it's not, you know, they're not going to take on risks when they don't have to and they have so many opportunities. But yeah, I'm going to hear your thoughts on that.

27:40.64
Peter
Yeah, um I think the first thing to put this in context is just understanding the amount of money and margins that data centers make, especially cloud service providers, you know they're spending billions of dollars on infrastructure for things that pay back in a ridiculously quick timeframe, you know maybe on the matter of months in terms of the investment. and so You know, it, there is seriousness and one getting it right when it comes to your decision-making about, you know, what qualifies as, you know, the design that we want. Say you want to have two N like redundancy or you want to have two N plus one, you know, whatever N plus one, you know, whatever the requirement for a certain part of data center is, there's just a lot of money at stake here.

28:25.48
Peter
um And two, you know, trust really matters. um You know, if. if you have a streaming service on your cloud platform and it goes down, well, they're gonna be really pissed off and then they're gonna be pissed off at you. And then they say, well, this giant contract that is actually basically all of our expend, we're gonna move it to another provider. And so reputation really matters. And so um you know data centers, especially on the cloud siding for what they call five nines, reliability or uptime where you know the system has to be basically bulletproof

29:02.89
Peter
Uh, in terms of the design and the supply and the critical inputs. And so, you know, when, uh, you build a data center and you say, let's take a 300 megawatt data center. We're going to build in. I don't know, Columbus, Ohio. Uh, well, what we really need to do is we need to get to.

29:22.43
Peter
probably high voltage 230 or above lines that are each capable of carrying 300 megawatts brought to the on-site substation. So if one of those goes down, you can get the power from another. Now that wreaks havoc on system planning, of course, ah but it also, um so really you know it's not cheap and it's not easy in a lot of cases where you know you have to study those flows. um So that that's a huge challenge with um you know that that just understanding of the root resiliency. You keep going and you go all the way from sort of the hot extra high voltage system or high voltage system through the substation where again, there's redundancy in every step to the medium voltage system on the campus of the data center where you have you know UPS systems and you have industrial cooling equipment and and all sorts of things. And you get down to the low voltage system within the walls of the data centers that run the compute equipment

30:17.99
Peter
everything is resilient and redundant. So there's a failure, you know, you have a system to catch it. You have a system to so make sure that you don't lose uptime. And so that is highly capital intensive. It is not like, you know, how you build other systems and, you know, the power grid. Yeah, of course, you don't want the grid to drop if you're just a regular, you know, i imagine a world of 30 years ago when there were no massive data centers.

30:45.73
Peter
You don't want the grid to drop, but, you know, are you really prepared to double or triple your CAPEX so that you avoid five minutes of tail risk where, you know, Hey, sorry, the grid's out. And so, you know, the traditional way of thinking is sort of upended when you're talking about data science.

31:00.39
James McWalter
Yeah. the And so you have like these multiple kind of entities, which we refer to, right? So you have, you know, the amazing kind of like skill set of the power developer who's like, I can get electrons delivered at a price that an optic will want. um I can build, you know, a system. I understand how to connect into the transmission distribution system, depending on the size of it.

31:22.49
James McWalter
and all the skills associated with that. You then have like the capital side, right, where money is trying to figure out where to go. You know, when it's very, very heavy, the risk projects, there's like, you know, not quite unlimited, but close unlimited capital for these projects, right, because the the ah ROI, as you said, is in months.

31:38.75
James McWalter
Then you have, like I think, a lot of speculative capital coming from commercial real estate right now, where folks are making a lot of early bets on, hey, you could build ah this site in you know a part of air cost, or you know you could do something funky here. um And then there like what we're seeing is like this big gap in the middle from both a capital and a skill set perspective. right So you have a lot of folks who are like, OK, these are sites that could be great projects for to to Path to Power for the Data Center. um And then there's often struggling to figure out, okay, who's going to be the buyer of that to develop it to the level needed to like access that capital on the back end, right? ah And then you also have just you shrink on this, this

32:17.57
James McWalter
I'll call it it like the Silicon Valley software viewpoint. right And you know that's that's my background. right So like you I know that viewpoint most of all, where there's just like a fundamental misunderstanding about like why can't this stuff go faster, right where people are like, hey, you know we are building a digital god like in the desert.

32:29.95
Peter
Yeah.

32:32.41
James McWalter
Why is that taking more than two years? And like the sentence I just said, I've had a person literally tell me. right and and they're And they're from a name brand company that people would recognize.

32:38.18
Peter
Yeah.

32:41.97
James McWalter
And so I guess like as you kind of think about this, like what where can like better orchestration between these different parties, like how can like these perspectives be better kind of joined together? And and I guess what are you saying on the market?

32:52.75
Peter
Yeah, that's a really wonderful question. If I had the answer, I would probably be doing a lot better than I am now, right? I think if I had had to think through it,

32:57.28
James McWalter
There you go.

33:00.08
Peter
you know There's things that are, yeah I wouldn't say unavoidable, but really humongous challenges. So there's supply chain, not only with sort of highly you know tech heavy chips type things, but also things like generators or switch gears, which you are electrical equipment vital to the data center. And then there's, you know obviously we talked about grid. There's only a ah finite number of power plants and transmission facilities and substations that can get you the load that you want. And so, you know, I think, you know, there are things that are really hard challenges to solve. And until you solve those, the people who want things fast and whatever fast and

33:43.23
Peter
quick and cheap and easy and all of this are going to have to read readjust their expectations or they're going to have to say, well, let's roll up our sleeves and let's solve the grid bottleneck issue. you know Let's try to make permitting a transmission line easier or let's try to you know build a new generator factory or or we're going to you know, try to try to get another chips facility in domestically or whatever it is. These are the massive problems that need to be solved in order for sort of tech like sort of expectations and a heavy infrastructure world begin to take form.

34:22.07
James McWalter
could not agree more um And I think, you know, one of the things that folks who are, you know, and interacting and you know I've seen you, you know, speak at different events and you know and and and and you know via conferences, and I think like having folks who have been at the the larger the hyperscaler and is now kind of like on the development side, especially when you have like that power background, I think has been like really powerful to kind of see and like has been you really, really strong, yeah really has improved you know our ability here at Paces to understand the market um and not just you know be kind of reactive and like, oh, let's build what people ask, but try to like build yeah software solutions for like what actually are going to solve some of these problems.

34:59.36
James McWalter
um Last thing before we could have wrapped up, I would be remiss without just mentioning it because it was everything that we were talking about two or three weeks ago was like deep seek, right? And the reaction on on the markets. I feel a paces and myself like we where this does not really affect the actual flow of capital for data centers.

35:22.05
James McWalter
Its efficiency is going to be driven, um but that means there will be ever more demand. Do you have this paradox and all that good stuff? But yeah, I'd love to see if you have a particular or take on DeepSeek and how that essentially has changed you any of Europe's existing priors.

35:35.09
Peter
Mm hmm. As I said before, you know I did not go into the really really smart guy like chip design, AI, CS type world of electrical engineering. I went into power. And so I can't speak intelligently about how you know AI models are trained or you know the and semiconductor physics that allows for these chips to work one way or another. um you know My initial reaction was,

36:01.61
Peter
you know Is this too good to be true? It sounds like it is. um you know And all you have to do is say a cost lower than it actually was. And if people believe you, you move markets. So yeah, I don't know anything about deep-sick. And I wish I could speak more intelligently about it. But you know I think you're right to say that the demand for computing infrastructure will be there, regardless if deep-seeks claims are real or not. It's just going to drive up the table.

36:27.61
James McWalter
Absolutely. Peter, it's been absolutely fantastic. I guess before we finish off, is there anything I should have asked you about that did not?

36:34.73
Peter
um Well, I mean, there's plenty. I may not have known the answers of like, where should I be ah build a data center today? you know I don't know the answer to this. um you know I think what we're seeing is a lot of the rotation into generations.

36:50.74
Peter
And I think yeah if we had drilled into that and say, you know what do you think of generation? What I would say is, I think we need an all of the above type approach. um you know We need more generation on the grid. We need more transmission on the grid. We need more generation behind the meter. We need more nuclear. We need grid reform. We need all of these things to make sure that um you know data centers are being built to keep up with demand, especially in the AI world.

37:20.72
James McWalter
Totally agree. And I guess if folks ah want to kind of engage with you or are interested in like learning more about your offering, I guess, yeah, we'll include your contact details in the show notes.

37:32.23
Peter
Yeah, absolutely. And, you know, where Impact ECI does a lot of work, like I said, in energy and climate and infrastructure, doing work all over the world, which is really exciting.

37:42.64
Peter
And so, you know, we're looking to grow. Clients of Impact ECI are looking to grow, so we'd love to to chat with folks.

37:50.74
James McWalter
amazing Thank you very much, Peter.

37:52.56
Peter
All right. Thanks, James.