AI Data Center Development Faces Grid and Community Tests

Hey guys, thanks for the intro. We're here with Jerome Parkeerson, who's the managing director for Arcova. And we're having a little chat about, well, AI data centers and what it's going to take to actually build all of them.

Hey, Jerome. Welcome to the show. Hey.

It's great having you and having me with us out there, and it's good to have a great discussion about the topic. We've seen that these data centers are proliferating regardless of whether some people like them or not. But I think the issue in my mind is maybe a lot of them aren't actually going to be built because there's a lot of things that go into that, that maybe haven't been planned for or accounted for.

So what are some of the things that trip people up on these programs? Yeah, sure. When you start looking at data centers, I think, across the board where we've seen a lot of, I think the biggest challenge is taking data centers and just splitting up in different segments.

Arcova's always stood, our position is that you need to look at the full spine. So, we look at just the engineering design piece, and we separate that from the environmental site permitting to the interconnection studies. But when you start looking at some of the biggest challenges we've seen has been community impacts, where we had pushback.

So you have right now, looking at all the statistics, about 48 states are from a local community level, so pushing back on data centers. You have huge discussions in terms of the power requirements they're looking from when you bill out from in the front of the meter, meaning that you're connecting directly to the grid, to behind the meter, where they're actually purchasing and doing the self-generation. But even that stuff has presented some significant problems, even of itself.

Even though you have hyperscales, neo-scales that are throwing huge dollars at that, and you consider the fact that construction costs, supply chains delays, and then nobody's really talking about how do we integrate the operational technology and cybersecurity element into some of the design perspectives. And then I think the other piece is that, just given where data centers are going, we now have issues where the regulatory landscapes are changing just as dramatically and as fast across the board in the States. And so whereas we need to build a data center in 2025, the same elements and inputs are not the same in '26, and it won't be in '27.

Mm-hmm. So to your point about that, didn't we kind of try to rush this through? And I ask the question because a lot of these issues are coming up were known, and yet we seem to just kind of plow through them in many places, and now it seems like they're all coming back to bite us in the proverbial behind.

So, what happened here and why didn't we think all this through a little bit better? Yeah. I think from the perspective is that, when we first went into this, the industry went into this, they went into trying to meet meeting demand, and really trying to address a need.

However, the market is very much more complex than with the way we see it today. And so, we built a lot of these data centers. We started looking into it.

We purchased land at the federal site. And then we started looking at, okay, what's the generation mix here? And then, problem number one, communities were saying, "Hey, you're building these big things in my backyard.

Am I going to pay for it? And is it going to be too loud? " On the other side of that, you had where a lot of these GPUs that are being demanded, there was not an understanding of what about all of the, what I consider the formula to making the success.

We build and then we've had examples where you've had project sites being built and then environmental issues came up later on with diesel generators being part of that generation, and so projects were halted there. I think the other part is, is that we already knew the grid was flat prior to this. And now we are doing an exponential ramping up with AI data centers, and the grid was never really prepared for it.

And so we are now asking utilities to generate a significant amount of power within 18 to 24 months cycle that took them over the last 100 years. So we're trying to move faster than that grid is, and then subsequently now we want to move faster with self-generation. But the issues don't change just because you're not connecting to the grid.

You have slightly different problems, but we never addressed the grid issue in the beginning. If you look at an international perspective, they're actually facing the same issues, which is the supply and demand, and we just can't keep up with it. When you look at just generated diesel generators alone, that has a four to five-year supply chain requirement, and you're doing your design.

But not considering, okay, I'm doing design, I have a supply chain issue, and then local resources in terms of water and energy, can it be all done in a 66-window plus? And the answer is no. Mm-hmm.

There are some folks that say that these AI data centers are part of a national security imperative, and therefore we need to go full speed and build them regardless of the impact. But is there a way to do this that's smart in some ways that we can service the mission without necessarily destroying everything in the, that gets in the way of that mission? The answer is yes.

I think there's a smart way to do this. I think what we're learning, over time we've learned that day one, it's not just about the land, and it's not for every megawatt that's being generated here. We have to think about what are all of the components that needs to be done, so that we can actually complete what I consider to be the complete puzzle here.

And so I think you have the collaboration of not just all of the entities that are involved across the seams in terms of developing a data center, but you also need to bring in that community aspect of it. Because there needs to be a part of this where you are actually investing in the local community, so they can see the benefits of that and the analysis that goes along with that. There is what I consider to be a security imperative, but I think where we really need to do is to really push on the education front of why we're doing this, so that the communities across the board can really understand the necessity for it.

I think what's happening is, we said data centers, we say AI, nobody really is understanding what that means other than those who are investing in it or actually building it. " And I think that's been a huge part. At the EI conference back earlier this year, one of the CEOs said that utilities have had a big challenge in just educating their consumers in terms of how they can be smarter consumers or in terms of leveraging some of the energy-saving elements that utilities are having.

And if you take that same formula and place it on the data centers, you have the same problem. Everybody needs to understand why we're doing this. And I think it's very simple.

We need to educate everyone along the way. " Because you're going to just get pushback across, all the way from the federal, all the way down to the local state level. Are we in danger of over-provisioning the number of data centers we need on some assumptions about what AI workloads are going to be like in four or five years when the technology itself keeps moving and evolving?

Yeah, the answer to that is absolutely yes. We have several examples of where a lot of the projects, companies out there, so hyperscales, neoscales have built, and they've over-provisioned. And the end customers, the large load customers, manufacturers and others, have said, "Hey, we don't need all this demand," and subsequently pulled out.

And so you lose a lot of money as a result of that. And I think that's where the smarter part of the analysis really needs to occur, because traditionally, we go and we develop these AI solutions, and then we retrofit as things change. And so design impacts may come into play, interconnection may also come into play.

And so then, so that adds to the construction schedules and the delays for construction schedules. " They have everything that's needed. So they purchase these, what they call ready-made lands, and then they start putting in there.

But nobody's really asking the end customer how much capacity is really needed, how much capacity is actually there in place today in order to meet that, and is that capacity really necessary for us to meet ongoing demand in the... And do we look at how supply chain impacts does that? And then at the local level, when we start looking at these regulatory impacts, especially around computational load, and how you have all of the, what I call the ISOs, such as PJM in Virginia and other states.

When they start looking at what they need to just meet current demand versus the exponential demand that's going on there, you still see that shortage that is going on. And so I think we're moving so fast that we are obviously going to reach to a point where we're going to hit a road, but we're going to hit a wall, and it's going to crash. And I think we have to be a little smarter.

We're not doing enough lessons learned in terms of, hey, if we know we need X, however, if X is not capable today, what's going to happen? Because there's a big debate about, we have battery energy storage, we have liquid natural gas. Now we're bringing in hydrogen, and we're talking about nuclear.

And when you start talking about all these other energy sources, the practicality is even if you did small modular nuclear reactors, that still takes a seven- to 10-year schedule compared to the normal schedule of 20 to 30 years. So there's still a lot of inputs that needs to be put in here that I think needs to slow down and really, I think holistically, because I think that's what's the problem. We're not being holistic here on the approach.

Speaking of those issues, one of the top of them is the energy grid itself. And in a lot of places, you've already seen reports where the demand far outstrips the ability of the energy grid to support it, and the projections don't look like it's any better. So do we maybe need to start from a different direction and maybe say, if this is a strategic initiative, maybe as a country, we need to rebuild the grid system?

Yeah, that's an interesting challenge. It's an interesting proposition for, I would say, because rebuilding the grid is going to create a significant investment, just looking at it across the board. You're absolutely correct.

Nobody permitted it. When you looked at most utilities, when they did their IRPs, their integrated resource planning, which is 10 years out, and then they do a refresh every three to five years. Nobody really expected or anticipated that you would have this much demand occurring just on the data centers perspective.

And so now that's changing the whole planning model in terms of how we add this additional energy in there. I do think that there was not enough investment in the infrastructure grid as we saw today. Even if we put AI on the side, there still needed to be some significant investment.

Now we're laying on top of that. And I think when you look from state to state, because you're correct, there's congestion on one end versus the other. When you look at California, the West Coast, and you go on the East Coast, you still see that as populations grow, the demand is growing, and we become much more connected on the grid with the use of EVs and more electrical devices on there.

When there was a discussion when we first started talking about EVs, could the electrical grid even handle, say, 50% of the consumers buying an EV and connecting and charging their cars? And the answer was no. We didn't have that capability at the time.

So we're still having that problem where the ball pain kind of kicked down the road where infrastructure was not meeting current demand. We still have a lot of transmission issues in terms of power across the board that still needs to be addressed. So remaking the grid is an interesting perspective.

I think we just need to be smarter about our technology, and we need to move a little bit faster. Again, though, that's largely related on rate base, the rate cases that are there, the PUCs coming together, the Public Utility Commissions, and really allowing the utilities to help increase the amount of advancement. I think utilities are moving faster.

Some are not moving as fast or being as deliberate. I think where the biggest gap you may have to your question is that we still have the small co-ops, the rural areas that still need to have some significant investment built up in order for them to be catching up, because I think right now they're in catch-up mode. Are we going to wind up maybe having to ration AI compute because the demand will be so far ahead of the supply?

That's a possibility. When you really think about it, I think we're trying to meet demand as quickly as possible, and we're running into a lot of the projects are being delayed or even being stopped altogether for any number of reasons. And so if you start talking about rationing, just like the grid where a lot of states, Texas and others, are looking at making sure that when you put these large loads in there, should there be an issue where they come offline all of a sudden, you don't want them to disrupt the grid so much so that it impacts the normal functioning of the grid, in the case what we had back in around July, about 1,500 megawatts.

And so there is the reliability, the stability of the electric grid that actually still needs to be there because there's a safety element that's always first in priority. And so there may be that case where we might have to slow down the channel after a while when it comes to that. And maybe because for us from technology, we're moving faster and quicker than we ever had before.

But the infrastructure that supports that fast thinking, that innovation, doesn't move as fast. So you're absolutely correct. " So I think that's not a realistic proposition.

And would we have to go so far as to look at the workloads and the use cases where maybe we're prioritizing defense and medical research over, I don't know, the latest and greatest social media meme? Well, that's a great proposition. I think that's why you have a lot of these large social media companies in of itself is really going through, especially the Neo clouds and hyperscalers, for example, are really looking at how they build their own self-generation and then sell to their customers.

And so I think they're trying to address that issue that you're bringing up in terms of prioritization. But I do think that we do some of that today, especially when we start moving into the summertime and peak loads, where we have to now start to start managing in terms of how much load goes to one customer versus others. So we do that today.

I see the same aspect of function crossing over into the AI element. If we are not able to answer the question of how much of new energy that's out there that we can really tap into. We even had some discussions where we're taking old plane turbine engines and using it as an energy source from that perspective.

So I think we're getting to the point where the capacity is going to result where we're going to be peak capacity, and we will have to slow down our build-out and demand, to your point, and I think that is an eventuality. How soon? I would say probably here in the next 24 or more so months or so, you'll start to see where it's really going to have to start to slow down.

So what would you tell our politicians about all this? Because clearly it's become a hot button issue, and it's going to affect their next campaign and whether they get reelected or elected in the first place, and it doesn't always seem to me that the conversation on either side is completely rational. Yeah.

I know. When you think about it on communities, when we talk about this community levels, and they do a breakdown in terms of, web pushback. If you just look at it pure from a Democratic, Republican, the concerns are equal.

They all have concerns. I think from a politician perspective, I think it's number one, you really need to educate yourself about this issue, and really be able to speak, what I would say, to the place where you're intelligently letting your constituents know in terms of why and how that is going to benefit them. " And, there's not what I consider to be a great cost-benefit analysis that's being done out there, that you can explain to the local constituent.

And you have to explain it to the one who lives out in the country and the rules, to the farmers, all the way out to those who are in the main cities. They all have the same concern. So I say this, politicians, you really need to first educate yourself, and let's not create what I would say to say everything bad about data centers.

They're not bad. I think it just hasn't been explained, and it just carries a bad name attached to it because we highlight more mistakes than anything else. Well, the paradox is such that the politician, what they hear is that there's going to be some new entity in town that's going to increase their tax base, and away they go.

But at the end of the day, they don't seem to take into effect the impact that that new entity is going to have on their constituents, and the constituents then either start selling, moving away, or throw them out of office, and we're caught up in this circular loop. So is there an answer here, or is this just the way things are? No, I think there's an answer.

I think there needs to be a reset, to be quite honest with you. In the state of Missouri, where I live, we've had where there was a vote, to keep to ban data centers, and then there was another county where they allowed two allow data centers, and then the local constituents voted all of those who voted for it out of the office so they can go back to it, right? And if you sit down and have the same conversation with them, they all say the same thing.

They only hear what's being told to them, and most of the time what's being told to them is really negative. And at the end of the day, the whole scenario, I go back to something very simple. What's in for me?

Is really the question. And is there a way that we can answer that question? Yes.

It's going to take some time because I think there needs to be a reeducation that needs to occur. Because right now, if we continue with the same information that's being passed out there, you're correct. " And I think we don't want that because we want some innovation.

There's a balance here. To what degree that balance is, I think that's still a debate. So a wise man once said that all politics is local, and the local folks seem to understand that there's more jobs because somebody has to build a data center, but once the data center is built, there aren't so many jobs.

In fact, some people are saying that the data center itself will be run by robots pretty darn near soon. So, what is the long-term economic value proposition, and what's the thing that people aren't appreciating? Yeah.

If I use Amazon, for an example, when Amazon goes into these local communities, they build these large warehouses, and they're fully automated. The people that they're hiring, they're bringing in and they're upskilling those persons so that they advance their skill sets, so they become a lot more smarter in terms of the use of technology. So I would say, the long-term benefit here is that these local jobs that they're creating is only going to help advance the skill sets of the people out there, and it's helping them to, what I consider to be creating a greater opportunity so that they themselves can bring the knowledge that they have that's been out there, bring into the local community so that they can stop the kids that are leaving to go for better jobs outside of the community, in terms of population flight, and actually bring some of those kids back into the local communities, because you're correct, it is local.

But if I am a college graduate and I see that there's opportunity back home, why not go back home and help build that community and make that community stronger? So I do think that there's a better connection from that perspective on a community level. Well, to fill out that thought, other folks would say that a lot of those functions are going to be remotely managed, and it won't be local folks who are doing it.

It'll be somebody five states away or maybe halfway around the world. So how do I ensure that those skills are in my locality, and then I can actually build something on top of that? Yeah.

It starts with the power of one. That's my philosophy. A lot of these rural communities, they need to be built up.

They do need some of the children that have left to come back. And I think more involvement in the discussions at the local levels, getting involved in the community boards and really asking some of the questions. You're absolutely correct.

You could argue they don't have the skill sets. " So how do we increase our workforce development, so that the local training, the high school levels, or the community colleges there, how can we add that into the curriculum so we can make them much stronger as their matriculation level? Or taking people who have just the experience, haven't gone to college, just probably graduated from high school, or even didn't even graduate from high school, bring them into the conversation in full and create skill sets and upgrade them for it.

Because I think it's not always going to be just one answer. It's going to be a multiplicity of approaches that occur, at the local level, to help build those communities stronger. Because you're absolutely correct.

Yes, they can come in and say there's no skill sets and automate everything out of it. But AI, no matter what, will always require some level of human interaction, believe it or not. There you go.

Hey, folks, I think you heard it here. I think the one thing that everybody agrees on is we just need a better way and a smarter way to build data centers. Hey, Jerome, thanks for being on the show.

Thank you. I appreciate it. And back to you guys in the studio.