Building the Power Behind AI: Nuclear, SMRs, and the Delivery Challenge

Electricity demand is entering a sharp new growth cycle. After years of relatively flat demand, the economy is becoming more electric at the same time AI and data centers are exploding. New manufacturing, reshoring, EV charging, building electrification, air conditioning growth, extreme weather, and 24/7 reliability expectations are all pushing more load onto the grid. 

AI is the loudest accelerant right now, and for good reason. Data centers need massive computing power, cooling systems, backup power, transformers, substations, and high-reliability grid connections. But the bigger story is broader: the energy transition is no longer just about generating cleaner electricity. It is about delivering enough reliable power, fast enough, in the places where demand is actually showing up.

That is why nuclear power, SMRs, gas, storage, renewables, transmission, transformers, and interconnection are currently front-page issues.

But there is a harder question behind the hype: can we actually build the infrastructure fast enough, safely enough, and predictably enough to meet the moment?

In this episode of Green Giants: Titans of Renewable Energy, host Wes Ashworth, President of Lee Group Search sits down with Todd Zabelle, author of Built to Fail: Why Construction Projects Take So Long, Cost Too Much, and How to Fix It. Todd brings more than 35 years of experience in complex capital project delivery and is the Founder and CEO of Strategic Project Solutions, Founder of the Project Production Institute, Founder and CEO of Pacific Contracting, and a founding equity partner of the Lean Construction Institute.

Todd’s perspective is clear: many major projects do not fail because the technology is impossible. They fail because the management system is wrong.

That matters now because AI, data centers, nuclear power, and SMRs are all converging around the same challenge: the physical world does not move at software speed.

Wes and Todd unpack why traditional project management tools often create false confidence, why schedules can become disconnected from actual production, and why measuring progress by money spent can hide problems until it is too late. Todd explains why critical infrastructure needs to be managed as a production system, not just an administrative exercise.

The conversation then turns to nuclear power for AI. Todd shares what he heard at Datacloud Global Congress, where nuclear for AI was a major topic, and explains the cultural gap between fast-moving data center developers and the more conservative, safety-driven nuclear industry.

They explore the real-world barriers standing between AI’s energy demand and nuclear deployment, including siting, permitting, grid connection, fuel supply, specialized labor, nuclear-grade quality standards, and the mismatch between data center load profiles and steady baseload nuclear generation.

The episode also takes a sober look at SMRs. Todd sees real value in smaller batch deployment, especially as data centers grow over time, but warns that SMRs are not plug and play. Even smaller reactors require serious infrastructure, containment, supply chains, fuel strategy, skilled operators, and disciplined execution.

This is not a conversation about whether nuclear matters. It clearly does. It is a conversation about what it will actually take to deliver nuclear-grade infrastructure in the AI era.

In this episode, Wes and Todd cover:

• Why U.S. power demand is rising after years of relative stagnation
• How AI, data centers, electrification, manufacturing, and reliability needs are reshaping the grid
• Why major capital projects often fail long before the public sees the delay or cost overrun
• How traditional scheduling, earned value, and percentage completion accounting can create false confidence
• Why project delivery should focus on production systems, not just administration
• What data center developers may be underestimating about nuclear power
• Why SMRs solve some problems but leave many execution challenges intact
• The fuel supply, workforce, and quality-control bottlenecks behind nuclear scaling
• Why nuclear-grade welding, rebar, and construction tolerances demand a different level of skill
• What capable owners must do differently on complex infrastructure projects
• Where AI, digital twins, sensors, and automation can actually improve project delivery
• Why the energy transition depends on execution discipline as much as technology innovation

If you are working in renewable energy, nuclear power, data center development, infrastructure, construction, utilities, project delivery, or AI energy strategy, this episode is a grounded look at the gap between ambition and execution.

The future of AI will require massive amounts of power. The future of clean energy will require massive amounts of infrastructure. And as Todd makes clear, the winners will be the companies and leaders who understand how to actually build.

Links:
Todd Zabelle on LinkedIn
Project Production Institute's Website
Strategic Project Solutions Website

Wes Ashworth: https://www.linkedin.com/in/weslgs/

• Email: wes@leegroupsearch.com
• https://leegroupsearch.com/green-giants-podcast/
• https://leegroupsearch.com/

Transcript

Wes Ashworth (00:25)

Welcome back to Green Giants, Titans of Renewable Energy. Today's conversation is about AI, nuclear power, data centers, and the hard reality of building critical infrastructure. Our guest is Todd Zabelle, author of Built to Fail: Why Construction Projects Take So Long, Cost Too Much, and How to Fix It. Todd has spent more than 35 years in complex capital project delivery. He's the founder and CEO of Strategic Project Solutions, founder of the Production Institute.

Founder of Pacific Contracting, and one of the founding equity partners of the Lean Construction Institute. Todd's lens is project delivery. His argument is many mega projects do not fail because the technology is impossible. They fail because the management system is wrong. That matters right now because AI is creating urgent demand for reliable power, and nuclear is back in the center of the conversation. But Todd's warning is clear.

You can't solve a physical world infrastructure problem with software era optimism, a slide deck, and a schedule that pretends uncertainty does not exist. So today we'll talk about why nuclear projects keep running late and over budget, what the data center world may be misunderstanding, why SMRs are not plug and play, and what it would take to actually build nuclear grade infrastructure at scale. With that, Todd, welcome to the show.

Todd Zabelle (01:35)

Well thanks very much, Wes. Very excited to be here and boy, that was an excellent introduction. Let's jump into it.

Wes Ashworth (01:41)

Absolutely. Excited to get into it. Obviously, this is a very hot topic and excited to talk all through it. So you have spent really decades around complex capital projects. What first convinced you that most big projects are already failing long before anyone publicly admits it?

Todd Zabelle (01:56)

Yeah, so you know, this is not really that difficult and it's profound that we find ourselves in this whole situation. Going back over a hundred years ago and to recent times, we believe as an industry, and I'm gonna say two industries, construction and then what they do in the project management world, that we could create a schedule, predict what's gonna happen years in advance, and control to that schedule, and then we're gonna measure our progress based on how much we spent. Now, this is really important. So between

earned value analysis and percentage completion accounting method, it basically says if we have $100 to spend and we spent $80, we're 80% complete. The whole construct is absolutely, as the book says, built to fail. So we don't need to have any it's called a crystal ball to figure out that all projects fail for the most part and they're gonna continue to fail. I think McKinsey puts these major capital projects about 98% failure rate.

Wes Ashworth (02:46)

Yeah, it's such an important place to start. And again, I think that public usually sees the failure really late, you know, in that delay, the cost of run, or the dispute is often just the visible symptom. And so I want to get down to kind of the root cause of this and dig into it a little bit. And as you just noted there too, as well too, like you said it's profound that people believe they can create the schedule in advance and simply track progress against it. Why is that belief just so dangerous? Like what else is there to add there?

Todd Zabelle (03:11)

Well, why is it there and why is it dangerous? Let's just say why it's there. I actually started doing some research on this and I came to this idea that's a tribal mentality. We all know that the schedule's not gonna work, right? We know that it's flawed. We at the Project Productions Institute in SPS, we've actually used queuing theory and some mathematical equations to prove that it can't work mathematically. Okay, so that's out there.

Why do people still believe in it? I think it's a tribal thing and it's also a cottage industry. There's tens of thousands, hundreds of thousands of people that are in the business of making schedules and collecting data. I'll just give you one story that was a mind blower. We're all involved in a project in Guam when they're gonna do the base out there. And one of the admirals said, you know, what's your role out here? And the guy said, I'm the scheduler. He said, And he said, Let me tell you what I'm gonna do. I'm gonna give you a schedule, it's gonna be 60 days.

in the rears and you know, be so much accurate and the Admiral says, Well what value is that to Said, sir, that's not a value to you. That's just a contractual obligation that we won the contract on, right? So I mean it's pretty crazy when you think about it, right?

Wes Ashworth (04:11)

Yeah, I think that story kind of sums it up perfectly. It gets right down to the heart of your critique. You know, and I think that yeah, I like the kind of tribal aspect to it to as well. People kind of married to that just because it's everybody else is doing it, it's how we've always done it. But I like kind of shaking those norms as well, too. So we'll touch on your book a little bit, which great title called Built to Fail. What did you keep seeing across projects that made you conclude this was not

You know, bad luck or a few bad actors, but a system producing exactly what it was designed to produce? Like what was that early on thing that really led you to say, All right, time to write a book and trying to tend to get into this even more so?

Todd Zabelle (04:45)

Yeah, it's just so let's just build on this what we might call the recipe for disaster, right? So we have this idea that we could create a schedule. We're going to control to the schedule by measuring progress based on how much we spend. And then the really interesting thing, and what we've learned through as we started take the field of operation science and apply it to this. I'm not gonna get far into that, but I'll explain one thing. What operation science tells us is the higher the amount of whip that we have, or the more stuff going on in the project. And if you're the owner, everything's work in process to you, not progress, work in process.

The higher we drive the whip, right? And the higher we drive the capacity utilization, the more cars we put on the freeway, the longer it's gonna take. So what do we do? We put together a schedule, we measure progress based on how much we spend, and then we incentivize everyone to do what we call burn and earn. So we have productivity consultants that are trying to figure out how do I get people to work harder, more time on tools, how do I get more stuff started, how do I get more people on the freeway. And so the whole thing, when you step back and look at it is designed.

To produce exactly the results that we get. And not only is the result that projects are taking longer and cost more, they're consuming more cash. And as you said earlier, we don't know what's gonna happen until the end. And that's when it takes the last 90% to do the last 10%, right? Takes 90% to the last 10% of work because the whole thing's completely flawed. Now living in that vibe as a contractor back in the 90s, it just became very difficult to figure out when as a subcontractor would you ever even be able to get on the project.

Because the schedules were swinging so much, right? Each way.

Wes Ashworth (06:13)

Absolutely. I think that whole distinction really matters. And in infrastructure, we often blame the contractor, the owner, the regulator, the market cycle. But if that same pattern just repeats across different sectors and different project teams, and the system itself deserves some scrutiny and is what we're looking at. So you mentioned this a little bit ago, and part of it, but you've called out CPM, earn value, and percentage completion accounting for listeners who are not project delivery specialists.

How do those tools create false confidence? Like what else is to that in that story?

Todd Zabelle (06:41)

Yeah, so I mean there's just so much to these fundamental things and you know, I we don't want to be beat a dead horse, but I mean first of all, let's start the first thing, which is we're gonna create a schedule. Okay. Ha yeah, how do you actually create a schedule? And you know, there's a element throughout the world due to insurance and other

situations where the people that do the design, the architects and engineers, depending on what kind of thing we're building, they're not allowed to get involved in means and methods. The people that know the most about what you're gonna build are the people that design it, believe it not. But then a third party comes in and says, I'm gonna create a schedule based on what it is you're doing at some particular point in time. Now, these schedules, how are they created? We're involved in a massive project. The scheduler was in Manila, and the project was somewhere on the west coast of the US, okay?

Never been to the project where it was creating the schedule. So now you have these people that are creating schedules that are completely disconnected from what's being built, and now I guess where it's being built. All right. And then everything starts to build on top of that, including the claim at the end of the project where the attorneys and the consultants, the accountants get involved, and they're trying to figure out who did what on the particular project. There's a whole aspect of this thing that you know I spent my life trying to figure out, which is on the production, on how do you actually design, make

Move stuff around, install it, construct it, and commission it, which almost isn't even part of the conversation, right? It's very interesting how this is how this has come to be. And so the universities teach this idea of project management being an administrative function. And so there's a lot of administration going on, but not a lot. The technical work for whatever reason, whether it's engineering or actually welding, has lost its luster, right?

And so we sit we'd say on projects, you know, you have people that are journalists, submittal stampers, and so on and so forth, because they focus on that aspect, not really the technical.

Wes Ashworth (08:23)

Yeah, very useful translation for the audience. I think, you know, many people hear even those terms and they assume it means, you know, discipline, but that real question is whether they really help leaders understand the actual work or whether they mostly just create a polished version of uncertainty, which you've started to talk about. You also mentioned too, so you know, you've said that project management is mostly administrative. What is missing? That I guess you haven't mentioned yet so far, like what is missing from the way projects are actually managed today?

Todd Zabelle (08:48)

Yeah, so let's go back to the Project Management Institute. And they say that the purpose of the project is to create a need or to take a need and create an outcome, service, product, whatever the case may be, right? In this case, let's say an asset, if we're talking energy type projects. What they don't talk about ever is how you actually and do the engineering or the design, how you make things, how you deliver these things, how you install them like we said, what I call the value added type elements of a project. So there's almost we for over the last forty or fifty years.

We've gotten away with that, right? If you just look at what a project manager earns or even a scheduler, a scheduler's probably the billing rate's two times of a structural engineer. Now, if a scheduler gets the schedule wrong, what happens? Not much. If a structural engineer gets something wrong, people die, right? It something could collapse. So I think really at the heart of this thing, we've just gotten away from how do we do the work? And we don't appreciate the people that do the work, whether it's technical

knowledge or craft. It's just not respected anymore. It's not part of the equation. That's just some other thing happening on the project, even though it's the purpose of the project, right?

Wes Ashworth (09:49)

Yeah, without a doubt. So what does it actually mean to you to manage the production system instead of just managing a schedule?

Todd Zabelle (09:55)

Yeah, so that's a great question. This idea of schedules are a bunch of tasks that are interconnected and we can you know, going back to Kellyan Walker, we could figure out how long the project's gonna be by looking at this critical path is crazy. At the end of the day, what we gotta do is we've gotta identify the requirements, create the design, produce design in some kind of and this is changing as thing goes on, in some kind of format that people that are gonna go make the stuff and install the stuff

Could read it, understand it, and execute it, right? That is what we call the production. So those are all different types of production. So it could be the structural system from requirements through it's complete. It could be just the assembly on-site. It could be the fabrication off-site. We could define production systems how we want, but they always have this idea of things flow through them and they use resources to make things happen. And that's a very different view of how people in the project management world would look at it.

That there are these production systems and that managing the production systems is really the key. So just to wrap this up, the schedule tells us what the demand is. The production system will produce to that demand, whether it can or can't, will dictate what's gonna happen on the project. So the schedule is nice to have. The production system will dictate what's gonna happen at the end of the day. It's that fundamental.

Wes Ashworth (11:08)

Yeah, absolutely. That's a shift I really want listeners to understand. I think the goal is not to just have a better looking plan. You know, the goal is to manage the work in a way that makes constraints visible, reduces variability, protects flow, keeps the project grounded in what actually what is actually happening. So we've covered that. I think that foundation matters because nuclear may be one of the toughest tests of any project delivery system. You know, the tolerances are tighter, the quality expectations are higher, the operating requirements are more specialized, and the consequences obviously of poor execution are more severe.

So now with AI and data centers are adding that new layer of urgency to that already difficult equation. And so with you, I know you've spent some time, you're just coming out of Datacloud, where nuclear for AI was a major topic. What did you hear that felt grounded in reality and what maybe felt like the market getting ahead of itself?

Todd Zabelle (11:55)

Yeah, that's that was a very interesting experience. And you know, I'm not a big cultural guy, but let me just put this out at just because I think this will help your listeners understand. You have data center people that you know these guys were burning disks and then had downloadable software that decided at some point we should put everything in the cloud and go from a very low capital intensive business to a very extremely high, the highest ever in history capital intensive business. And their gung ho people are gonna go get it done, right?

Then you have and people probably in their solar system, the bankers and the rest that are right there with them because they want this to succeed. You have the real estate developers who are really the guys that build the data centers, if you will. It's kind of a real estate general building contractor type group of people, right? And then you have the people that put the technology in there. When you bring the nuclear people in, they're in a whole different world. They seem more like the original technology people. Now I was born and raised in San Francisco, I grew up here in the middle of Silicon Valley.

When you look at the people that were the technology, the computer people back in the day, long time ago when I was young, those are physicists and engineers and computer scientists. That's who you have in nuclear. So what's interesting is you got the hype coming of I'm a data center builder, I'm a hyperscaler, yeah, I got I want my power, and I want my power now, and damn the torpedoes, what's the problem? And the nuclear guy is saying, Look, we're not writing software code. Part of what we do is that'll be included in the project.

But we've got a lot of things that we need to go through here because a lot of people could get dead really fast if we got this wrong. And most of the places where we want to put this, they're scared to death that's what's going to happen, anyways, because of things in the past, right? So I would say, and I don't know if I'm answering the question directly, but this cultural difference is really fascinating between nuclear people that are very conservative, measured, go slow to go fast, and

People are more on the digital infrastructure side, which are damn the torpedoes, we're gonna get this done. And it'll be interesting to see what happens. So, what you hear from them is the damn the torpedoes from the real estate data center type crowd over there. You hear the nuclear guys saying we gotta be measured, right? This is serious stuff. And then you bring in another group of people that have shown up, which is the civil engineering people and the people that do concrete and grade sites.

They don't they haven't done a lot of nuclear work because there hasn't been a lot of nuclear. So what I saw on a very large nuclear project the other day was someone said we have a cultural problem between two different countries. I said you don't have a cultural problem between different countries. I propose you have a cultural problem from guys that put in rebar on concrete and move dirt around and guys that are in high precision nuclear type systems. So I think really what you got here is a huge buzz for all this wonderful opportunity.

But the reality is this is a little bit more complex and people appreciate.

Wes Ashworth (14:32)

Yeah, that's perfect. That paints the picture. And what I was kind of hoping you would get into. And I think it's the readout people need right now. I think there is this real urgency around power for AI, but there's also a risk that excitement starts moving faster than execution reality. So separating those two is very important. I think understanding some of those even cultural differences, as you just outlined there, that really does paint a great picture of what's happening. So we know this data center world is desperate for power.

and nuclear companies are excited to provide it. And you started to talk about this little dissonance too between all those groups, but what are both sides underestimating about the path between interest and actual deployment?

Todd Zabelle (15:09)

Yeah, this is a complex systems engineering problem, all right. So the first thing we gotta get our head around is the demand profile of a data center is a lot like this, right? In milliseconds taking power. And you know, a power plant just wants to go at a base load and go down the road and be happy. I learned something that I didn't fully understand, which is you know, you need to connect a nuclear power plant to the grid because it needs power to start up and it needs power for backup.

Okay, so a nuclear power plant is still connected to the grid. It's gotta be plugged in the wall too. Okay? And I didn't fully appreciate that until I understand how all this works. Okay. So I think you gotta understand that there's not a you can't just plug a data center into a nuclear power plant, and a nuclear power plant needs to be plugged into the grid for a variety of reasons, the way I understand this, okay? Now you're gonna have to get some kind of technology.

That goes between the nuclear power plant, whether it's an SMR or not, and the data center to deal with this modulation and what how the demands current compared to the baseload. Okay. Now that power plant's gonna keep producing. So if you want to go out on the grid, the grid operators don't want you going in and out either, right? Because they're trying to keep that thing stable. So I think you have a somewhat significant.

systems engineering problem. The supply chain has been decimated. So you have a fuel challenge. Now we heard in con the most important challenge, the most critical challenge is fuel. How are you going to deal with fuel? You need to get fuel. And there was a whole presentation on fuel that was quite fascinating how that works. And the supply chain for fuel has several players that are involved in it. And the customers of fuel are utilities and you just don't show up and start buying fuel.

Yeah, it's not potato chips, right? So that brings some complexity. I think you need to appreciate that we don't have enough electricians and welders and so on forth so forth to deal with data centers. Those humans aren't at the same level that you need when you start messing around in data center with nuclear. That's a whole nother level. You need the cream of the crop of welders and so on and so forth to get involved at that level. Okay, so there's that's just three things that I think we need to

understand and appreciate. Now whether the data center people were listening or not, I'm not certain.

Wes Ashworth (17:20)

Hopefully they were, but it's a helpful way to kind of frame the gap and sort of really understanding the complexities that are going into this. And you've described this sort of matching problem between data centers and nuclear plants. To like what has to match for this really to work in the real world?

Todd Zabelle (17:34)

The first match you gotta have, because there's demand, right? So the first match is you gotta find a place where they're gonna allow you to put a nuclear power plant to begin with. Now, a lot of people think an SMR is like your refrigerator. An SMR require anything nuclear requires containment. So how big is a site for a reactor? It's not three meters by three meters, okay?

So you gotta find somewhere where you could put this thing, right? And you gotta have people that are gonna want allow you to be there. You also are gonna have to bring fuel in and out, right? And so you gotta have the ability where they're gonna be happy with that stuff happening in their community. So the first match is you gotta have a place to put it. The second match we just talked about, which is the systems engineering. The third match is you're gonna have to be able to get all the bits, right, in the supply chain to get it all there so that you can build this thing.

And you're gonna need this the part of the supply of the labor. You're gonna need to match, create the supply chain to provide not only the fuel, but the people to operate it, because we've decimated that as well, right? And then one of the other matches to me, which is really interesting, is the life cycle of a nuclear power plant is measured in decades, and a data center in maybe less than a decade. And what technology is coming out that might replace

power, whether it's quantum or whatever the case may be, right? The power demand. So I think there's a match on the actual design of the network, meaning, well, if we put in a bunch of SMRs, are the data centers always going to be there to take the demand?

Wes Ashworth (18:59)

Yeah. Again, I think, you know, understanding all that, you really start to see again, just how much complexity is there. And it's not just a simple question of like whether nuclear is clean and reliable. There's a lot of other things that we have to consider and look at here. And you said this earlier, you know, that data center load can change in milliseconds. You know, while that nuclear power plant wants to run as a steady base load. How big of a, I guess, a design challenge really is that mismatch as you see it?

Todd Zabelle (19:24)

Yeah, so the design challenge is you gotta put something in between, whether it's batteries. People again were talking about we could send it onto the grid, but the grid guys are saying we don't want that thing jumping up and down like that. We're not your solution. So there is a research and development ongoing. There is research development ongoing on how do you buffer that in between there, right? And is it batteries, what kind of batteries is it, fuel cells,

People are working on that problem as well. So you're gonna ha so that's gonna require some development as well as part of this.

Wes Ashworth (19:52)

Yeah, absolutely. And I'll think about this kind of like in a practical sense of you know, something playing out in the real world is so if a hyperscaler came to you and asked, you know, we want nuclear power for AI capacity, what would you ask before you ever even talked about reactor technology?

Todd Zabelle (20:06)

Yeah, I would say first of all, do you have the time? Because I would say that these SMRs are five years out, all right. And you know, the other thing is so you take let's take Sizewell being built in the UK right now. That reactor would power one of these one gigawatt data centers, right? That's a fifteen or twenty year situation, right? So do you have the time, which I don't know how many do. Do you have the place?

How are you going to deal with this systems engineering problem that we're talking about? Can you get a permit? All right. So a lot of people don't seem to be as concerned about the permitting. That's something that we keep kind of putting under the deal. And I get that the federal government's moving quick on that, but that doesn't mean your local government or your state's gonna go along with that wherever you may be in the world. So I think it gets back to the match. I would just go through exactly what you're saying, the questions of the match and say how many of these things can you match?

Wes Ashworth (20:52)

Yeah, I think it's a good discipline for the market. You know, before anybody picks a technology, they need to understand all those things, like timing, location, system design, demand profile, whether the site and communicating realistically support a nuclear project as well. I mean, there's a lot there. We've touched on SMRs a bit there. And I know, you know, you've said SMRs offer some real advantages, and I think everyone would agree. Maybe not the easiest plug and play solution like some people make it out to be, but you know, it can include some smaller batch deployment. I guess from your perspective, like

SMRs, what do they solve and then what do they not solve?

Todd Zabelle (21:23)

Yeah, so I would say that where SMRs really add a lot of value is exactly what we said, this lower batch production. So Oklo and those guys, I think they're down around seventy-five megawatts. I believe Rolls-Royce is up over three hundred. But this idea that you can build out as you would the data center over time by adding, let's call these, you know, the reactors is very interesting and it's a better way to go. And you know, this is what the Japanese did to us in manufacturing. They showed us the

lower or smaller batch production beats large batch. Okay. So that's really the attribute that I think people are excited about. But I think the challenge is, and we see this in all new tech, okay? So my company gets a call every few months on somebody's got a new thing, pyrolysis, RNG, you know, solar. There's always something that someone's gonna deploy. And it's always the same story. What we're gonna do is we're gonna build it in a factory and we're gonna deploy it. Well, it's more difficult than you actually realize. I think there's a great

case study on BP when they changed the name from British Petroleum to BP, it was gonna be a billion dollars to do all the signage. I think it was seven billion or whatever it was, don't quote me when they were done, right? Deployments are extremely complex things. Because you can make something in a factory doesn't mean you could deploy it. And the question becomes when you go to reference design or standard design, how far, how much of it's standardized and how much of it has to be localized due to building codes. We were working the data center company a long time ago and they had a standardized design. I think they're

What do you call it? The red conductor for the hot lead, if you will, the positive was red. They went into somewhere in Texas and they said, no, we're blue. And so their whole design had to change for just the wiring in this thing because of a local municipality. So I think there's challenges with deployment. And I'm highly confident that modularization and offsite assembly, though it may be good for the core element of this thing, the reactor or whatever you're calling it.

The other parts of it aren't gonna be any less challenging and probably will be more challenging than what we're doing now.

Wes Ashworth (23:16)

Yeah, that nuance is important. I mean SMRs definitely a lot of advantages there, but good to understand the full picture and how that all plays out. I know some examples like Oklo are and others are talking about smaller units that could expand with demand. Why is that appealing for data centers? And maybe why again, is it still not just plug and play?

Todd Zabelle (23:34)

Yeah, so again, what the Japanese taught us is if we could reduce the work in process, we could get the cars out faster. So what we're doing here is we're actually deploying at 75 megawatts, let's say, or 100 megawatts, whatever the solution may be, and we could bring that data center on and get to revenue faster. So if we have to build a you know AP 1000 or 1100, whatever it may be, reactor, and it's gonna take 20 years, we've got to wait a long time to get to that.

If we could get deployed and let's say things are up and running and if they're up and running and it only takes six months to get, let's say, this is the you know, ten years from now, or whatever, but if you can get one delivered every six months and you're bringing on 75 megawatts and your build cycle for the data centers is about as matched, if you will, like you're saying earlier, then that's you're getting the revenue faster, right? I think the challenge though we talked about that other part of the question remains.

it's not really the factory that's gonna be the issue, it's how you deal with the site. And then the decisions on the site. So, you know, this is where people they don't understand until they get into it. So if we're going to build out let's say a gigawatt of 75 megawatts, do we clear the whole site and prepare the whole site for the gigawatt of the 75 megawatts, which is clear if you're gonna do a gigawatt, right, of power. So there's a lot of factors in here and costs. So you may say, wow, our civil costs are really expensive. Well, you build civil.

for what inevitably be will be a billion. But then if you go back down or for a gigawatt, rather excuse me, for a billion or whatever it may be, and then you go back down and you say, well, I only want to deploy 75 megawatts, but then you got containment around the 75 megawatts. Now you may have a real challenge with how do I go in there start grading again. So again, this is more complex than people realize.

Wes Ashworth (25:11)

Absolutely. And that's why we're bringing this home a bit and just like really do spelling it out. Cause I again I think way more complicated than most people realize. And I think that answer really gets to the appeal without letting the conversation become hype, which I appreciate. The smaller units may fit the growth curve of the data centers better, but again, that surrounding infrastructure and nuclear great execution still really matter and all the other things we're talking about. So we've painted this picture. Nuclear sites are often larger and more complex than people realize.

Anything else, you know, in terms of like what the market tends to misunderstand about the physical footprint or infrastructure around these projects that we haven't hit on?

Todd Zabelle (25:41)

Yeah, I think we've probably hit most of it. I think part of what you're saying in that question, which is interesting to me, is the excitement and the for it without really understanding the technical challenges, if you will, that are all part of it. Now, I learned something because I'm not a nuclear guy, I'm a project guy, like you said when we started. Tony Roulston who's a friend of mine and who used to run the nuclear program at Rolls Royce and now is over at Cambridge.

There's a lot of different types of reactors, even for SMRs, okay, and different type of reactor technology. You should get Tony on the show because there's whole thing in just understanding that and it was quite fascinating on how those reactors work. Lucula technology's been around in the Department of Energy, the DOE in the US like fifty or sixty years. And what they're doing is they're commercializing that. Okay. And that's a certain kind of reactor that people are taking advantage of. But there's

Many others, right? The other thing that was interesting, just give me some riffing on some data there, I think there's like a hundred and eighty SMR companies in the world. Hundred and eighty companies chasing SMR technology, right? Only some of these guys can make it. Now I don't know who the judge and jury and who's gonna make it, but you can't have a hundred and eighty of them out there chasing permits and doing design and the rest, because there's just gonna be lot of confusion. So I would say if I'm looking to go going back to your other question, if I'm gonna go get involved in

Todd Zabelle (26:54)

SMRs for data centers or SMRs for anything, there's just a certain amount of getting up to speed on the technologies out there and what the pros and cons are of the different nuclear reactor technologies at you know at the core.

Wes Ashworth (27:05)

Yeah, absolutely all great context there. I appreciate you kinda like just sharing a bit and we'll have to feature another episode to get more into to some of that nitty gritty as you said. So you have again sort of like as we're kind of going through, I want to make sure we covered different pieces of this.

Another thing you noted, is you know that fuel came through as a major issue at Datacloud. I wanna just dig into that a little bit more. So why is fuel supply just such an important constraint in the nuclear scaling conversation?

Todd Zabelle (27:29)

Yeah, so first of all, it's been decimated, right? The people that really do anything with nuclear fuel these days are utilities and the DOD, right? Or I guess the DOW now. So what's interesting is fuel, and you can't quote me on this because I don't totally recall, but I would say there's four or five players that create the fuel. It has to go through a process of extraction.

And then there's enrichment. And then you have people that fabricate it. And then it finally moves into where they're gonna go ahead and use it. And then of course, when it's done, it goes back out. There's just not the capacity in the market for the stuff. We talked with a guy that does the enrichment, that was quite interesting. I spent lunch with that guy and he was explaining how it works with the centrifuges and the rest, right? And they just there's just not the capacity to go

do what needs to be done if these SMRs come along. Now I think they could build the capacity, but these are billion dollar plants as well, and very technical on how these centrifuges work, and they're not telling anybody how they work because they have a closed business. But basically all your customers for fuel are either the Department of War or utilities, and it's all kind of connected and everybody's happy. So it's I think it's like you know, it's like a leveled system that's all working very well. It's kind of like the grid. Everybody's happy, they're all moving down the road, and now we're gonna disrupt that, what's that actually mean, right?

But it was very clear that the biggest challenge in nuclear fort is availability of fuel. That was across the board very well communicated, right? And so that's probably one of the questions people need to understand is if you're gonna talk with one of these SMR guys, is where you're where's your fuel source, right? Where are you gonna get the fuel?

Wes Ashworth (28:59)

Yeah, great clarification there. And again, to your point, like that may be one of the most overlooked parts of the story or the most overlooked part of the story. And I think people talk about reactors and demand, but fuel supply determines whether many of these deployment pathways can actually scale. So that's a huge one. And I want to go back to something you said earlier. Like people talk a lot about construction labor, but you've raised the issue of people qualified to operate nuclear plants, a higher level of skill. When we talk about, you know, welders, electricians, skilled trades, you know,

As our recruiting firm, like we started taking on a lot of that work for companies because they're like, we're desperate. We need these people bad. Can you please help us? And so you're seeing that happen. And now you're talking about just the cream of the crop, as you said. Like, tell us a bit more about that. Like, how serious is that bottleneck? And maybe what are what do you see could be done about?

Todd Zabelle (29:41)

Yeah, so you know, I have a lot of passion for this and I was actually on the board of a school that I was resigned from because I kept pushing this was I don't know, probably a year or two before COVID. We need to get back into the vocational stuff, right? We really need to do this, and they thought it was crazy. But you know, being in the software business, I kinda know where things are headed and I could see this AI thing was coming and so

The first thing that we need to do is we need to understand that we've got to get the vocational stuff in there to find the people that actually are interested in this, right? And you know, welders now are making more than other professions, right? So we need to get people that actually and we need to celebrate and say not everyone has to be a venture capitalist or hedge fund manager. We need to say if you're a welder or an electrician, that's a good thing. And I think there'll be more robotics than the rest, So it's not gonna be as labor intensive. But

We need to celebrate that. Then once we have the people that are interested, we need to put the means in place to develop these people into what needs to be done for nuclear, which is not the same as a guy working on a house. We're not talking carpenters banging nails. We're talking people that need to be able to go in and do work in a live facility that you just can't take the risk of what may happen. Now you might have 10 75

megawatt reactors, whatever the case may be, but you just can't have anyone roaming around the range there. So we've got to go create these people. And it's almost, I believe it's gonna be like special operations in the military, if you will. It's gonna be the guys at the top of the top when it comes to welding, electricians, or whatever, that are gonna get into that. And so there's a whole program that needs to be put into place to support that. It's a real challenge. And I don't know if anyone's really working on it yet.

Wes Ashworth (31:17)

No, I agree. That like this is a topic near and dear to my heart. We see this from the inside all the time and just the need for those individuals, how critical that is today, that is only going to exponentially get worse and worse as we go. So it needs to start like you said, early and often in schools and kind of getting people interested in and starting those careers because highly rewarding, high paying jobs, high demand. You can kind of write your ticket, you know, if you're a skilled tradesperson. So agreed. I wanted to definitely make that point.

Todd Zabelle (31:44)

this is how serious this is. It was about maybe 10 years ago. I went when Stevie and I had bought Shaw and I went to one of those fab shops, that big fab shop in Louisiana, I think it was, and they actually had two or three psychologists on staff. Okay. And this welding is so critical that if Wes or Joe are having a bad day.

Because they got in a beef with their wife or whatever the case may be, they have psychologists to go there and counselors and talk with them just because it comes out in their welds, right? And so I just don't know if people really appreciate how technical this is. Now I've decided after many years my life and knowing a little bit about welding, I said I'm gonna go learn how to TIG weld. And a friend of mine said, Hey, look, if you could TIG weld stainless, you're like a made man in the welding mafia. And I've learned how to do it.

And I can't tell you how complicated this is. You almost have to be like an athlete to be able to do these kinds of things. So we're looking at some very special people that understand math, physics, metallurgy, material science. They have the ability to hold a foot pedal, to move something this way, to see, to just even the welding of putting the rod and how that needs the filler material, they call it just to learn that.

And have the skill to do that. It's an incredibly complex thing. And I have a lot of respect for those guys the more I physically, you know, develop my skill and just for fun, because I wanted to see what it was all about.

Wes Ashworth (33:10)

Yeah, I love that story. And that's a company that gets it, right? You know, it's to that point, like you could kind of compare it to like a surgeon. Like it's the level of precision and care and skill that those individuals have to have at the highest levels. I mean, it's it is absolutely incredible. And so we need to value them more, hopefully encourage more to come in and start that path as well too. Another thing you kind of hit on was mentioning this sort of nuclear grade work.

And so we look at things like rebar tolerances and welding standards and quality standards, like how different truly is that from the broader just construction workforce challenge?

Todd Zabelle (33:43)

Yeah, so let's take we're involved in a couple of these things now, and it's interesting. Rebar has to be within a tolerance of one millimeter. So, you know, maybe you're taking a 20 or 30 millimeter thick piece of rebar and it's gotta be lined up within one millimeter. It can have no corrosion or rust on it whatsoever. So let's just start playing this out. So what do guys that do rebar work, right? Steel fixers, whatever you want to call them, iron workers, yeah, whoever the guys, rod benders, whatever you want to call them.

These guys aren't dealing in a tolerance of one millimeter, right? And then if you want to take and do the work off site because you want it in a factory involved environment, now you gotta weld it. So let me give you a for instance, okay? We're involved in a product the other day and they're trying to figure out how do you weld two pieces of rebar together. And I'm like, well, you know, that's pretty the first thing you gotta do is in welding, you gotta cut the chamfer in the thing, right? So you could weld it. Well that reduces the

strength of the rebar. So we can't necessarily do that. So there's challenges out there to say, we're going to move the work off site and we're going to modulize this thing. We're going to go build precast panels that we're going drop in. Hey, the guys are actually doing it are trying to figure out how the hell do we weld the rebar, right? And have it within a one milimeter tolerance. Now the next thing we start thinking about is I was on the phone about this with an engineer buddy of mine this morning. And we were talking about something else they said, but I'm really intrigued with how do you take and mount mechanical equipment

The concrete. You can't just go out there with a roto hammer and start drilling holes and putting epoxy anchors in because you're talking about rebar, it's one meter apart. You know, and so it's just a different level of complexity across the board and standard that nuclear is held to compared to everything else. And how do you get the people and the thinking to get up to that? And I think that's why this cultural thing that I was talking about earlier, the civil guys are like, we're just putting in rebar and

the engineers in nuclear are saying, Yeah, it's just rebar, but make sure there's no corrosion, it's gotta be one millimeter with intolerance, so on and so forth, right? I mean it's really serious. It's very interesting.

Wes Ashworth (35:26)

Absolutely. And I think, you helps listeners understand why more workers is not a complete answer. You know, nuclear requires that different level of discipline, documentation, quality control, technical execution. The good thing on the plus side of like that means typically higher paying jobs and higher demand. And you know, people want to say, like, AI taking jobs or what have you. Like, I think it's creating new jobs, creating new opportunity, higher paying opportunity as well, too. And this being one of them. So again, I like that you've gotten into that.

We'll bring it back a little bit to the project side. So if you looked beneath some of these like major headlines that are out there, these projects, what failure patterns maybe would you expect to find in some of these that we're seeing?

Todd Zabelle (36:04)

Yeah, so here's what's gonna happen, and I think you're gonna see this on the data center side, though no one's talking yet. But I'm extremely confident about this. If we go back to where we started and we say we got a guy that creates a schedule, okay, man or woman, and then we're gonna use earned value and percentage completion accounting method to measure progress against that schedule that probably has no reality with what's happening in reality, right? There's no relationship there.

And then we're gonna declare that we're 90% complete based on how much we spent and we're way over budget. I think you're gonna start seeing, you know, for lack of a better word, the shit hitting the fan here pretty soon on some of these projects. Now, I think the nuclear guys have had so much of that, they're a little bit different and they're probably a little bit more sophisticated. I think the oil and gas guys who are coming across, who are probably the closest to the nuclear, right? They live at that level of sophistication, especially offshore. I think you're gonna start to see some

people are adopting more of this production view that we're advocating. But I think you're gonna see the challenges on the cost and schedule overruns. The thing, you know, this is just my experience the last couple days. I've been using some LLMs, I won't say which brand, but basically all of them, because I can't get enough capacity, no matter how much money I spend, invest in, these services, I can't get the thing to run. It keeps kicking me out saying come back three hours later, right? And it's driving me crazy because now I got a bunch of boxes open and whip going on.

So we're not gonna get the infrastructure deployed fast enough, but we're gonna believe we are until we're not. All right. And here's the big headline to me. Now I just saw a thing come out from Delaware this morning that we're looking at data center investment of north of a trillion this year, right?

This is not a throw of money at problem. It's just not gonna happen. It's not gonna happen. And what's happening now, you know, at some point we're gonna see, despite all the optimism, these projects are delayed, over budget, and things gotta change.

Wes Ashworth (37:49)

Yeah, absolutely. And I think this is where the industry can actually really learn and hope kind of paying attention to this and seeing like hopefully we can solve some of this before it becomes too much of a problem. And again, to your point, you're seeing those things already start to trend that way. That you're probably under understanding what's happening, you know, underlying and some of those issues that are gonna cause these massive delays and things like that as well. Too. I'll ask you this: like, what does a capable owner look like on a really complex project?

And where do owners usually give away too much control?

Todd Zabelle (38:18)

Yeah, so that's an excellent question. And at the Project Production Institute, and people go check that out. It's a free thing to join. We basically put all our IP out there as a open source thing. But at the Project Production Institute, we're have Gary Fischer, who was involved in capital projects at Chevron for 38 years, he's the executive director. He's releasing a paper that we just produced on that with a kind of a maturity matrix benchmarking thing. And it's really an important question. And so

And it's important to understand the history. In about the it started in the 80s, but in the 90s and early 2000s, this idea of getting to your core business made a lot of companies that had capability to do things, get rid of their competencies in engineering and in construction, right? And you think about capital intensive businesses, that is your part of your core, but they were convinced that's not your core, you can go hire an EPC firm. And then the EPC firms, they got out of doing direct hire stuff. And so what you have going on now is a whole

mishmash of people that come from different companies. You look at a big project and the engineering guys show up, you know, they used to work at a different company. They don't even know how this company necessarily operates. Okay. So that's kind of the setting of the scene. I'm gonna just you could pick this up out of public information. The startups that are involved in AI can't raise enough money and if who do they raise the money from?

They raise money from people that have a vested interest in their business, which is a good thing, right? However, they have to make commitments to raise that money. And so they have to buy things from them, right? And they have to have a place to put those things. And so the people that are providing the funding, right, for some of these guys, whether it's NVIDIA, SoftBank Energy, or whatever, they want in on the action. All right. So what you might call some of these guys is ecosystem.

orchestrators, right? They rely on a bunch of people, or entities rather, some of the biggest in the world now, to go do stuff for them. So the question is, how do you coordinate? How do you make that happen? And you can see a lot of frustration in at Datacloud where people are saying, I can't rely on these guys, I gotta take it back in house. But if you look at how the funding model is working, they can't take it back in house because they made commitment to a bunch of people, right? So I would say look for that paper. It really lays out like

40 different things to go do. But at the end of the day, at the end of the day, the most important thing that anybody who's been in a position, like let's say Gary has, the owner is ultimately responsible. So whether you're rebuilding your house, making a new house, you're building a liquefied natural gas plant, power plant, data center, or you have to live with what you got. And it is extremely important that one, you're clear on your requirements, two, you stay involved in the technical

development and the engineering and the design, right? The supply chain you want to have your fingers in to make sure. So I always talk about you get down to third tier decisions, if you will, and you pick out the stone or the tile, but the guy decides what the grout joints are going to be, right? And the next thing you got a half inch grout joint on this beautiful stone because he was trying to make it fit so he didn't have to cut it, right? So you gotta be involved in the design and the procurement. You got to be paying attention to what's happening.

A friend of mine built a house the other day, and the guy got the concrete wrong, and now he's got to step up to go into something in his house that he didn't even think he was gonna step in. But yeah, you got to pay attention to what's going on, right? So, do you need to do everything? No, but you better have your fingers in the thing because at the end of the day, you're the one that's gonna be held responsible. And this is the last thing I'll say about this: every contractor's got one foot in the grave, and these projects were so big nobody couldn't can't underwrite them, anyways, right? And so let's go to nuclear.

What happened? The French government took EDF over because they needed to keep them in business because they gotta go build twenty reactors. So nobody can take the risk of what these projects are all about.

Wes Ashworth (41:54)

Yeah. That owner role is so critical. So I appreciate you going through that, highlighting those things and where that where the owner needs to really stay involved, especially as we're seeing that unfold and some of those evolutions of just how the industry has gone through those iterations. Another thing too, I mean, it we've touched on this a tiny bit, but you know, AI, automation, digital twins, industry 4.0 tools, they're being sold as answers to project delivery problems. Where can they actually help?

And where do they become, you know, expensive decoration and not quite getting the full job done?

Todd Zabelle (42:24)

Yeah, a few years ago I wrote a paper on this. This is great question. To me, you have to get to the production. So if you want to automate the administrative work, I guess you can automate the administrative work and I could tell, I don't know, chat GPT to send you an email. That's great. But at the end of the day, if we're gonna drive value, which has to do with getting projects delivered more predictably, more economically, faster, safer, environmentally friendly, whatever the case may be, we gotta focus on the production.

The challenge is the humans don't understand the production aspect of projects, and so they don't think to apply AI to that portion. But that's where the value is gonna be. So when start taking IoT sensors, an IoT technology that's gonna give you telemetry and tell you where things are with connected GIS, what is that? That's the world of production, not administration, right? If you start connecting sensors and get feeds from fabrication equipment, orbital welders, whatever the case may be, robots, that's production information, right? So where

these ideas are really being implemented and adding value is in manufacturing, right? In retail, in where people are working on problems that have to do with making and designing and moving stuff around, not on administration. As long as these as project management stays administrative based, there's not a lot of value any of this stuff could add anyways.

Wes Ashworth (43:37)

Yeah, good distinctions there. I think understanding, you know, where they can really help and you know, where again it might just cause more complexity or you might quite could have done. But again, that focus on production has been really fundamental throughout the conversation. So I like it coming back to that as well. And as we've talked through, as we'll get kind of to the conclusion here, but you know, the bigger issue is not whether nuclear matters. It clearly does. The question is whether the industry can move from enthusiasm to execution. And I think that's been a common theme on the show as we've talked, you know.

hundred plus episodes, but that requires a different level of honesty about what it takes to build and operate infrastructure that is governed by physics and not slogans. But when you look ahead at the next decade, are you optimistic that we can deliver the infrastructure AI and the energy transition require? Or is it only if we fundamentally change how we build and I'll just let you run with that. Any kind of like final thoughts around that and anything you want to share?

Todd Zabelle (44:28)

Yeah, I have two thoughts about that. First thought is we need to be figuring out how do we use less energy. And so I think things like quantum computing is where I'd be spending my time. So a lot of people talk about AI and you know, to me, quantum is really where you need to be focused. And I'm glad there's people out there that are focused on that. Okay. So how do we use less energy and how do we become more efficient? I think that's the role of quantum and other technologies like that. When it comes to the building side of this thing.

and the rest, I think we just gotta rethink how we're doing things. I think we are gonna rethink how we're doing things because people are addicted to their AI now, right? We're trying to get work done and we've got to deploy the infrastructure. So I think we're gonna have to solve the problem, but we're still in denial about whether there is a problem.

I think what you're saying is excellent, right? There we got a lot of slogans going out there. And if you listen to the earning report earnings reports of these companies, it's in the Q&A sessions, you know, at the investor meetings, it's the quarterly investor sessions, it's interesting to see how excited they are. But this is complex stuff and we've gotta really rethink how we're doing things to make this happen on the technology side, but as well in how we do the work, right?

Wes Ashworth (45:36)

Yeah, absolutely. And I think this is the right place to end. And this is the point we're trying to make, is the demand is not going away. We know that. AI, electrification, domestic manufacturing, decarbonization are all pushing on the same infrastructure system. So that hopeful version is this will not be easy and some things have to change. And that's the point we're making here. So it's that we can get better if we're honest about what has to change and accepting of that. So we'll end there. I think that's a great conversation and really has painted that picture perfectly.

But Todd, thank you so much for joining us. This has really been a valuable conversation. You know, I appreciate how directly you challenge the assumptions underneath the nuclear and AI conversation. The big takeaway for me is that the energy transition is not just about technology, it's also about delivery discipline, and we really have to change some things there. And thanks for to everyone out there for listening to Green Giants, Titans of Renewable Energy. If this episode helped you think differently about what it really takes to build the future of energy, share it with your network. Don't forget to subscribe to the show and leave us a rating.

And with that, we will see you next week.