In Part 1 of my conversation with economist Lynne Kiesling, we traced how monopoly utilities and central planning helped electrify the country. That model worked. Economies of scale and guaranteed returns brought capital into the system, and within a few decades, nearly every home had electricity.
But the world has changed. Technologies are smaller, decentralized, and more flexible. Risks are more complex. Consumers expect more than just “the light turns on.” In some areas, the old model now creates perverse incentives: rewarding capital spending over performance, insulating utilities from risk, and slowing innovation.
So the question is: what comes next?
In Part 2, we explore how markets can evolve beyond wholesale and retail competition to tackle the next frontier: risk allocation, demand-side flexibility, and performance-based regulation. And we look at how AI-driven data centers are testing the limits of the old model while creating new opportunities for Texas to lead.
Markets as Error Correction
Markets don’t just allocate resources, they correct errors.
As Lynne explained:
“If someone has made an investment and… we’ve built too many gas power plants, and we’re not earning profits on that, that’s a signal to me that I need to take my money and put it somewhere else.”
That’s how we avoid repeating mistakes. Yet in the utility model, many risks never reach shareholders. After Hurricane Beryl, for example, CenterPoint admitted its failures but still posted a billion dollars in profits. Consumers bore the outage costs, while investors stayed insulated.
The missing piece: markets for risk. Today, outage risk, rate risk, and weather risk aren’t fully priced or traded. Post-Uri, some generators took huge hits while others profited. That’s markets working. But for regulated utilities, risk rarely lands where it should.
Of course, markets don’t solve everything on their own. Consumers need protection against fraud and market manipulation, and regulators still have a vital role in setting guardrails. The goal isn’t to remove oversight, but to let markets do what they do best, deliver solutions faster than central planning.
Demand Flexibility
For decades, demand seemed inelastic. People flipped a switch, the light came on, and rates averaged out costs. But digital automation has changed the game. Devices from EV chargers to air conditioners to fridges can now respond to prices automatically.
“We could find that there is a lot more latent flexibility on the demand side that would not inconvenience or discomfort consumers.” - Lynne Kiesling
Imagine refrigerators with backup batteries. When the grid is stressed, those batteries could keep food cold without drawing power, creating resilience for the household and flexibility for the grid.
Markets can unlock this value. Today, no one pays you for your fridge’s flexibility. But if performance-based regulation and transactive energy systems take hold, millions of small, automated actions could add up to major resilience.
Performance Over Spending
Rate-of-return regulation rewards utilities for spending capital, not necessarily for delivering better outcomes. Lynne contrasted that with price-cap or performance-based systems:
Rate-of-return: utilities get a guaranteed return, no matter the outcome.
Price-cap: utilities must meet quality requirements under a certain cost
Performance-based regulation: rewards improvements in reliability, efficiency, or customer service, usually removes incentives for capital spending and removes disincentives for operational expenses
“If I were rewriting utility regulation, there would be a penalty structure on your ROE depending on your [reliability] scores.”- Lynne Kiesling
Aligning incentives with performance instead of capital spending could drive innovation from transmission and distribution utilities.
AI and Data Centers: The Demand Tsunami
Perhaps the most urgent shift is the rise of AI and hyperscale data centers. The International Energy Agency projects global data center demand will double by 2030.
In the U.S., McKinsey forecasts a 23% compound annual growth rate through 2030, adding 400 terawatt hours of new demand, the equivalent of adding another Texas in just a few years.
Utilities, designed for less dramatic acceleration, can’t match that pace. Data centers are already seeking alternatives: onsite solar + storage, natural gas peakers, geothermal pilots, and even small modular reactors.
Texas has a leg up. In many states, large customers are captive even for their generating resources to the monopoly utility. In ERCOT, they can contract directly with generators. That flexibility is why AI companies are flocking here and why Texas can continue to lead.
From Capacity to Capabilities
The old system was built on capacity: how many megawatts you could generate or consume at peak. But AI and automation are shifting the paradigm.
Most data centers won’t run at full tilt 24/7. Their true value — and the value of generation — lies in capabilities, how much load (or supply) they can flex, when, and where.
“Capability and flexibility, with that intersection with time and location, that is kind of everything going forward.” - Doug Lewin
Texas must evolve its mindset. It’s not just about building more capacity. It’s about enabling capabilities: flexibility, automation, and responsiveness that can balance reliability and cost in real time.
Final Thoughts
Texas built a world-class wholesale market by letting price signals communicate. The next step is to let that principle flow into risk markets, demand-side markets, and distribution markets.
Markets, as Lynne Kiesling reminded us, are a discovery process. If we reward performance, enable innovation, and let capabilities speak louder than capacity, Texas can not only handle the AI and data center surge — we can do so while increasing reliability and lowering costs for residential and small commercial customers.
That’s how we keep build out the grid and meet the challenges of the 21st century.
If this resonated, share it with a colleague who cares about Texas energy. And if you haven’t yet, subscribe for more conversations and insights on the future of the grid.
Sponsored by Aurora Energy Research
Aurora Energy Research provides leading analysis of global electricity markets. Explore their insights on the Energy Unplugged podcast and join their Energy Transition Forum in New York on October 21. Details at auroraer.com
Timestamps
00:00 – Welcome and Part 2 overview
01:30 – Why central planning doesn’t work; next frontier: demand side
04:30 – Markets as error correction, markets for risk including for fully regulated monopoly utilities
08:30 – Demand flexibility via automation vs. customer actions
12:00 – Transactive energy and user-friendly customer interfaces
14:00 – Price cap regulation and performance-based regulation
17:00 – Metrics for price cap and performance-based regulation
20:30 – Sponsor: Aurora Energy Transition Forum
22:30 – How AI data centers are reshaping demand
25:30 – Make-or-buy decisions for AI infrastructure companies
30:00 – Contracting for power in Texas
32:00 – Crusoe, flare gas to power
34:00 – Data center flexibility: reducing peak while overall energy use increases
35:45 – Why we should talk about capabilities not capacity
37:00 – Closing, where to find Lynne
38:00 – Credits and thanks
Resources
Guest
• Lynne Kiesling - LinkedIn, Knowledge Problem (Substack)
Company & Industry News
• Google, Kairos Power, TVA collaborate on advanced nuclear
• Reuters, Google to buy power from Kairos SMRs
• Google, Fervo geothermal project operational
• Crusoe secures 4.5 GW for AI data centers
Books & Articles Discussed
• Alfred Kahn – The Economics of Regulation (two volumes)
• IEA report on data center energy consumption
• McKinsey report estimating 23% CAGR in U.S. data center electricity demand
Related Energy Capital Podcasts
Turning Waste Into Power: Crusoe's Cully Cavness on Revolutionizing Energy Use at Data Centers
In April, ERCOT's Regional Planning Group published projections that ERCOT's electricity demand may double by 2030 compared to 2021. Initially anticipating a peak demand of about 90 gigawatts by the end of the decade, ERCOT now expects a staggering 152 gigawatts, up from approximately 75 gigawatts at the beginning of the decade. This would mark the larg…
Related Substack Posts by Doug
• Why Are Utility Bills Rising So Fast?
• Large Load Queue Has Tripled in Texas (Grid Roundup #76)
• Solar and Storage Help Reliability (Grid Roundup #68)
• Rapid Demand Growth Outpaced by New Supply in Texas (Grid Round #70)
• Energy Pragmatism, A Path to Abundance
🌐 Doug’s Platforms
• Substack
• LinkedIn
• YouTube
• X (Twitter)
Transcript
Doug Lewin (00:06.05)
Welcome to the Energy Capital Podcast. I’m your host, Doug Lewin. This is part two of my interview with Lynn Kiesling. She is a professor and historian of economics at Northwestern University and a non-resident fellow at the American Enterprise Institute. She is one of the smartest people on markets, particularly electric markets anywhere. If you haven’t listened to part one, go back and listen to that first, because we get into some of the history kind of bringing us up to the present. And then in this conversation talk.
Doug Lewin (00:32.642)
whole lot about the state of current electric markets, data centers and AI demand, risk allocation and why that really matters for the grid of the future. You know, obviously this is a discussion going on all around the United States and around the world. And a lot of folks think that it doesn’t apply to Texas because we do have competitive generation in retail, but there is the potential for markets and price discovery on the distribution grid. And that is, as I talked about in the previous podcast with Charles Hua,
Doug Lewin (01:02.082)
The fastest rising part of the cost of the grid is on the distribution side. And even in a state as competitive as Texas is, there is no competition on that side. And there is a really important set of questions which we get into here as to whether or not that should remain the case. So with nothing further, enjoy part two of my discussion with Lynn Kiesling. And as always, please, please like and share the Texas Energy and Power Newsletter.
Doug Lewin (01:30.7)
the Energy Capital podcast, please become a paid subscriber if you aren’t already get access to lots of paid episodes that otherwise aren’t available as well as great roundups, reading and podcast picks, the full archives and leave us a five star review wherever you listen with that. Enjoy the show. Thank you.
Doug Lewin (01:51.5)
And I just think that sometimes, particularly in the systems where they’re fully vertically integrated utilities, you are putting the regulator in a position where they have to be perfect and they can’t. That’s why central planning doesn’t work. So what I’m trying to figure out is like, what is the next evolution of competition? Because I think Texas, and again, you study this stuff all over the world, so correct me if I’m wrong, but I think Texas is pretty cutting edge on...
Doug Lewin (02:17.954)
wholesale and retail competition. seems to me like that next frontier is the demand side, is the distribution grid, is the next frontier something different? But if it is that, how does some amount of that price discovery and system of telecommunications that a market actually begin?
Lynne Kiesling (02:38.466)
Yeah, I agree with you on the demand side. We’ll come to that. The other thing that I think a lot of people who aren’t economists, and even a lot of economists actually don’t think about this, but a lot of people who don’t think about markets don’t necessarily realize that the other important role that markets play in our lives is error correction. Right? So if someone has made an investment and they’ve, know, we’ve
Lynne Kiesling (03:07.0)
built too many gas power plants. We’re not earning profits on that. that’s a signal to me that I need to take my money and put it somewhere else. That’s a really, really important and fundamental role for helping us to get the most benefit out of what we do. And that’s important. the other thing that, and I think this is really crystallized post-Winter Storm Uri and Winter Storm Elliot in PJM,
Lynne Kiesling (03:36.142)
with the kind of growing concern about winter peaking and summer peaking systems, where it just used to be summer peaking, is one of the things that these dynamics are revealing to us is that we’ve designed electricity markets, and I’d use the word design deliberately because we as humans have been engaging in exchange for 14,000 years. so exchange and markets
Lynne Kiesling (04:04.552)
are a very evolved, emergent social phenomenon. And so over the course of 14,000 years, we’ve developed certain practices, certain behaviors, certain things that revolve into rules that get codified into laws and get codified into market design. A lot of that gets reflected in the fact that these electricity markets are very designed, but it’s because they’re kind of born in captivity. They’re not emergent phenomena.
Lynne Kiesling (04:34.722)
But one of the rules that is very, very, and I would argue the thing that has been the most crucial deliverer of benefit to consumers and of producer surplus is just the move to security constrained economic dispatch. So if you say, all right, we’re going to run the power flow optimization and you generators, you submit your bids and we’re going to stack you from lowest to highest bids and we’re going to run the power flow.
Lynne Kiesling (05:03.482)
And that’s how we’re going to tell who gets dispatched. mean, that in and of itself, even if you didn’t do anything else, that would account for most of the benefit. The place that we haven’t, and Uri and Elliot have really revealed this to us, the place that we haven’t really done as good a job is we haven’t paid as good of attention to markets for risk and that we have very incomplete markets for risk allocation. So things like the ability for
Lynne Kiesling (05:31.426)
gas supplier to claim force majeure. And so force majeure is a big wedge that fractures risk allocation. And the fact that we as customers, bear outage risk, we bear rate risk on the regulated portions of the bill. And frankly, and here I’m channeling my colleague Josh Macy in saying this, the utility shareholders are insulated from a lot of those risks.
Lynne Kiesling (06:00.162)
because of the kind of utility ROE dynamic. And so figuring out and digging into how we can do a better job of assessing risk and allowing for transactions in risk to be better is, think, the next frontier on that side, even before we get to demand. But in SCAD, when you think about security-constrained economic dispatch in markets, that’s the supply side and there’s the demand side.
Doug Lewin (06:19.264)
That’s so interesting.
Lynne Kiesling (06:29.324)
And that demand side, as I know you and I have both written about for years, and I mean, you are fantastic on this post-URI especially about how much unexpressed flexibility there is in the diversity and heterogeneity of consumer behavior. And digital automation makes it easy to access that. We need to just do it.
Doug Lewin (06:51.534)
Now all the plates are spinning on my end. I’m like, okay, which of these threads do I want to pull on? are so many. Risk market, do want to go a little further on that because I actually think in Texas, URI, of course we didn’t have outages during Elliott. It wasn’t nearly as cold or as icy or the duration wasn’t as long. But during URI, yes, there were aspects of that, a large amount that was securitized and spread around all the consumers.
Lynne Kiesling (06:57.089)
risk demand.
Doug Lewin (07:21.538)
But on the generation side, a lot of generators took major, major hits. And some generators did very, very well. So I would argue that there’s some amount of that risk that is being communicated, again, back to the system of telecommunications. That price signal is there. Interestingly, though, back to the distribution utility, if you look at a hurricane barrel, center points performance.
Doug Lewin (07:49.452)
by their own admission. They’ve said this in their hearings. don’t want to just beat up on them here. They’ve said it, that we failed. It was a failure, right? A couple hundred thousand Houstonians were without power for a week. Did they suffer financially for that? I don’t think so. I listened to their investor calls, maybe not everyone, but most of them. And I think they had a billion dollars in net profit in the year after Hurricane Barrel. So again, I really think so much of it is on that.
Doug Lewin (08:19.768)
fully regulated utility. And again, I do not mean this to like beat up on them. They have a very important function and I think they’re going to be hyper relevant in our system for a long, long time to come. The question is, how do you create markets? that’s an interesting one. What a market for risk would look like for a regulated utility is something I frankly hadn’t really thought about before. So thanks for that. Now you’ve got my brain spinning on that one. I will say on the demand side,
Doug Lewin (08:49.674)
in those markets, you said earlier in this conversation that, and I think what you meant was more like in the early days, but I want to clarify, you said something along the lines of like demand has proven to actually not be super flexible. And I think for most of the history of electricity, that’s true. I think where we are now, just as you were saying, with all the automation, with chips and so many different things, with batteries and so many different things, I think this is going to be one of the big mega trends that we’ll see.
Doug Lewin (09:17.07)
fridges just being delivered with batteries in five to 10 years. This will be out on the internet and in five or 10 years, people can tell me I was an idiot or I was Nostradamus or whatever. like, you know, I think like electric induction stoves are going to have batteries built into it. Like we’re going to have batteries all throughout our homes because customers are going to demand it because they’re just not willing to throw out a fridge full of food every time the power goes out and batteries are cheap enough now that
Doug Lewin (09:44.224)
It doesn’t cost that much to integrate it. And then the question becomes, well, wait a minute, if every fridge has a battery in it, what does that mean for the grid? But there’s no market. My fridge is dispatchable, right?
Lynne Kiesling (09:59.65)
Give me a contract. My fridge is dispatchable.
Doug Lewin (10:02.178)
But seriously, where is the... So I think there’s a market for that based on the resilience value of that battery alone. I think consumers would pay for that. Again, I don’t make many predictions for your quote about Yogi Berra. It’s probably stupid, but I’ll make this one. I think we’re gonna see that because there’s a resilience value in it. If there’s a market that values that battery, then you create a virtuous cycle, right? I would think it virtuous for the consumers, virtuous...
Doug Lewin (10:31.512)
for grid reliability might not be virtuous for the utility if the regulation, if the incentive structure is not correct. So this is what I’m like trying to get at is like, how do we align those incentives? And this was in your markets as minds, your number two lesson for regulators was like reward performance, right? Not technology. So how do we get to this performance based system? What do we need to do there?
Lynne Kiesling (10:58.818)
We’ve
Lynne Kiesling (10:59.028)
been trying for a long time. So, so many things. In the electro-mechanical, you know, pre-90s power systems, right, the value proposition to the customer is I flip the switch and the light goes on, and it’s a mechanical thing. And so if you want to send me a price signal, say to turn up or down my thermostat or my air conditioner, I need to physically be there and do it. And that’s a pain in the neck.
Lynne Kiesling (11:28.174)
And I have a life. I’m not going to do that. so for the 20th century, right, the value proposition is, and we got habituated to this value proposition of you don’t have to think about your energy. You you flip the switch, you’ll pay a volumetric price and in a well-designed regulated rate, it’s a two-part tariff. Thank you, Ramsey Pricing. Thank you, Bond Break. In an ideal world, you pay your fixed
Doug Lewin (11:31.616)
Nobody’s gonna do that.
Lynne Kiesling (11:55.032)
costs in a fixed charge and your energy charge in a variable cost, but a lot of times that all gets swished together into one single volumetric rate and that causes distortions in and of itself. And so in that universe, of course, between kind of behavioral economics saying we’re habituated to this through regulation and through averaged fixed rates and plus the ubiquity of electricity, demand is going to be inelastic for sure. But with
Lynne Kiesling (12:24.034)
digital technologies for automation with remote control, with the ability to schedule, and of course the stuff I work on with Transactive Energy where you can have the devices in your premises themselves be programmed to be price responsive and automated. So they have automated price response. And so for the engineers, this is PID control theory, right? You can use the price, that emergent discovered price.
Lynne Kiesling (12:53.128)
as the device control signal, right? You set dead bands around the temperature setting. You set dead bands around where you do and don’t want your fridge temperature to go. And otherwise, you figure out a user-friendly interface for enabling devices that are gonna be meaningful, right? Like your HVAC, your water heater, your electric vehicle, obviously. So vehicle to grid.
Lynne Kiesling (13:20.632)
know, bi-directional V2G fits in this category as well. just enable a local energy market to exist and enable retailers to offer me products and services that enable me to have my devices participate. And I can figure out the ways that that works for me. And again, it’s going to be trial and error. Just like Hayek said, you know, I’m going to maybe set some trigger prices and
Lynne Kiesling (13:49.71)
find out that, I don’t want to have that because it makes me uncomfortable after two hours, or I need to set this price higher or this price lower. But you figure that out. And I think we would find that there is a lot more latent flexibility on the demand side that would not inconvenience or discomfort consumers. And I think we want to go through that discovery process that it would be beneficial.
Lynne Kiesling (14:17.598)
And as you say, having a more performance-oriented regulation would do that. In the UK, they use price cap regulation. And you’ve got rate of return regulation and price cap regulation. And performance-based regulation is kind of on the continuum between them, right? Because the price cap regulation is basically you go through the same kind of revenue requirement exercise as you do for rate of return regulation. But you just say, okay, here’s the revenue that we’re going to approve you to earn.
Lynne Kiesling (14:47.174)
and we’re going to ratchet it down by a little bit every year because you’re going to get more efficient because you’re good at what you do. And you figure out how to run your business. We’re not in the business of figuring out for you how to run your business. so I think, yeah, moving it more in that direction would get us the kind of stuff that you and I are talking about.
Doug Lewin (15:07.246)
In the UK system or in a system that’s similar to that, it doesn’t have to specifically be about the UK, but it could be. If they come in underneath that beyond just what the regulator expects for efficiency, do they get to keep a percentage? They keep it. So the incentive is to actually... They keep all of it or don’t they need to give some back to consumers? Don’t they share that?
Lynne Kiesling (15:29.898)
It has changed over time. may still have a savings sharing thing. And this is where, this is not a book recommendation for everyone because this is a bit in the weeds, but Alfred Kahn, right? Sort of the poster child for public utility regulation. The guy who famously, after being in the public service commission in New York, under Jimmy Carter in the Carter administration, went to Washington and was the chief of the civil aeronautics board.
Lynne Kiesling (15:58.562)
when we were still regulating airlines. And he kind of looked around and he’s like, why are we doing this? And even though he a kind of, know, center left Democrat, he’s like, we shouldn’t be doing this. And he navigated the economics and the politics in such a way that we closed an entire regulatory agency and turned out the light, limed the soil, done.
Doug Lewin (16:20.706)
I mean, honestly, it’s one of the reasons why I love this area, this issue area. It’s like Democrats, Republicans. At various times, liberals, conservatives, centrists, moderates, whatever will embrace competition and other times not. I think it’s really just one of those fascinating areas that defies the usual conventions. But what’s the con book? it profits and regulation or is it-
Lynne Kiesling (16:43.086)
He
Lynne Kiesling (16:43.286)
wrote a book called The Economics of Regulation, two volumes. he has a very important part where he defines what are the dimensions of regulation. And the one that is always important and I think doesn’t get as much attention as it deserves is defining the quality level of the service provision. And so under rate of return or price cap regulation, you want to define the quality level of the service.
Lynne Kiesling (17:12.642)
that customers are going to receive. And it’s a little difficult as systems get more complicated. But under price cap regulation, yeah, I mean, you have to meet a quality requirement. then if your costs are lower, then you keep some or all. I don’t remember the exact details of the difference.
Doug Lewin (17:30.126)
Give me an example of a quality requirement. What does that mean?
Lynne Kiesling (17:33.378)
The nerdiest one is just voltage and frequency, right? Your voltage and frequency have to be in these bands. I’ve seen some conversations around some of the failure to trim trees and the wildfire issues as being a failure to provide quality service.
Doug Lewin (17:52.61)
So I mean, outage minutes, right? Just like loss of load hours on the distribution system caused by extreme weather. Like those could be one of those. And I mean, that is fascinating to me. Right. Again, like we really, as far as I can tell, and I’m sure you’ll let me know and people in the comments let me know if I’m wrong. That does not really exist. Certainly in Texas, I can’t speak to other states, but like, you know, Encore is in a rate case right now.
Doug Lewin (18:20.918)
I don’t think there’s any discussion of, if your performance improves, if the outage minutes, I was in a news story, Spectrum News did, and it was about base power, how some homeowners in Plano had gotten a base power system because they were experiencing outages at their home like every few weeks. It was just like routine and they complained about it, but nothing got better. So they spent some money to get based. That’s a cool innovation in the competitive market where instead of spending
Doug Lewin (18:49.614)
10 grand on a battery, they could spend 700 or a thousand or whatever it is in some amount per month. And they sign up with a new retail electric provider. They’ve got a battery. That’s a cool example of the market delivering a product that it’d be hard to imagine a regulator like dreaming up, but that’s what markets are good at. But I just wonder, like, I don’t even know if Encore is there’s different areas in DFW that have different munis or whatever, but there are certainly areas where outages are higher. Is there any kind of performance metric that’s like, Hey, you want to
Doug Lewin (19:19.598)
10.5 % rate of return, well, you’re going to have to hit this benchmark and then you can get it. It’s wild to me that that conversation is nowhere.
Lynne Kiesling (19:28.492)
mean,
Lynne Kiesling (19:28.682)
in theory, reliability is the most of the service quality metric, the SATI and SAFE, right, duration and frequency of outages. If I were the benevolent social planner and got to rewrite the public utility regulation, one of the things would definitely be that there would be a penalty structure on your ROE depending on your SAIDI and your SAIFI scores. But there might be some reasons why I’ve never been nominated to the PUC.
Doug Lewin (19:52.632)
Yeah.
Lynne Kiesling (19:58.072)
Ha ha ha!
Doug Lewin (19:58.606)
Doug Lewin (20:00.422)
man, that would be fun to watch, Lynn. I would endorse such a nomination.
Lynne Kiesling (20:05.752)
But on your point about price cap, don’t have anything like this. There is a price cap regulated utility in the United States, in Iowa. Iowa is a vertically integrated state. I think it’s called MidAmerican.
Doug Lewin (20:17.164)
Okay, that’s a Buffett company. That’s interesting. Okay.
Lynne Kiesling (20:19.926)
Yes, it is. And they, I think, generally have good performance, not strikingly different from other utilities, but as a data point, there is one.
Doug Lewin (20:35.63)
Before we get back to the conversation, a quick word from our sponsor. You’ve probably heard on previous episodes of the Energy Capital Podcast, my discussions with Aurora Energy Research. They’re one of my favorite research outfits, doing tremendous work looking at electric markets around the world. They also have a great podcast called Energy Unplugged, and they will be hosting their second annual Aurora Energy Transition Forum in New York on October 21st. They’ll have senior leaders from utilities, finance, government, technology.
Doug Lewin (21:03.576)
They’ll be talking about grid reliability, AI’s impact, capital flow, strategic investments, all in one high level gathering. There’s an event you’re not gonna wanna miss. You can register now by visiting Aurora’s website. There you can also find all of the great research I was referencing earlier. It’s www.auroraer.com. That’s A-U-R-O-R-A-E-R.com and we’ll have a link in the show notes. Now back to my conversation with Lynn Kiesling.
Doug Lewin (21:32.462)
Okay, so a couple of things I want to talk about, data centers. You’ve written a lot about data centers. Really data centers is almost like a misnomer at this point. I’m trying to say more and more like AI infrastructure or hyperscalers, because there are data centers that are 10, 20, 30 megawatts and they’re still being built. But really we’re talking about something much, much bigger and really kind of transformational. You’ve written a lot about them. You wrote an excellent series, I believe last year.
Lynne Kiesling (21:59.448)
Last summary,
Doug Lewin (22:00.404)
In one of your pieces, you talked a lot about hyperscalers, make or buy decisions, and how these AI infrastructure companies, AI companies themselves, are quote, redesigning the market itself, creating new hybrid forms tailored to their economic and operational realities. Talk a little about the market impacts that data centers are having. And you’ve written about it. We’ll put links to your articles so people can go back and read those. I’m also just interested, like, as we’re recording and...
Doug Lewin (22:26.892)
September 2025, like what are you looking for? Like what are the kind of canaries in the coal mine or whatever that you’re like looking for for how this is going to play out? Are you looking for announcements from AI data centers? you looking for decisions from PUCs, bills from legislature? Like what do you think is gonna really sort of shape the future of AI power use in the US?
Lynne Kiesling (22:49.102)
I think it’s an all of the above and it’s going to vary a lot by state, but I have some statistics for the magnitude of this. So recently the international energy agency put out a report and they estimate in 2022 that actual data center energy consumption was 460 terawatt hours. And their forecast, because they do not obey Yogi Berra’s rule, their forecast is that in seven years in 2030,
Lynne Kiesling (23:18.958)
that that would more than double and the data center energy consumption globally would go up to 945 to over a thousand terawatt hours. So more than doubling in eight years. Looking in the U.S. specifically, McKinsey did some estimates and they estimate that in the seven years from 2023 to 2030, because we’re all benchmarking on 2030, the compound average growth rate of
Lynne Kiesling (23:47.15)
energy consumption from data centers is going to be 23%. Wow. So that’s a lot. mean, this is an industry where like 3 % a year is a big number. And over that seven years, that itself will amount to 400 terawatt hours of additional electricity consumption. So the demand growth is here and it’s not going to go away anytime
Doug Lewin (24:11.598)
This last year, by the way, was like 460 terawatt hours total. So when you talk about adding 400 terawatt hours, look, adding a Texas worth of power consumption in the next five, six years.
Lynne Kiesling (24:23.794)
And for various reasons, think this is an anomaly in the history of investor-owned utilities, both in terms of magnitude and pace. know, growth in construction happened more slowly. And it’s as you would expect for a network infrastructure industry that they scope projects cautiously. They design and engineer them carefully and build them carefully. And that takes time. And so both from
Lynne Kiesling (24:53.39)
prudence and just the business model, but also culturally, just the kind of culture of the utility. I don’t think utilities, the vertically integrated utilities that own generation, that they are going to be able to build enough generation fast enough to enable the tech companies to meet their speed to power requirements because a data center build is 18 months to two years.
Lynne Kiesling (25:20.468)
And so if a utility is going to build a power plant and it’s going to take eight, nine, 10 years versus two years, that mismatch is a challenge. so not surprisingly, data centers are going to look to alternatives. And there’s a big field within institutional and transaction cost economics literature that focuses on the make or buy decision. This is kind of an area of economics where we study vertical integration. And so if you’re a firm and you do a thing,
Lynne Kiesling (25:50.014)
And you basically decide, you know, do I want to buy the supplier of the input to my thing? And so that’s the make or buy decision. And this all goes back to Ronald Coase, along with Hayek, my other favorite economist. And Coase very famously created the field of transaction cost economics. And it’s his analysis that I was putting into play when we were talking about quarantining the monopoly, that all these technological changes are reducing transaction costs.
Lynne Kiesling (26:20.02)
and therefore making market exchange possible and profitable where it wasn’t previously. And I think that is the case in this question of data centers make or buy power. And this is naturally happening in ERCOT because ERCOT has quarantined the monopoly. And so if you want to buy generation, you don’t go to the regulator’s utility. You go to a power provider and you go to like an NRG or someone else. And I’m going to shout out Enchanted Rock.
Lynne Kiesling (26:48.888)
while I’m being a Texas booster. And so I think kind of the economic calculation for the tech companies, and when we say make or buy in economics, I don’t mean that they literally make, that they change their business into vertically integrating into becoming a power producer as well as a tech company, right? That’s a vertical integration that would require lots of assets on their balance sheet that are different from the other assets they operate. It would require
Lynne Kiesling (27:18.028)
Different HR.
Doug Lewin (27:19.084)
You know what though? Well, I wonder though, like, I mean, we might see that. I’m not saying it’ll happen, but like with like Magnificent Seven, quote unquote, I mean, you kind of already see it with like Tesla, right? They’re also an AI company that has their own storage, right? Like, I mean, Musk famously in Memphis was like putting little temporary gas plants up and stuff like that. So, I mean, you could start to see that though with some of those companies.
Doug Lewin (27:45.842)
If AI takes off, which is a huge F, that’s another market, right? Another back to that system of telecommunications. Like they may or may not make as much money as they think, but assuming they make as much money as they think they’re going to make, you could start to see some of these companies buying EPCs and you know, any number of different companies and just say, yeah, we’re just going to bring that function in house. You guys did such a great job for us over here in this place and in this place and we can see you can replicate it.
Doug Lewin (28:12.888)
Come on in house and build these for us all over the place and now our competitor can’t access your product either. mean, wilder things have happened. I’m not saying it will happen, but it could.
Lynne Kiesling (28:21.486)
You make a persuasive case. I’m more persuaded than I was before. But I think in the more immediate term, it’s going to be through long-term contracts. here’s my land. I’m building a data center. I want you to build your power plant on my land and serve my facility. And there are all sorts of interesting design aspects of that. And this is the other thing about the supply chain issues that
Lynne Kiesling (28:49.378)
know, gas turbines are on like a seven year supply chain at the moment. Substation transformers are also on a five to six year timeframe. And so if you are talking about building grid capabilities to be able to connect offsite generation to large loads, some of those pieces of infrastructure are also going to be hard to get in the near term. So I think onsite is going to have to happen.
Lynne Kiesling (29:17.866)
And I think the progression is going to be start with solar and here I am predicting when I said I wasn’t going to predict. But what looks logical to me is building solar and storage, unless you’re a huge 500 megawatt data center, that’ll probably serve you okay. So start with solar and storage because you can build those quickly. I think there’s a role for natural gas. And you know, I’m absolutely a portfolio diversification in the generation mix person.
Lynne Kiesling (29:47.682)
Yeah. And so Google’s a great example of this where, you know, they’re partnering with Fervo Energy to investigate geothermal. You know, if that geothermal Cape Point is in Nevada, you’re building data centers in Reno, you know, build a transmission line there. Sure. So there’s that, but then there’s also the investment they have with Kairos in piloting the small modular reactors.
Doug Lewin (30:10.05)
Yep.
Lynne Kiesling (30:13.806)
So I think there’s a technology pipeline here across time. And I think there’s a technology pipeline, but there’s also a contractual pipeline. And that I think we’re going to see this vertical integration or not question come up. And historically, this is a big challenge for the vertically integrated regulatory system. Because historically, if you want to plot a land and you have a data center and you want to have someone come and build generation that isn’t the utility that serves you,
Lynne Kiesling (30:42.296)
they’re going to come and knock on the door, say, excuse me, you can’t do that because we have an obligation to serve you and therefore you have an obligation to be served by us. And so traditionally, the way regulation has been implemented, even large load customers are captive customers. And so I think we’re in the middle of working out how that is going to evolve over time. It’s going to have to evolve.
Doug Lewin (31:09.644)
Well, it’s going to have to evolve. I think in the meantime, while it’s evolving, Texas has a huge leg up on other states, right? For sure. It’ll be like music to the ears of the governor and the economic development office and lots of local governments all around the state. Like, we have a leg up because you don’t have to go do the whole mother may I thing. You can contract and have your power. Like you said, like 18 months to build a data center, like you can get solar and storage in that period of time. You might be able to get
Doug Lewin (31:37.262)
smaller gas peakers, derivatives, maybe not an 18 to 24, but maybe it’s three or four years. then, so yeah, I mean, but the point is you can do that. You can go to the market and procure what you need, right? Through a contractor, a company like a, like an Enchanted Rock or a Lancia or Caruso, all these different companies that are doing this right now. Whereas in a lot of other States, you’re just left to like kind of the whims of the regulatory process.
Lynne Kiesling (32:05.838)
Since you mentioned Crusoe, I want to one last pitch for A, Texas, and B, one of the things that I love about their business model, right? Because it’s just straight up market-based. But part of what they’re doing, because they’re siting in the Permian a lot, right? They’re building data centers in the Permian. And so part of the value proposition and part of their mission is to use gas that would otherwise be flared because
Lynne Kiesling (32:34.19)
If you’re out there in the Permian and you’re pulling up oil and gas together and there’s not enough pipeline capacity for the gas, there’s not enough storage for the gas, you got the oil, you flare the gas. And both for economic and environmental reasons, flare gas is extremely costly. And so they’re like, hey, that gas you were going to flare, we’ll buy that from you. Fill our data center. That alignment of economic and environmental incentives through market investment.
Lynne Kiesling (33:04.302)
just makes my heart sing.
Doug Lewin (33:06.722)
Totally. And to take it even further, they’re also doing solar and wind and batteries. I put this in an article I wrote that Caruso was at that White House meeting on their AI action plan. they had the slide with the wind turbines up there. It really is that diversified energy portfolio. And then to bring it full circle and we can wind down on this, there’s the whole demand flexibility piece of this too.
Doug Lewin (33:31.564)
Right? So like, you know, there is real value in that. So if you’re an AI company that is procuring these different resources and you know, within your load, within your demand, you can reduce. is the Tyler Norris and the researchers at Duke, that team that did that great study of like, if you have flexibility within Texas, it was 1 % of your hours of flexible 15 gigawatts of new load coming on the system. So
Doug Lewin (33:59.086)
Like there literally is a value to that demand flexibility that we haven’t quite priced into the market yet. So this is something that I think I could absolutely be wrong. Again, predictions are tough, especially about the future, but I think data centers actually and AI infrastructure more accurately will start to bring more demand flexibility into the market just because it’s going to become necessary. When you were giving all those numbers of terawatt hours, I think one of the biggest mistakes
Doug Lewin (34:28.588)
we’ve made, including in Texas, and this continues, is we keep talking about AI and their loads in terms of megawatts and gigawatts instead of in terms of gigawatt hours and terawatt hours. Because you’re getting all that load does not mean that your peak demand is necessarily going up. Exactly. In peak demand in Texas, really, at least in the summertime, it’s a big issue in the wintertime, but in the summer it’s not that big an issue anymore because we have so much solar.
Doug Lewin (34:58.25)
it’s kind of more net peak. So we really want to look at like having more terawatt hours is good. That’s spreading costs around the system more if there’s a whole other set of questions. like, if we do it right, it’s spreading costs around the system, it’s economic growth and all that. It’s that gigawatt number at the constrained hours, which aren’t that many hours that you really want to be looking to control. So if you could do that, now you’ve opened up huge additional amounts of potential.
Doug Lewin (35:27.992)
for your state.
Lynne Kiesling (35:29.272)
This is another one of the things I think is going to evolve and it’s going to be tough because it’s a deeply culturally embedded piece of how we do this. I think the data center demand flexibility conversation illustrates something I’ve been thinking about for a long time, which is I think we need to move away from thinking about capacity to thinking about capabilities. the more we just try to shoehorn everything into capacity because capacity
Lynne Kiesling (35:57.802)
is a concept that was relevant to that electromechanical technology set. And now with digital, like I know from talking to folks in the industry, if you’re building a data center and you’re in the conversation with the utility, the utility is going to ask what’s your peak demand going to be, and you’re going to give them the biggest possible number. If you’re running that data center all out, because that’s what you need to do for your own operations planning.
Lynne Kiesling (36:27.342)
Most of the time you’re going to be running at like 60 to 80 % of that. so it’s that what are you capable of in terms of your behavior? And that’s going to matter so much more than capacity given the diversity of the especially digital technologies that we have now available for controls and automation.
Doug Lewin (36:48.718)
Yeah, capability and flexibility and with that intersection with time and location, right? Like that is kind of everything going forward. Lynn, this is amazing. I loved this conversation. We went a lot longer than I intended, but I’m glad we did. Is there anything I didn’t ask you that I should have or anything you just want to say in closing?
Lynne Kiesling (37:08.802)
No, I’ve loved this. I’ve enjoyed our conversations in person before, and now I’m glad we get to do it here as well. So thank you for inviting me.
Doug Lewin (37:16.994)
Lynn, thanks so much for doing this and thanks for all you do. really can’t encourage listeners enough to go check out Knowledge Problem and just wherever. Lynn Wright, you’re doing some writing as a non-resident fellow at American Enterprise Institute, so people can find your stuff there. Anywhere else, they should look to find your stuff, Lynn.
Lynne Kiesling (37:33.794)
The Knowledge Problems Substack is the best place to find me. Great. I’ve got other kind of academic papers and stuff, once I get a little less busy, I’m going to try to start kind of distilling them on the Substack.
Doug Lewin (37:49.25)
You drop these little things into all your little articles. This is coming in a future article.
Lynne Kiesling (37:54.636)
And at some point I’m going to have to deliver on them.
Doug Lewin (37:57.215)
Because otherwise I’m going to write you and I’m going to call you and be like, Lynn, I’m looking forward to that paper. Great. We’ll put a link to Knowledge Problem. also recommend it through my own sub stack. It really is great. I’ve learned so much from you over the years. Thanks for all you do and thanks for being on the pod.
Lynne Kiesling (38:11.598)
Thank you.
Doug Lewin (38:14.2)
Thanks for tuning in to the Energy Capital Podcast. If you got something out of this conversation, please share the podcast with a friend, family member or colleague and subscribe to the newsletter at douglouen.com. That’s where you’ll find all the stories where I break down the biggest things happening in Texas energy, national energy policy, markets, technology policy, it’s all there. You could also follow along at LinkedIn.
Doug Lewin (38:38.188)
You can find me there and at Twitter, Doug Lewin Energy, as well as YouTube, Doug Lewin Energy. Please follow me in all the places. Big thanks to Nathan Peevee, our producer, for making these episodes sound so crystal clear and good, and to Ari Lewin for writing the music. Until next time, please stay curious and stay engaged. Let’s keep building a better energy future. Thanks for listening.
