Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
AI Companies Are Building Their Own Power Plants. Here’s Why.
18 Plays9 hours ago
In this episode, Eric talks with energy strategist Peter Kelly-Detwiler about a question that is quickly moving from the engineering world into everyday life:
Will artificial intelligence fundamentally change how we think about electricity?
Peter argues that the AI race is not primarily a software race. It is increasingly a race for electrons. As companies invest trillions of dollars into AI infrastructure, electricity is becoming a strategic resource that shapes economics, geopolitics, national security, and technological innovation.
The conversation explores why AI data centers are unlike previous industrial loads, why a single delayed data center can represent billions of dollars in lost value, and why some companies are willing to pay electricity prices more than one hundred times higher than normal market rates. Peter explains concepts like “compute heat rate,” co-located power generation, battery storage, transmission constraints, and why data centers are beginning to reshape energy markets around the world.
Eric and Peter also discuss the surprising fragility of the modern power grid, how AI changes long-term planning for utilities, why batteries may become the Swiss Army knife of the electrical system, and why cybersecurity has become one of the industry’s greatest concerns. They examine the tension between innovation and regulation, the politics of transmission lines, and the possibility that electricity may become as strategically important in the twenty-first century as oil was in the twentieth.
Perhaps the biggest idea is one that most people rarely consider. We tend to think of AI as software running in the cloud. Peter suggests a different mental model: AI is the conversion of electricity, chips, and data into digital labor. If that is true, then the future economy may be constrained not by algorithms, but by our ability to generate, move, store, and protect electrons.
Episode Links
- Peter’s article on Compute Heat Rate: https://www.rtoinsider.com/129994-metric-related-data-centers-electricity-that-may-matter/
- Peter’s book The Energy Switch: https://www.amazon.com/Energy-Switch-Companies-Transforming-Electrical/dp/1633886662
- The Grid by Gretchen Bakke: https://www.amazon.com/dp/1632865688/
- Peter’s website: https://www.peterkellydetwiler.com/
- Connect with Peter on LinkedIn: https://www.linkedin.com/in/peterkellydetwiler/
- Peter’s first appearance on The Unfolding Thought Podcast: https://unfoldingthought.com/42-peter-kelly-detwiler-the-future-of-energy-and-sustainability/
For more episodes: https://unfoldingthought.com
Questions or guest ideas: eric@inboundandagile.com
Recommended
Transcript
00:00:02
Speaker
Peter Kelly Detweiler, thank you for joining me. You are the first guest to come back for a second time, which I really appreciate. But for anyone listening to this that is not familiar with you, would you please tell me a little bit about yourself?
00:00:20
Speaker
Sure. I'm 65 years old and spent more than half of that time now, 35 years in in electric power markets. Before that, I was like developing country economics, spent places like Mogadishu and Accragan and so on, but ultimately got into power in the early 90s and then got into competitive power after spending a few years in Chile on efficiency projects and You know, in parts of this country, you can buy and sell electricity. So individual customers and companies can buy power. So I got into that game. And then ultimately, I started getting involved in paying customers not to use power during periods of scarcity, during periods of peak demand. And so my team and I, we built the second largest virtual power plants.
00:01:03
Speaker
in the world. We have the equivalent of almost two nuclear plants of capacity from customers. We could just call them up and say, or text them or email, shut off within the next half hour. We need you to reduce consumption to keep the grid stable right now. So I got to mess around with all kinds of different things.
00:01:19
Speaker
And then 12 or 13 years ago, I went off on my own and just started doing all kinds of researching and and trying to figure out how the grid really works. You sent me two numbers that seemed fairly striking to me, but I'm hoping you can help provide some con some context. So you said that an AI data center might continue operating until power costs reach roughly $6,500 per megawatt hour and per megawatt hour and
00:01:50
Speaker
that delaying a 1000 megawatt data center could cost around $7 billion. dollars So would you mind explaining this to me? Sure, let's explore the latter first.
00:02:03
Speaker
All right. So right now, you have this huge race, Anthropic. like they hey so They may do an a IPO and be worth, what is it? I can't even remember the numbers, just a ridiculous number.
00:02:15
Speaker
And we're talking expenditures this year for data centers and AI of somewhere around $850 billion. dollars And over the next five years, I'll be talking $5 trillion. Now, why is that the case? Because essentially, you're taking NVIDIA or Google or other chips and combining that with raw data and electricity to create these models, which are neural networks, to develop and understand relationships between different facts, between words, between pixels, et cetera, and create models that humans can use, whether it's for robotics, financial trading, making movies, whatever it is, or even stupid TikTok videos. But but there's there's expected to be, because of the efficiencies around this, all kinds of economic value created. We don't really know how much, but that's the bet right now. So Google's raising all kinds of debt. I just saw another $60 billion dollars fundraise they're trying to get right now for more debt so they can make more investments.
00:03:13
Speaker
So the key right now is to be first or as close as you can possibly be. And there's um a huge impetus for that. And it's kind of like during the search engine race in the 90s, everybody wants to win this thing. and And as I like to joke during my keynotes, you know, that race was critically important.
00:03:33
Speaker
And there was really only one major winner, and that was Google. And there were a lot of losers. And if you don't believe me, you can go ask Jeep. Remember, that that was one of the losers. So the same thing is going to happen here, which is there are multiple players in the space and they're all vying to have the best models that then consumers and companies, trading firms, et cetera, will adopt.
00:03:55
Speaker
Google's already really well ensconced because it's got Notebook LM and Gemini and people already use the search engine. So they're they're going to win. But then does Meta win? What does AWS do? Where's Anthropic in the game? Is OpenAI trailing, which it looks like they are now?
00:04:10
Speaker
At some point, there will be consolidation. But the winners are going to be the ones who get there first and develop the best models that then get embedded in everybody's networks.
00:04:21
Speaker
When I was at Constellation building the virtual power plant, Eric, we gave away our platform, our energy management platform, because my belief was if we get that embedded into our customers and they start to tie it into their accounting systems and so on, it could not even be the best system, but it could be so painful to rip it out because of the inertia. Once you're there, you win, unless you really, really stink.
00:04:47
Speaker
And so that's the game right now. And so these companies, therefore, want to build these models as fast as they can. And the main constraint right now isn't fiber and it's not water and it's not land. It's it's electricity. It's the next electron.
00:05:03
Speaker
And so the cost of not being able to access those electrons, of delaying that 1,000 megawatt power plant, which is about the size of a single unit of a nuclear plant, of of you know delaying that that AI data center and not getting access to the power, is estimated to be roughly around $7 billion dollars a year.
00:05:22
Speaker
So they will do anything. So that's 19 point something, I think $19.2 million dollars per day. So they will do anything they can to access electricity as fast as they possibly can. And so you see these crazy numbers where Dominion in Virginia Their peak last year of demand was 23,400 megawatts, and they have 47,000 megawatts of data centers chasing interconnections with their utility. In Texas, they peaked out at 85,500 megawatts, and their large load forecast, plus the load they have, gets them over 300,000 and I think maybe even closer to 400,000 megawatts. Now, will all this be built?
00:06:01
Speaker
Absolutely not. There are multiple players chasing multiple sites knowing they only have the capital to fund one, but they might have three or four applications on the board. And then one gets funded and built, they pull the other ones off. So there's a lot of so-called phantom load.
00:06:16
Speaker
So nobody really knows what to forecast for or what the implications are. But here's the second part of that number. These data centers get built. And then there's the question of what are their price elasticities. You and I walk into a store or a bar and a beer is $10 for a 16-ounce beer. Well, we might elect to drink that because we're kind of used to that. But if someone says, oh, that beer is $20, you say no.
00:06:42
Speaker
Okay, my elastic price function just kicked in. I'm not willing to pay $20 16-ounce beer. Same thing with these data centers. They all have revenue streams.
00:06:53
Speaker
And with open with available information, a professional that I've been dealing with who shared this idea with me, and then I wrote about it for, i write for an industry publication called RTO Insider, which is the grid operator sort of insider. So it's all pros and regulators that read this.
00:07:10
Speaker
And I write this article called Around the Corner. And so I wrote about this guy's concept called the compute heat rate, which is the price at which a data center would liked not to consume power because they don't make enough revenue for the same megawatt hours.
00:07:25
Speaker
And so that price, that weighted average price of $6,500 per megawatt hour, that's roughly 127 times the weighted average market clearing price of power last year.
00:07:36
Speaker
So it's $6.50 KWH in the terms we You and i pay to cents okay So the Frontier inference Frontier, like the leading cutting edge models on Frontier. Frontier is so the LLM, you're building the language model.
00:07:55
Speaker
Then what do you do with it? Autonomous driving takes that data and recognizes that's a cat, that's a dog, that's a woman crossing the street. Don't hit the woman. That's inference or trading using computers or robots using using those models to perceive what's around them and and move carts or whatever. That's inference.
00:08:15
Speaker
That training inference models, this guy, Hans Royal, he estimates the avoided cost there at around $50,000 a megawatt hour. So now when you've got sufficient supply and there's an overhang of supply over demand, doesn't matter what that cost is. If there's a full keg of beer,
00:08:33
Speaker
Then, you know, whether or not I'm willing to pay $20 or 10, it doesn't impact the market price of beer in the bar. But if you and I and all our friends went into the bar and said, we'll pay $30 for a glass of beer, the barkeep would say, well, you guys get it all. Everybody else too bad for you because these guys are going to pay more.
00:08:52
Speaker
Right. If there is more demand than supply, it can shift the fundamental fundamental clearing price of that electricity. And so this guy postulates that around 15 percent of total demand in the system, you actually create a scarcity function. Now you can see the prices pop up because there's more demand than supply and they end up setting the hourly prices for power and beggar the rest of us, right? Driving other industries out of the space. It's kind of like in Manhattan, if everybody's willing to pay billionaires high prices for apartments, what happens to the schlep, you know, who doesn't have the money?
00:09:28
Speaker
That's what's going on potentially in power markets real soon. And we saw the first potential instance Two weeks ago, NV Energy in Nevada told Liberty Utilities that serves Tahoe, we're not going to renew the power agreement that you serve your, I think it's roughly 50,000 customers with.
00:09:49
Speaker
At the same time, they have these massive demand forecasts for data centers. So you kind of go, well, they probably would have renewed that contract, except they have all this new demand willing to pay a higher price.
00:10:02
Speaker
What was the scale, again, roughly that you talked about just a minute ago between the, I guess this, maybe the $6,500 per megawatt hour number and what you or i would typically pay for? your typical average clearing price in the market is 127 than that.
00:10:21
Speaker
Wow. Yeah. Okay. And it's it's things like this that that drive these data centers that are being built also having their own associated power generation capabilities. Yeah, it's called co-located or self-supply.
00:10:39
Speaker
And if you look at the most recent graphs, like last year, there were maybe 5,000 megawatts of announced gas fire generator projects that were going to be behind the meter on premise for data centers.
00:10:51
Speaker
Now that number is catapulted up to 100,000 megawatts. Now, again, will it all get built? Who knows? Because among other things, the turbines are in short supply. If you look at GE Vernova and Siemens in Mitsubishi, they're the holy trinity of gas-fired turbine manufacturers. Like, GE Vernova will tell you right now, you don't get one unless you put down a $30 million dollars deposit. And even then, you're not going to see it until 2029 2030.
00:11:16
Speaker
So, you know, there's an articulated demand there, but the reality of the facts on the ground because of supply chain and also regulatory issues and so on will shape the landscape based on another reality.
00:11:29
Speaker
The example of, I think it was Nevada not renewing a contract to, it sounds like, provide power for residential use, yeah then do you think that that is likely to happen in many other places?
00:11:48
Speaker
That one, in some ways you could argue, well, Liberty should have, you know, they had that contract renewed a couple of times, so maybe they should have done something, but they've been renewing it again and again and again for years. So that looks like a fairly clear case where, well, this looks better than that. See you later.
00:12:04
Speaker
I think what we'll see over time as we um get closer to scarcity on the grid, it's going to happen more often. The first place it's already happening right now is you see data centers wanting to cut deals with power plants that already serve existing markets.
00:12:24
Speaker
Okay. So right now you have a supply demand function. And when supply and demand cross, that's where your equilibrium price is. Well, if you all of a sudden yank a whole bunch of supply out of that equation to feed data centers directly, you just shifted the demand and supply curve, right? You actually moved the supply curve up and to the left or right. I can't remember. I'm talking to you backwards on the screen. But what you do is you fundamentally set a new equilibrium price because you just yanked a whole bunch of supply out of the equation.
00:12:54
Speaker
Conversely, if you add a whole bunch of demand and supply stays limited, you can do the same thing. You're going to set a higher equilibrium price. So one more data point on that. In PGM, Pennsylvania, Maryland, the mid-Atlantic power pool, every year, supposed to be, they have an auction for capacity for enough steel in the ground generating capability or demand response, people who can shed load, to meet expected demand.
00:13:20
Speaker
okay So there's a price for the ability to supply power during the critical peak period. And for the last three years before the dynamic changed, that clearing price was somewhere between $30 and $50 per megawatt day for this capacity, ah but ah ability to serve load.
00:13:38
Speaker
Then PGM comes up with this. forecasts for all this new data center load that was given to them by all the different utilities that these data centers were approaching. And they say, here's what our new demand forecast looks like.
00:13:51
Speaker
So then they run the next auction and the prices go to $269.90 per megawatt day. So they're up like seven, eight, nine fold what they were. Wow. And everybody screams, ah, you're killing us. And Josh Shapiro, the governor of Pennsylvania, the technical term is, he freaks out.
00:14:10
Speaker
And he goes to the Pennsylvania, Jersey, Maryland, the PGM power pool, and says, I want to negotiate some kind of a cap and a floor to make it fair for generators. And so we'll cap we'll put the floor at $175, which is historically higher than they'd ever seen in the past, and the ceiling at $325.
00:14:29
Speaker
So for the next two years. So they run the next auction. Bam, they hit the ceiling. And then they said, what if the ceiling weren't there? Let's run a simulation. Instead of $325, what would the price be? I said, without the cap, it would have been $388.
00:14:42
Speaker
So then they run the next auction. Prices hit the cap again. They go, okay, now let's run that again. Let's simulate it again. Now prices are at $533. Okay, so then the caps expire, but then he goes back and says, i want to negotiate those caps for two more years.
00:14:59
Speaker
So the next auction happens this month, and we'll see what what transpires there. But the there's ah there's an economist group that's job is to look at the market and say, okay, what's going on? Is there market manipulation between suppliers? What's fair? And they report directly to the feds, to the Federal Energy Regulatory Commission.
00:15:22
Speaker
And they've looked at the last three auctions and said roughly half of the total revenues paid for capacity $23 billion dollars is because of new data loads that are forecasted, not even built yet, mostly forecasted, which means that all these other customers are paying these new high prices because of these loads that don't even exist yet.
00:15:47
Speaker
So this is the kind of political pressure that's now mounting where, you know, what's why the 13 states sat down with Trump and said data centers got to pay their own way. they You know, they got to get their own supply and they can't be messing with these market dynamics that are raising the prices for everybody else.
00:16:03
Speaker
But even if they do that, Eric, if they come up with the tariffs and the approaches that are meant to inoculate everybody, They still don't mitigate the impact that an 800-pound gorilla just jumped into the bathtub that or the swimming pool that you and I are in.
00:16:17
Speaker
It raises the water for everybody because it'll push up gas prices because we're going to be using more natural gas. It pushes up the prices of turbines. It pushes up the prices of transformers and switchgear and copper and everything else because we're now in this superheated inflationary dynamic, at least for the next you know three to five years where we have this $5 trillion dollar spend taking place.
00:16:39
Speaker
How much ability do we have today with deal with and you know, peak, peak need, uh, you know, when everybody's AC unit goes on in the summer and now we add an AI data center on top of it, or when, you know, in a year when battery storage capacity improves, how will that impact our need for power generation? Like what, do what do things look like today?
00:17:07
Speaker
and as, technology changes, what kind of pressures will that put on the need for new materials or its a grid petroleum?
00:17:18
Speaker
So we're fundamentally dealing with two different concepts here. One is the peak capacity, which is like when your AC load is cranking and you got to serve all that load. And then you have to have a 15% reserve margin, give or take on top of that, in case you lose one of your largest transmission lines or generating asset, right? So you always hold that reserve margin.
00:17:37
Speaker
The estimates are that the cost of serving the top 1% of load, so there's 8,760 hours in a year, so like roughly 90 hours of peak demand, is about 8% of the total capital costs.
00:17:50
Speaker
Okay, so your listener can't see what I'm going to do, but I'm going to explain it. All right, I'm holding my two arms up and my fingers are meeting at a V. The V is where the peak demand is, okay? What I don't want to do is push the V up.
00:18:05
Speaker
But if I had data centers with batteries, for example, or they could operate flexibly so they didn't use more power during peak demand because either they charged the batteries during the off-peak and released that energy into the data center during the on-peak, they shifted their compute function to another location that doesn't have as much demand, Or they simply flexibly operated so they didn't use as much power then because they're going to crunch the numbers later. Then the peak, the apex of where my fingers meet would stay the same.
00:18:35
Speaker
But they would still consume more electricity. So my elbows go up, right? And so essentially, I'm flattening the curve across my two arms. What this means is the space under my arms is bigger.
00:18:49
Speaker
I'm flowing more megawatt hours across the same infrastructure that I had before. Or I'm not building that much infrastructure, but I'm utilizing it more heavily, which means the delivery cost per unit would go down. And in fact, there was a post on LinkedIn this morning from Google, from one of their representatives, and she said in Indiana, Detroit, and one other area, the utilities are saying we're going to actually be able to reduce our costs, at least for the next year, our rates, or the next couple of years, because those first data centers are going to allow us to flow more megawatt hours across the same infrastructure.
00:19:25
Speaker
That's great for now, but we're only in the first inning of the game. Once we start to bring even more data centers in, you're going to have to build new stuff. Whether you use it at 50% utilization factor or 75%, you're going to have to build more stuff.
00:19:42
Speaker
And then the question is, how do those costs get allocated? And that's where the regulators come in with their tariffs and say things like, if you want to build something here, you enter into a take or pay agreement, or you want to build your data center, you want to consume power. In order to not impact other rate payers negatively, which they would be if we had to build that stuff and everybody paid for it, we say, okay, data center, large load.
00:20:09
Speaker
You come in you commit to, say, four-year ramp period because you can't build the data. It's not like it springs off your forehead like that, you know, Athena off Zeus's brow. It takes a little while.
00:20:20
Speaker
But then after that, it runs at a forecasted flat level. So basically, in Ohio, they were the first ones out of the box with this tariff about a year and a half ago, and the data centers bought it like hell. But now they've accepted it pretty much everywhere. It was a four-year ramp and then eight years of committed demand. And they said, you got to pay for 85% of that forecasted demand over the 12 years, irrespective whether you use it or not.
00:20:47
Speaker
And you've got to put like a letter of credit down, and there's exit fees, and so on. Now, some companies are saying, 12 years isn't long enough. We want 20 years. Because when you're building new infrastructure, like transformers and transmission lines and generators, they're usually 30 to 40 lifespan assets.
00:21:05
Speaker
So if you made the investments and then the load went the demand went away, you'd have these so-called stranded assets that everybody else would have to pay for. So they're thinking longer-term contracts help to insulate the rate payer against the otherwise incurred cost for infrastructure.
00:21:22
Speaker
when I think of AI data centers and we think of some of these hyperscalers, you know, the the i feel like the stock market is sort of telling the story of this is all about Google.
00:21:36
Speaker
This is all about, you know, Microsoft, not so much, but some of these big, you know, magnificent seven type businesses. But, It also seems to me like no matter how much as a hyperscaler I'm thinking about potential power generation, generating our own power, that this must be driving stock prices, for example. It must be driving crazy revenue for existing power companies as well.
00:22:07
Speaker
Yeah. So, for example, let's take something that happened the other day, which I sold my stock too early. Fluence, which is a JV between Siemens and AES Power Corporation, does storage.
00:22:18
Speaker
They announced a tie-in with Siemens and with NVIDIA the other day, and the stock popped 40% overnight. overnight Bloom Energy, which is a fuel cell company, they were trading in the $15 range. And then they announced that they were selling up to 1,000 megawatts of fuel cell boxes to AEP. Then they got an investment from Brookfield. Then they made some big announcements with some data center companies. And their stock was at $15. Now it's at $300 and change. just fell back to $275. So... There's a lot of examples of everybody, and certainly the utilities, stocks, NextEra, the largest you know generator in the country, they've all been swept up in this rising tide, lifting all the boats affected with the power infrastructure. What's interesting too, though, Eric, is if you think about the bill of goods, the power piece of what the data center companies are paying, it's only 5% of the total costs.
00:23:12
Speaker
It's the chips that cost you. and When you're spending as much for a car as for each one of these Vera Rubin chips, $30,000, $40,000, and they depreciate over five years, sometimes less because the new chips come along and make the last ones look like idiots, you're You know, that's where that's where the money and the risk is. Energy is necessary. It's critical. Without it, the system doesn't function. But from an economic perspective, it's an afterthought, which is why you see these companies saying, well, we'll put the stuff on, we'll put the generation on site.
00:23:45
Speaker
Now, if you think about that, Remember I mentioned that 15% reserve margin figure before. That's because you have a large system. If one unit fails, you still have a lot of other assets and you have diversity.
00:23:58
Speaker
What if you're a data center and you're not connected to the grid? Or maybe you want to connect to the grid to the future. But today, you put generation behind the meter on site.
00:24:09
Speaker
Well, what if that generation fails? Now you're down for maintenance or whatever. So there's a data center in the Midwest that wants, they actually, it's a metadata center, 200 megawatts of supply.
00:24:22
Speaker
They have 30 generators on site. And I think their figure is the number is 320 megawatts of nameplate capacity. So they're carrying like a 50% reserve margin because they know that units are going to go down for maintenance. It's a whole bunch of reciprocating engines and some simple cycle generating turbines.
00:24:42
Speaker
And, you know, so they've got 30 of them and they're never going to connect to the grid. They have a special terra for it, but they pay a huge amount for that extra reserve margin that the grid would never carry.
00:24:54
Speaker
That's the economic tax if you want to be feeding yourself forever. How much does the source of the power, whether it's, you know, a ah coal power plant versus solar power or whatever, how much does that differ by location? You know, state laws, the the laws by country, something of that nature.
00:25:17
Speaker
If you were to put it behind the meter, the main the first thing you face is whether or not there are EPA regs. So Elon Musk put XAI, he put a data center with turbines that were violating the EPA rules, and he's just like, I'm not going to pay attention to you.
00:25:32
Speaker
And they didn't do anything about it. You know, when it's a low-income area affecting people's air quality, and and rich people sometimes can just ignore the realities that the rest of us face. But for the most part, though, they do have to play a ball. And the more responsible actors, and I wouldn't put XAI in that bucket, the Googles and the Microsofts and the Metis of the world, especially Google and Microsoft, AWS also, they they have been... um pretty good about maintaining their carbon targets and making commitments to um to be good actors in the space. And while some of them have now said, we're not going to meet our short-term targets because of this massive growth, but we're not chucking the targets. You know there are those players that have always been sort of the visibly good actors and they're the ones that drive all the renewable energy contracts. And then there's other buffalo in the herd that are a little bit mean-gier and uglier.
00:26:30
Speaker
you know And that's always been that way. But yeah, there are certain state regulations. And there's a number of states. I think there are 14 now that I was looking at And not just blue states where they're talking about moratoria on new data center construction.
00:26:45
Speaker
Have you seen, whether it's in the United States or you're hearing stories from around the world, where there are places where it's just open season on, you know, any power that you can produce, who cares what the downstream negative impacts are?
00:27:01
Speaker
Nobody's saying no. But for example, California has still a goal of our an articulated and legal goal of 100% carbon-free grid by 2045. And that means ultimately but first you know integration of 40% or 50% or 60% renewables, that's feasible And the first thing you do is you make your gas turbines work harder and more flexible. So you integrate the wind and the solar, which is intermittent. But what do you get to 100%? what do you do? You can't have the gas turbine anymore, right? Unless you have huge hybrid reservoirs or ginormous batteries, which don't exist yet.
00:27:41
Speaker
The lithium ion batteries we now have are like good for four to six to maybe eight hours. But what if you have five days of an atmospheric river in California where it's raining cats and dogs and your solar capacity is down by two thirds during the daytime because the cloud cover?
00:27:57
Speaker
Now the raw energy is not there for a long period of time. Now, unless you have transmission or you have other assets with lots of raw energy, or you could have one of these reverse rust batteries that form energy is now producing, that'll give you 100 hours of duration. In fact, Google just bought...
00:28:16
Speaker
300 megawatts and 30,000 megawatt hours of batteries from Form Energy combined with 1,400 megawatts of wind and 200 megawatts of solar, if my memory serves, in Xcel Energy, Northern States Power, Minnesota.
00:28:31
Speaker
And that number, the 30,000 megawatt hours, that one contract to put that in perspective, that's 60% of all the battery storage installed last year in the United States.
00:28:43
Speaker
A single contract. you know So you can do, that would enable you to create a carbon-free grid, that kind of long-duration storage. But for the most part... Solar you know has a production curve in the middle of the day and not other times. And so you really have to look at this as a symphony of different players in the orchestra because you know unless you have a fossil fuel that you can feed it to the plant and dispatch all the time. Once you get into renewables land, the mixing and matching and the smoothing over sort of with storage as peanut butter gets more challenging.
00:29:16
Speaker
Peter, do we think about energy differently today from how we did even just five or 10 years ago? And when I say we, I'm thinking both the public, but also people in your industry.
00:29:31
Speaker
Yeah, we do. I mean, Mikey Sherrill, who came to power as the governor of your your New Jersey, she ran on lower power bills. You know, Graham Plattner, who's running for Senate in Maine, he talks about lower power bills. Like, energy is now front and center in a lot of ways. And I said I wasn't going to talk about it, but i actually I'm working on a book proposal that I should have the contract signed within next few days. It's about the electrostates, right?
00:29:59
Speaker
which is And the thesis is, yeah, we have molecular fuels like coal and oil and natural gas. And they've been the thing we've built our societies on in the future. But now when you fuse data and chips and, pow, electrons together, there's a much higher value creation opportunity because what you're doing is you're creating a digital workforce.
00:30:21
Speaker
Like everything you have AI do for you, yeah that's new labor that didn't exist before. Right. And so China, in some ways, they're depopulating because of the one child program. They need to find ways to expand the workforce. And artificial intelligence is a really good way of doing that.
00:30:40
Speaker
among other things. Because what do you have? You have capital and you have labor, right? And you have raw materials. Those are the three things that make up an economy. You can't expand your economy if labor is finite. But with digital labor, you can do that.
00:30:57
Speaker
And so that creates the possibility for electricity to become way more important than it was before. And China is moving really, really fast in terms of it already installs half the world's solar, half the world's wind, makes 80% of the world's batteries and installs half of the batteries.
00:31:15
Speaker
They're putting in all kinds of nuclear power plants. And 10 years ago, it was a 700 megawatt coal plant every week, some of which weren't even registered with the central government. So I think we're at this phase right now where, yes, we all think about electricity differently. Certainly everyone with an EV does already.
00:31:34
Speaker
And then you put in heat pumps. And now also, by the way, you might have seen these laws that are now getting passed allowing people to buy balcony solar. little solar things they can hang, they can put on their porch or the balcony and plug straight into the house, into the 110 outlet to offset some of their power bill. Germany's got millions of them all already. So I think our relationship with the electron is fundamentally changing right now at an individual level, residential consumer level, and then more profoundly at the nation state geopolitical level because of this AI dynamic and all the
00:32:10
Speaker
both risk and opportunity that that fusion of chips and power and data creates for us. I imagine that as a regulator in one state or another, but also as a a power generation company, trying to think about not just you know, what the the growth rates have been in power demand, the supply of um petroleum or whatever it is that we're burning, the cost of solar cells, you know, to be able to install, not just thinking of the growth in that, but also the rate of growth or decline yeah in population and how that's going to influence things. It seems like
00:33:00
Speaker
It's just so incredibly complex to try to plan for something like a 15 or 20 year contract then. It's impossible. I mean, think about the fact that chat GPT 3.5 popped out of the Jack in the Box in November of 2022.
00:33:16
Speaker
So we wouldn't even have had this conversation two years ago because it wasn't meaningful yet. You know, certainly three years ago. So you're talking 20 to 40 year forecasts.
00:33:27
Speaker
It's even stupid to conjecture them because the technology, especially now with AI technology is going to accelerate so fast because it helps us with material science and just innovation in general.
00:33:40
Speaker
We're on an accelerating path that we'll start to soon see. I mean, right now, there's only if a handful of large language models out there. There's dozens, but it's small. Relative to the multi-trillion dollars of investments over the next five years, that can't help but change the innovation landscape itself, right? And then, yes, to your point, that the there was a great...
00:34:02
Speaker
slide that American Electric Power had probably 18 months ago. And I just, I'm an inveterate report reader and slide clipper, and I remember visuals. And this one, and I'm doing a four-hour training in a week at the, well, a couple weeks in the American Public Power Association. This slide will be in it on on data centers.
00:34:20
Speaker
And what they said was, traditionally, we forecast with our macroeconomic tools. we look at pop We look at demographics. We look at GDP. We look at these macro indicators that typically tend to move more or less slowly unless you have a shock like the 2008-2009 calamity or something like COVID.
00:34:41
Speaker
right These sort of black swan things. But for the most part, it's visible, it's statistically supportable, and it generally moves, not quite glacially, but at least visibly so you can forecast it.
00:34:53
Speaker
Then you have a data center, which all of a sudden, you know, in northeastern Louisiana, in two years, you're building something that's the same as Minneapolis on, you know a mile of farmland.
00:35:05
Speaker
Like, it's a yes or a no. They're binary. They're short in duration before they sprung up like mushrooms. And they're huge. So that makes it almost impossible to forecast because,
00:35:19
Speaker
You just don't know. And then you get the numbers like the stupid land numbers like Texas, where they're talking about, you know, tripling the grid in five years. Well, it's physically impossible to do. So like, why even throw that number out? Big is big and bigger than big doesn't matter.
00:35:35
Speaker
Is there something that makes an AI data center fundamentally different from other large consumers of power up to this point?
00:35:47
Speaker
There's a couple of things, except for, say, aluminum smelters. So aluminum smelters typically like to run twenty four or seven And you can't shut the pot lines down because if you do, they freeze up, right? so And they are what the Hydro-Quebec used to call them when I worked on Hydro-Quebec stuff, an An archivore, like a carnivore, but it's an energy eater, right? i love the term.
00:36:08
Speaker
But most loads in this country, most demand, industrial demand, typically runs on shifts. And it's usually one shift or two shifts. So it's a nine to five type thing. Sometimes you want to utilize your equipment. You might run two shifts or sometimes three, but that's usually on a wartime type thing.
00:36:26
Speaker
Large language models, they run at a really high capacity factor, north of 80%. So they're running most of the time, except when they're shutting down and you know adding more load. Now, inference, that follows human behavior much more because like trading models are going to be trading when the markets are open.
00:36:45
Speaker
Autonomous driving, that's going to be when people are moving around. You're not going to have as much autonomous driving at 2 in the morning as you do in 2 in the afternoon. So depending upon what the use of the AI is for, the models, either the model creation, you know, the LLMs, the models in use and inference are going to have different characteristics. But there's one other fundamental piece that's kind of fascinating here. These LLMs, what happens is the chips process the data, and then for macro seconds, they pause. And the demand drops down like it plummets. And then they pop back up again and crank through the next bit data. And so what does that do?
00:37:23
Speaker
Well, our grid is kind of like a human heartbeat. It needs to operate at 60 hertz, which is 60 beautiful sine wave oscillations per second. And that's called harmonics or power quality.
00:37:36
Speaker
And because if these were small loads doing the eh, eh, eh, it wouldn't make that much of a difference. But when they're big loads, it's like a cardiac event. And what happens is it actually impacts power quality of other players in the space.
00:37:51
Speaker
And so now the data centers are being asked to put batteries in because the batteries can absorb that and actually feed, take the power from the grid, store it, and then feed that interruptible, like up and down, up and down pattern without screwing with the grid.
00:38:07
Speaker
And then one other piece about the big data centers that's kind of important to understand is, so like during Winter Storm Yuri, when all of a sudden we lost all those generators. Well, what happens when you have a delta or a big difference between demand and supply? There was tons of demand and supply kept falling off.
00:38:27
Speaker
Your frequency falls off from the nice desirable 60 sine waves per second, 60 hertz. It dropped down to 59.4. Another four, four and a half minutes and the Texas grid would have crashed because the machines connected to the grid, those gas turbines, they're actually fairly delicate.
00:38:45
Speaker
And they need to be connected. They're all connected at the 60 hertz. They all function together as one machine. even though they're multiple units, but they're all doing the same sine wave that are harmonized. So if the sine wave gets messed up, it can break them. And so they would they would have started disconnecting from the grid. Well, equipment in data centers, those racks of servers, they're super sensitive.
00:39:11
Speaker
And so there's basically these disconnect functions between the grid and the data centers. And so in July of 2024 in Virginia, a lightning strike, and lightning's got massive voltage, way more than the grid. It could, you know, lightning hit my little invisible dog wire. And the only way I knew that I hadn't grounded it was when I came home, the entire lawn had this, the so come of the perimeter was blown out of the out of the lawn, there was burnt wire every place, and the little connector was blown off the wall with a huge black mark. I'm lucky it didn't set my house on fire, right? So so anyway, this this lightning struck the fault arrestor, and the fault arrestor that was meant to absorb that and protect the grid, it failed.
00:39:56
Speaker
And so there was a voltage surge in the system. And 1,500 megawatts, so the equivalent of a nuke and a half, sensed the disturbance in the force immediately. And to protect themselves, they disconnected from the grid.
00:40:10
Speaker
Well, the generators didn't change their behavior. So suddenly there's an oversupply relative to demand. And we went into an overfrequency situation and voltage was messed up. And the federal regulators said they almost crashed the grid.
00:40:23
Speaker
So we have to think about them not only in terms of how much energy they consume in this capacity and energy consumption issue. We also have to think about how they behave on the grid in terms of the physics every single day. So now the North American Electrical Liability Council issued a very rare...
00:40:42
Speaker
level three alert within the last month saying, we got to get this situation under control because it's not the only time we've had this kind of voltage issue where they disconnected. And so what they're demanding that they do, and Texas is already saying it has to happen, is what's called a voltage ride-through. If there's a disturbance in the force,
00:41:02
Speaker
They have to deal with that with batteries or uninterruptible power supplies and not take the grid down because that could be potentially catastrophic and life-threatening. The they here is like an AI data center. Yes. Well, yes. The big ones especially. Yeah.
00:41:19
Speaker
Some of these things, batteries or whatever else, is that I've seen, you know, with uninterruptible power supplies, for example, that there's this phrase power conditioning. Is that basically what they're doing? Is they're smoothing things out?
00:41:35
Speaker
That's what it does. Yeah. It takes the voltage and the frequency and cleans it up. Yeah. And so they're doing this then on a massive scale, and you know, just much larger than like if I have a UPS here at my desk, that's doing it on a very small scale.
00:41:51
Speaker
Yeah. Your you you your UPS s protects you from crummy power coming in from outside. And if there's a lightning strike that hits your distribution wire, it's meant to disconnect and save your computer. If development of data centers and AI continues at its current pace, does that imply something about electricity demand and how we should be planning over the next decade or so?
00:42:20
Speaker
Yeah, I mean, one of the things it implies is we got to hire more trained people. There's a really good article today or yesterday Substack by an economist called Lynn Kiesling. And she said, you know, everyone's focused on the shortage of copper and specialty grain steel and, you know, turbines and all that stuff. We ought to as well be focused on the regulatory and planning capabilities within our utilities and our grid operators, et cetera, because quite frankly, they're overloaded.
00:42:46
Speaker
You can't have three or four X your existing demands. you know, of new data centers wanting to connect to the grid and assume that you can do it with the same people you had before and maybe you add some AI tools to it. No, this is completely different animal than we've ever seen. And we have to respond across the entire board in order to get this right.
00:43:06
Speaker
But then i think what we're also going to see is you're going to need more transmission. There's no way you can't There's no way you can do it without more transmission, right? And today, in today's world, there was a 3,500 megawatt line that connected ah wind energy from New Mexico to Arizona and eventually to California called the Sunzea line. And it just got energized within the last month or so.
00:43:31
Speaker
It took 17 years from the initial planning and first applications for licenses, went through 10 federal agencies, and finally got built. And there were another few lines in the Midwest that are also now in their midway through their second decade.
00:43:48
Speaker
And part of the reason why, like the one in the Midwest, it was almost all approved. And then Josh Hawley and Missouri said, well, we don't want that thing because it's not going to benefit us. And it's going through our farmer's land and it's an eyesore.
00:44:01
Speaker
you know, but you, now you have this issue of, all right, We have typically in this country prioritized states' rights and sometimes individual rights. But hey, when we want to build a highway or a transmission line, oftentimes we invoke eminent domain. There's one company that's not doing that. It's actually going out and talking to all the farmers and the ranchers first and then saying, if you want to hear, don't want to hear, maybe we move it 10 miles to the west and try and get across the checkerboard without taking anybody off.
00:44:32
Speaker
But at some level, If we want to, and it's a big if, let's assume we want to find all these AI data centers, which ultimately will take over humanity and kill us all. So let's assume we want to do that, though. Then we're going to need to, in some cases, ride roughshod over individual rights.
00:44:51
Speaker
Because you can't build things like transmission lines without eminent domain, except in some rare instances. Like, for example, there's a 400 megawatt line taking wind power from Iowa over to Chicago area and they're going to do it by burying high-voltage DC lines ah along the right the railway line.
00:45:13
Speaker
Single right-of-way. The court already said that all the abutters gave away their rights, you know, when they gave away the rights to the railway. So you could take a look at all the railway or, you know, Senator John Ossoff in Georgia, he's saying, why don't we do the same thing with highways? and Under the Biden administration, they mapped out all the highways, all the crossings, where there was hard granitic rock, where there was soft, et cetera. And possibly you could create these interstate corridors for power by burying these lines. It'd be expensive.
00:45:43
Speaker
But right now we're up against such a cultural impasse of getting this stuff done that it may be the only way forward at this point in time. You know, short of ultimate in the future when modular nukes come in, these little guys.
00:45:57
Speaker
you know you could put those, but then a lot of people are going to be like, nope, I don't want to nuke in my backyard. So energy becomes, that's the other thing we were, you were asking about electricity before the electrons never been as politicized as it is today. And it's going to be more politicized tomorrow. I guarantee that.
00:46:13
Speaker
If we are able to build these new transmission lines from, you know, across multiple states, Does that then make the American power grid more resilient or healthier in some way?
00:46:28
Speaker
Well, there's one thing you can do, which they're doing already. So in 2000, I think it was three when first energy didn't trim their trees and it was a hot day. And when you have heat, you're actually your lines physically sag.
00:46:41
Speaker
They droop. And the line touched a tree and it arced. right And so then that line went down. And then what happened was it cascaded to neighboring states and neighboring power grids. And only New England was able to disconnect in time. But New York went down and a whole bunch of the rest of the eastern part of the country. After that, what they did was they put in something called a synchro phaser network where they have these phaser measurement units. And what they do is they're synchronized to the atomic lock and they measure the state on the grid. And if there's a delta between a and B, they know to open up the relays and disconnect this system from that one and protect it. So that was the first time we created situational awareness. So of things like that you can do that make the grid safer and more resilient that way. The other thing you can do I forgot to mention. So the transmission lines, just like the power grids,
00:47:32
Speaker
They're underutilized in the sense that you rate, let's say you a thousand megawatt line, Eric, it's rated for the hottest day of the year. Okay. And power, the ability to move power is affected by heat, by so-called thermal violations.
00:47:49
Speaker
So you can only move, you know, you you can't move more than a thousand megawatts on that hottest, most humid and still day. Okay. Now, if it's the wintertime, you could move 50% or 60% more power across the same line, especially if it's windy, because wind is awesome at dissipating the heat away from the line, right?
00:48:06
Speaker
So what they're doing is they're putting on these things called dynamic line ratings. Some of them are just AI software. Other ones, I saw, i was at a conference and a woman had a bowling ball case, leather bowling ball case. I said, that's a bowling ball case.
00:48:21
Speaker
She goes, yep, it is. I said, what's inside it? And she's like, oh, you walked into my trap. She unzips it. And it's this little thing made by Heimdall. And it's a sensor that knows the ambience, the wind, the temperature, et cetera, and how much current's moving through the line. It's actually powered by some of the current through the line. Drones fly it up.
00:48:40
Speaker
They clip it onto the line, and now you have real-time awareness. So you know, even though the line's rated for 1,000 megs, you can move 1,500 megawatts through the line on January 3rd when it's 25 degrees out, the wind's blowing. Now, it doesn't help you on those hottest days, but it helps you the rest of the time.
00:48:59
Speaker
And then there's also other ways of, for example, you can take your existing wires and poles, and get the wire that's on there and take it down and put on these new ah wires that have a fiber center in them. And they were, fiber is ceramic. It's ceramic, I believe. And they'll give you almost double the power over the same right of way.
00:49:19
Speaker
And it's the right of way that's the limitation. So there are ways that you can actually move more power through the system we have by doing it more intelligently and better tools than we have today. And it gets you part the way down the road, but it's kind of like HOVs on the highway.
00:49:35
Speaker
It'll solve the problem for 10 years, and then ultimately you you outgrow the value of the HOV. You'd be crushed without it, but now you still have all kinds of traffic like we do coming into Boston. But at least it's a temporary fix, and we should do that before we're building a whole bunch of other stuff.
00:49:50
Speaker
Are there things that we have talked about already that would keep a grid operator ah awake at night? Or are there other things that are out there that if you're a grid operator, you're worrying about something else?
00:50:07
Speaker
Yeah, so when OpenAI came out with that neat little um um software or or LLM release recently that was only given to, what was it called, Mystic or something? I can't remember the name of it. But essentially, they only gave it to 40 financial institutions because it found so many weaknesses and ways to hack into software, etc. And you know there's this theory with quantum computing that someday...
00:50:33
Speaker
You'll look at your bank account and you won't see any ones and zeros anymore because they don't exist. Because somebody hacked your password and took all your... Because you can brute force a hack a password in seconds with quantum.
00:50:44
Speaker
It takes forever now. It could take centuries to attack a good password. But with quantum, it's like that. And so there's some concern with AI that... really good software could exploit weaknesses and you know take down the financial system. Well, the grid is super vulnerable cybersecurity. In 2015, the Russians attacked the Ukraine and they hit three distribution companies. I think it was December 23rd, around three o'clock in the afternoon. And what they did was in the previous May,
00:51:19
Speaker
They had sent a whole bunch of you know attachments that people clicked. could be a birthday invitation or whatever. They clicked on the attachments, and that allowed the viruses to penetrate the system.
00:51:30
Speaker
And then there's this thing called like, there's ah ah there's a they have certain phrases for this, but what happens is then you're still in the IT system. Then what happens is the viruses harvest the credentials of the operator so they know their logins. And they get in and then they borrow into the operating technology system, the SCADA system, supervisory control and data acquisition system that manages our transmission lines and our power plants and all the hardware.
00:51:57
Speaker
And in the old days, SCADA was not connected to to the grid. So it was fairly safe because you couldn't find your way in. But then there was an attack in and the UK and in Saudi Arabia on a Schneider system that almost killed someone there. well anyway, the Russians figured out how to do this.
00:52:14
Speaker
So that day, the grid operators and these distribution utilities are at their screens, and all of a sudden, their cursors are opening up the relays, de-energizing the system, right? So they're trying to click, and they realize they've lost control of their screens because they're no longer theirs anymore.
00:52:30
Speaker
because the Russians had hacked that all into it. So they had, I think it was 23 relays that were open. And at the same time, well, how would you know where you lose power? Well, your customers would call, right? Well, the Russians did two more things. First of all, they shut down the power to the control room and took out the back-out power, so they didn't have power. The next thing they did was they went after with um but a botnet,
00:52:56
Speaker
you know basically getting everybody's electronic toaster to send messages and so on. They just flooded the customer care group with useless information. So the utilities had to mobilize and send their crews out and physically re-close all the breakers to re-energize the system.
00:53:14
Speaker
So then a year later, and that time, there were actually physical Russians doing the stuff. A year later, they had another attack. And this time, it was a logic bomb that was planted to go off at a certain time, again in December, but without anybody involved.
00:53:31
Speaker
And what happened in both cases was they found the virus and they brought the software code back to the United States. And I talked to someone involved in it was involved who said it was kind of and worrisome because they didn't know if maybe there were viruses in that code.
00:53:49
Speaker
So they had to be really careful about what they were doing with it. But anyway, the second one opened up a breaker. And when they went out to close it, it was meant then to surge the line and blow up the transformer and kill the worker.
00:54:03
Speaker
And so this one was way more weaponized than the first one. And so there's a concern right now that in this country, because there are over a dozen called APTs, Advanced Persistent Threat Groups, that is people who go to work with suits and ties, essentially. You know, they're not hackers with the hoodies. These are state-sponsored agencies like from Russia's GPU, and China has a bunch of them, and North Korea has a bunch, and Iran has a bunch. And they're doing this full time.
00:54:28
Speaker
The Iranians opened up a dam in Westchester, ah New York, which was a tiny little irrigation dam. It had the same name as a big dam in Oregon.
00:54:38
Speaker
And so maybe they did the wrong one. Right. So these are the things that keep people up at night is the cyber security issues of of how do you protect the grid in the spy versus spy game where there's constantly I mean, there are millions of attacks on utilities on annual basis in terms of penetrating probes, trying to find the weaknesses, where the flaws in the code are and so on. And with ai you accelerate the attack, but you also accelerate the defense. So it's like member Mad Magazine had spy versus spy, the little black guy, the little white guy. Imagine those two on steroids. That's what's going on right now in the AI world around grid protection.
00:55:17
Speaker
The chaos that could be created either at the state level or the you know national level, yeah it seems really dramatic. And then the the economic impacts, the health impacts, you know if hospitals are hit. And it's still difficult even for me for that to really sink in when I hear some of these stories.
00:55:40
Speaker
because it's too dark for us to really let it penetrate beyond sort of the story level. I mean, the one last thing we forgot to talk, I don't think I mentioned in the Texas example, because I really didn't pay attention to it till later.
00:55:51
Speaker
you know, I thought about the fertilizer issues and all that. Credits. Take the ninth largest economy and have nothing being paid. That would have crushed the entire global financial system. Every bank would have been, well, probably insolvent because the daisy chain of IOU, you owe the next person. The whole velocity of money is how fast does the dollar bill move through society? you know Within a year, like how often does that same dollar spend?
00:56:17
Speaker
Well, our institutions are so inextricably intertwined that if you yank out the ninth largest economy for a month, Good luck finding an ATM that works anywhere, right? So, but, you know, by the same token, you and I wake up every day and we rarely check our pulses and don't think about our hearts until we have a heart attack because it's not something we want to do every day.
00:56:39
Speaker
We want to just assume that our ticker is going to keep on going and, you know, maybe we won't eat as much red meat, but for the most part, we just ignore it. And yet, our electrical system in our body That's predicated on the hard functioning, you know, and feeds our brain and everything else. it's It's as critically important to us as, I would argue, the global electric system is critically important to us as our own internal nervous system and electrical system. But yeah, they're there're without it, without juice, nothing happens anymore. so And because now we're so digital, you know, there's like five preppers who could figure their way out of the the first week of this thing if if something went bad.
00:57:19
Speaker
Peter, if you had to make a bet on one energy technology or trend that maybe is underappreciated or is just coming along, is there one that you would bet on this is going to have the biggest impact or we should be investing the most in in the next year, five years, 10 years?
00:57:39
Speaker
So short-term, it's storage, right? Because really, batteries can do so many different things. They are the Swiss army knife of the grid. They can help the distribution grid. Those transmission lines that are stressed out for 10 hours, 15 hours of the year, well, okay, put a battery in there. Charge them when the transmission grid's not stressed, release the power when it is. The data center we already talked about, you know, et cetera. So batteries, and especially longer-duration batteries,
00:58:05
Speaker
They will help a lot. But remember, a battery is not a power source. It's a storage source, and it loses 10% of the energy and the round trip efficiency for a lithium battery and 60% for that long-duration reverse rust battery I talked about.
00:58:19
Speaker
So they can help. But they're not the long-term solution because the long-term solution, we need more terawatt hours of power. And we need it to be carbon-free because, well, for the obvious reasons around climate change. And so then I think you get into the situation of, well, you can build out X amount of renewables if you have transmission because renewables, by definition, are far away from people.
00:58:44
Speaker
It's open space. has to be cheap. And that's far away from people who drive prices up of land And so then the question is, well, how do I get the next energy-dense resources that I'm going to need? And there, it looks like modular nuclear. And interestingly enough, people have been talking about fusion for decades as the energy source of the future that always will be.
00:59:06
Speaker
But you're now starting to see tens of billions of dollars invested in Fusion and some real a physical achievements taking place, like Helion, which is the company that Sam Altman is on the board of.
00:59:20
Speaker
They have a contract right now for 50 megawatts of power to Microsoft, in Chalon Public Utility District in Washington State for delivery in 2028 with teeth in the contract. And I was talking with one of the utility district employees and he said they just took lease on more land.
00:59:39
Speaker
And they also announced they reached 150 million degrees centigrade pretty warm, you can cook a chicken with that temperature real fast. that's I mean, that's the kind of temperature you need to achieve fusion. and And then there was a company that just announced the other day, they raised, I don't know how much it was, it but maybe it was $100 million. And so what they're doing is they're using software to position the magnets that hold the plasma in place.
01:00:06
Speaker
And so now it's not just like a physical piece, it's a combination of physical and software. And so I think what we may be surprised at is to see that, yeah, there's still more solar and more wind that comes on board. But ultimately, we graduate into the next energy sources, which could be modular nukes. And some of those designs are pretty safe. And then you might see your first fusion reactors commercialized by 2033, 2034.
01:00:34
Speaker
Google has a contract with Commonwealth Fusion, from Massachusetts. Commonwealth Fusion in the last capacity auction, or no, interconnection request with PGM, and actually they bid into the auction. They are um intending to connect to PGM.
01:00:53
Speaker
I think they've been to the auction or maybe they're just in the – no, they're only in the interconnection queue. My my my mistake. So they they're not committed to so delivering capacity to the market. They have requested to PJM to be interconnected. And I believe the year of interconnection they want is 2032. It might even be a little bit earlier. So some of these things are now advancing out of the theoretical drawing room.
01:01:16
Speaker
and creeping into the potentially into the real world. And I think we'll know a lot more in the next year or two how strong that thesis is. I typically look at stuff. It hits my radar screen.
01:01:28
Speaker
i don't pay a lot of attention. and then I see a couple of physical demonstrations. But then it's when the pilots actually get commercialized where I start to get excited because you still got to get to volumetric scale.
01:01:39
Speaker
right And that's got to be a while before we really, even if this thing goes along a pretty smooth trajectory, you know fusion of modular nukes, it's still going to be a while before these things develop at scale. twenty Mid 2030s is when you'd see a bunch of them. And then 2040s is where you'd go, okay.
01:01:56
Speaker
I mean, like solar 19, now 2009, we had a thousand megawatts of installed solar. Now it's like over 200,000 megawatts, but it it goes, it takes a little while before it really hits liftoff.
01:02:11
Speaker
Going the other direction in some respects, are there things in the you know near future, one year, five years, 10 years that concern you the most about energy or energy production?
01:02:24
Speaker
The thing that concerns me the most is the fact that climate is such a slow motion disaster until it's not. It's kind of like planning for retirement. You don't really think about it that much. And then all of a sudden it's like, oh, I'm 65.
01:02:37
Speaker
Luckily, i had some good help with that. But I mean, with climate, it's never the top thing on anybody's mind because there's always how do I feed my family and all the insecurity related issues and so on. And yet it sits there quietly And atmospheric carbon accumulates every single year. And then at some point, maybe you hit your tipping points, you know, where suddenly your artery is blocked and you need a stent, but it's not going to be that easy.
01:03:04
Speaker
Right. And so I worry that we're going to get to this point where we go, oh, we should have paid attention to this a long time ago. And now it's kind of late for that. And we only have so much capital. so we're going to spend it on mitigation.
01:03:19
Speaker
ah you know, an adaptation, I should say, excuse me, instead of reducing the amount of carbon coming in, we're just going to build more seawalls and harden our buildings. And then poor societies get completely destroyed. And the rich man, are just like right now in a heat wave, those of us with air conditioning do fine and other people in Greece die.
01:03:38
Speaker
Like that's, that, that's when I look longer term, it's that quiet, persistent accumulation of carbon that really bothers me beyond my lifetime. But when my kids and grandkids are involved,
01:03:50
Speaker
Peter, if you could plant an idea in someone's head, whether it's general public, regulators, anyone else that would actually have an impact, it would change their behavior or their beliefs.
01:04:07
Speaker
Is there something that you would want people to to believe or to think that would have a ah positive impact on the future as you see it? I mean, the first thing I would want people to understand the carbon related issue, but then because of, because now it's boiled down to an issue of, well, I believe in climate change, believing in God, that's different than believing in climate change. Like one, you can actually look at statistics and numbers. And the other one is you never know until you die. And maybe then we still won't, right? Like it's not a belief system. It's,
01:04:41
Speaker
Do you think that the scientific method of inquiry, despite its flaws, generally yields beneficial results that we can more or less trust, right? We predicated our society on science. So why do we all of a sudden discard it when it's inconvenient for us? You know, to borrow the phrase from Al Gore. But the thing that I would, if I could wave a magic wand, the thing I would do is I would make everybody educated,
01:05:09
Speaker
Like so that the conversations were fact-based instead of supposition-based. And same true for the power grid. I sometimes see people just sort of wishing that solar and wind are going to solve all the problems when they don't. And in fact, my own trajectory has been that way. The more I understand the physics of the grid the more and the economics, the more I understand how tough the challenge is.
01:05:35
Speaker
And you know you can model something, but that's different than the reality of the world oftentimes. And I've spent the last 13 years trying to understand this beast. And it's every time I look, there's a new frontier of my ignorance that I need to figure out. Now, not everybody has to be obsessed like i am and spend three hours a day on it, trying to understand it, but at least having some kind of a general understanding. The thing is so critical to our survival that I find it astonishing that people don't spend any time understanding how it works because without it, we're all dead.
01:06:09
Speaker
You know, like that just blows my mind. Peter, where should someone go to learn more, follow you, connect, or or otherwise follow up on this?
01:06:20
Speaker
So there's two books that I would recommend. One is The Grid by Gretchen Bakke that provides a great historical subject. And she's an anthropologist by training. So she talks about Edison and and sort of how the grid works. And I found it, and Bill Gates liked it, and I found it immensely entertaining. And then as that was coming out or maybe a little bit after that, I wrote a book called The Energy Switch was meant to do what we just did, which is short circuit the difficult learning process and contextualize everything that I had learned up until 2020 And so...
01:06:56
Speaker
and so People have told me, numbers of professionals and others, that it was a very approachable way because I just have a chock full of stories to understand how the grid works. like Hers is a little bit less technical. Mine, this is how a generator works. This is how an electric vehicle works. This is you know all the pieces and how they fit. And then um um also, if you want to reach out to me, I read 43 newsletters. We discussed that, I think.
01:07:22
Speaker
I have them all in a spreadsheet and they're ranked by topic. And I have the signup URLs for every single one of them. So if you want to go into modular nukes or data centers whatever, I got all all that information.
01:07:33
Speaker
My job is really simple. I'm not a doer anymore. My only job is to cross-pollinate information as quickly and effectively as possible and as approachably as possible so that other people become more effective at what they do.
01:07:50
Speaker
Right? That's all I do now. <unk> I'm no longer a doer. And it's, you know, it's something when you're sort of an emeritus, you can do. And so the that's that whip that spreadsheet. And then I also do a weekly video and they cover everything from solar perovskite or sodium battery technology or the latest thing in AI or whatever. ever Those are on LinkedIn. And then I have a website, peterkelly.weiler.com, where you can access all the videos there. so Those are some areas that might be helpful you know for folks who want to understand more about the space tied into what I'm doing. Then there's like Canary Media and Utility Dive, which I mentioned in the list. But there's a few of those that if you want to understand more about the space, they're good sources for that.
01:08:33
Speaker
Peter Kelly Detweiler, I appreciate you being here for the second time. You're the first person and I'm really glad that it's you. So thank you for joining me today. Oh, it's my pleasure. you are You come up with the right question at the right time, Eric. So ah thanks for leading the conversation.
01:08:50
Speaker
Thank you very much.