Transcript
INTRO
Jigar Shah: We just had one of these crazy experiences. I dropped my kids off at Legoland, rushed over to the recording studio to record the podcast, and honestly, Legoland is pretty decent, I have to say.
Jamie Nolan: Well, I appreciate your level of commitment. I see you're joining us wearing... what is on your shirt? Are those crabs?
Jigar Shah: Americana. Look at this. Rosie the Riveter.
Jamie Nolan: Americana, all sorts of stuff. Look at you with your patriotism, and your commitment to the pod that you're even recording from vacation. How does your wife feel about that?
Jigar Shah: She's not happy, but at the end of the day, there were only so many times we could reschedule, so I said let's get this done. My ten-year-old is doing great, though. He was excited. He said, "You're giving me the freedom to just roam around the park?" I said yes. He said, "This is amazing."
Jamie Nolan: Amazing.
Jigar Shah: The conversation we're having with Amit Narayan from GridCARE is a really interesting one for me. These topics can sound so boring, because it's really about an esoteric thing within the grid: we operate the grid around the worst possible outcomes. You say, if these five things occur at once, we need enough extra capacity to ride through and keep everyone powered. But what that means is there are one hundred thousand megawatts of extra capacity in the grid that could be unlocked right away with the right software. He went to Portland General and found 450 megawatts of capacity they didn't know was there, and as a result rates are going down five percent for everyone in that territory. He found 650 megawatts in National Grid New York's territory, same thing. It has real-world consequences for affordability.
Jamie Nolan: Those numbers speak for themselves. If you're a utility executive seeing that data, I don't know how you aren't calling them. Grid utilization is the topic of the year. You cannot turn around without hearing about how to get more value out of the grid we've already paid for, and here's a company with an actual solution and very clear ROI.
Jigar Shah: It follows on from the fantastic episode we had with Secretary Granholm, about what governors can do. In this moment, governors need to become PhDs in how electric utility systems work. They need to know solutions like this exist, because if they don't, it's hard to mandate it onto their utility.
Jamie Nolan: They either get the PhDs or hire the equivalent. A lot of these solutions have been waiting in the wings for their moment, and I'm excited we're finally seeing the real-world benefits. Amit has the best example of what this looks like and what it means for everyday people in money back on their utility bills.
WHY GET BACK IN THE GAME
Jamie Nolan: Amit, everyone is acting as if the grid is maxed out. Interconnection queues are years long, and the assumption is we need to build massive new infrastructure. You're arguing something different: that a big chunk of the problem is we just can't see what's already sitting there on the grid. Make that case for us.
Amit Narayan: First, I'm excited to be here. Jigar, you've been a personal inspiration over the years, across your startups and your public service. A big part of what I'm doing today is trying to follow your footsteps.
Jigar Shah: Kind words are appreciated, but we love what you're doing now. You've had all these success stories, companies you started, AutoGrid, which you sold to Schneider Electric. We're excited you're back in the game. Having the best entrepreneurs in the country back in the game is critical.
Amit Narayan: I've been on both sides of this AI infrastructure buildout. First I was writing software to design the semiconductor chips that drive the demand for compute. For the last several years I've been on the supply side, making sure energy supply can keep up with that demand. When the two movements collided, it felt too big to sit out. When I sold my previous company, the one thing I had explicitly ruled out was becoming the CEO of another startup. But with the AI race going on, the next three to five years will decide the shape of humanity for probably a hundred-plus years. We have a challenge I call the time-to-energize challenge in the US, and I think we have a unique insight into how to address it. That's what brought me back.
AMERICA DOESN'T HAVE A POWER SHORTAGE
Jamie Nolan: Your case is that a big part of our problem with the grid is that we can't see what's already there. Make that case, and how does GridCARE address it?
Amit Narayan: Our fundamental thesis is that America doesn't have a power shortage. We have a lack of visibility into what already exists. And I want to be clear: I'm not against building. If power demand is going to double or triple in the next decade, we absolutely have to build more transmission, more generation, more substations. But the pace of pouring concrete and getting permits is not keeping up with the speed at which AI needs to be deployed. So the solution may be hiding in plain sight. If we can optimize what we already have in the ground, that's the fastest, cheapest, and cleanest way to power this revolution. We published a study showing that when you look at the grid in full granularity, it's only about a third utilized. That doesn't mean it's idle all the time, and it doesn't mean reliability constraints go away. But if we shine a light on it and get smarter about it, we can squeeze a lot more out of it. And if we can do it in code, not copper and concrete, we can do it in six to twelve months, not six to twelve years.
THE VALUE OF A MEGAWATT
Jigar Shah: They call it speed to power, and you're right in the middle of it. You just got validation. You raised sixty-four million dollars in your latest round, with early wins at Portland General and National Grid New York. Talk about the investment environment. What are these investors investing into, and what's the revenue model that gets them the returns they want?
Amit Narayan: Start with the value of a megawatt. Right now, the revenue AI is generating is somewhere in the range of ten to twenty million dollars per year per megawatt of capacity that can be unlocked. For a hundred-megawatt data center, that's billions of dollars of value a year. The urgency is so high that hyperscalers are willing to spend money on almost anything that gives them hope, including sending chips to space because they think they've run out of capacity on Earth, and investing in everything from nuclear fission to fusion. I'm a technologist and I believe in those things, but even in the most optimistic scenarios, those timelines are five or ten years out. The AI race is happening right now. There's simply nothing else that can be done at this speed and scale other than making use of the infrastructure already in the ground. Our lead investor this round is Sutter Hill Ventures, who incubated NVIDIA, along with John Doerr, the original investor in Amazon and Google. They see that power has become the defining constraint for the AI revolution, and we're offering a solution that isn't speculative. It's already there and can be deployed at scale. If we capture even a small part of the value we create, it's a tremendous upside for us, our investors, and all our stakeholders.
WHY THE GRID RUNS AT A THIRD
Jigar Shah: Let's dig into the analogy. The grid is used more during peak and less the rest of the time. Drew Maloney at Edison Electric Institute used a bus analogy recently. The other one people use is a restaurant: you have a kitchen, staff, the ability to cook and fill seats, and when the seats aren't sold, that's low utilization. Paint the picture of why utilization is only a third or a half, depending on the market.
Amit Narayan: The restaurant analogy is apt. I use airplanes on Thanksgiving, but it's the same idea. The grid does get constrained, but the precise scenarios under which it gets constrained are fairly rare, and even then they last only a few hours. Historically, getting full visibility into when those rare events happen wasn't feasible. Go back twenty or thirty years and you had limited compute power. That led to planning methodologies that kept the system reliable and safe but left a lot of room on the table. What we can do now with modern AI is look across time and across many locations at once and find the precise conditions under which the grid gets constrained. To do some quick math: if you have ten thousand things on your grid, the NERC standard requires you to study all possibilities of two simultaneous outages. One thing can fail in ten thousand ways and the second in another ten thousand ways. Right off the bat you have a hundred million scenarios to calculate.
Jigar Shah: That's shocking. Honestly crazy. And people used to do that analysis by hand, with General Electric or Siemens engineers.
Amit Narayan: It's brute force. You can't run a hundred million simulations, so you make worst-case assumptions. You say my load might be a one-in-ten-year maximum, and if the two worst outages happen at the same time, what's the capacity of the grid? But you wouldn't look at the probability, or at all the hours across all the years, because that computation didn't exist. The other thing that's happened over the last ten or fifteen years, thanks in large part to your work, Jigar, is that a lot of new technology has been deployed on the grid. You funded batteries, you funded virtual power plants. There's much more dynamism on the operations side, and these aren't unproven; utilities dispatch them at scale. But they aren't properly accounted for in the part of the utility that calculates how much capacity is available in the future. That's a big part of what we do: create visibility across the utilities and the customers. Planning processes are worst-case, looking ten years ahead. Operators look seconds, minutes, a day ahead, but they don't look across tens of years. Now, because of the computation, we can do that granular, second-by-second analysis but look at quadrillions of scenarios over the next ten years and figure out exactly what can go wrong. Then we figure out the best ways to mitigate those rare events, maybe making the load a little flexible, adding a battery, or using the virtual power plants already in place and dispatching them more intelligently. It's not rocket science, but it wasn't possible with the tools utilities had until very recently.
WHY GRIDCARE EXISTS
Jigar Shah: Let's dig in further. Part of what you're saying is that there are multiple silos within the utility, each with different data and different responsibilities, and none of them truly understands how to fit a data center into their grid. So take us step by step: why do you need to exist? Why do folks need to use you?
Amit Narayan: This is not just planning, and it's not just operations. You have to bring together many parts of the utility that have historically lived in silos and don't share the same view. But it's more than the internals of the utility. We don't unlock capacity by telling utilities to become more efficient. We also expect customers to behave differently and become better citizens of the grid. One challenge is the communication gap between customers and the way utilities work across their functions. So we come in with a fresh pair of eyes, as a neutral third party, and try to build a bridge between the utilities and their customers, giving visibility to all the stakeholders, including regulators, ISOs, and the different silos within the utility. Once we create that common picture, we can address each stakeholder's objections. Planners want to make sure reliability isn't compromised if they add a new load, and we quantify that. Operators want to know whether the flexibility is qualified, whether it will show up, and who takes the risk if it doesn't, and we quantify that precisely. Customers want to know how much flexibility they have to bring, whether they'll be curtailed and how often, and whether their compute will be affected. Regulators want to know the impact on rates. Affordability is a big issue, and the current narrative is that data centers come in, take all the power, and drive up rates for everyone. A big part of our value proposition is that by using the infrastructure we already have more effectively, we can actually help reduce rates for everybody.
Jamie Nolan: When you say customers, do you mean hyperscalers, all types of large load, or homeowners?
Amit Narayan: Anybody who gets electricity from the network. The network is a fixed cost we all pay for based on usage. If we can increase utilization from thirty percent to forty percent, that brings the cost down for all consumers. We published a white paper recently showing that every new gigawatt of capacity added on an existing network using our approach brings rates down by roughly five percent for a typical mid-size utility, which is substantial.
THE UTILITY BUSINESS MODEL
Jamie Nolan: If utilities ran the airline business, they'd get paid for having enough planes for the busiest travel day of the year and charge everyone more all year to keep that capacity. The incentive is to build more, not to get more out of what they already have. So how do you sell a product that helps utilities use their grid better when their business model rewards them for building new grid?
Amit Narayan: That's another legacy assumption due for a challenge. In an economy with no growth, historically the only way utilities made money was by building more infrastructure, usually justified in the name of reliability, which by definition means overbuilding. What's changed in the last one to two years is unprecedented growth in demand for AI. Now utilities realize they're in a competitive environment. The earliest they can build a new transmission line might be five years out, but if they can earn more revenue on existing infrastructure, their earnings go up today. That's one big change, especially for investor-owned utilities. If they don't attract this growth, another utility might.
Jamie Nolan: We're going to call that your hot take of the episode.
Amit Narayan: The second driver may be even more important, and it's universal, including for public utilities: the crisis of affordability, which has become a bipartisan issue. Even when utilities want to deploy capital, they're not getting rates approved, because the historical approach was to rate-base it, which raises rates for everyone. The theory is that eventually the benefits outweigh and rates come down, but that mostly doesn't happen. So legitimate upgrades get pushed out because of community backlash on affordability. We make it possible to stop viewing data centers as a trade-off between growth and affordability. If we use the asset more effectively and bring large loads on sooner, everyone benefits. The data centers benefit, the utilities get their earnings sooner, and because we're amortizing the cost of fixed infrastructure over more electrons, we can reduce rates. Our white paper shows that every new gigawatt that comes online on existing infrastructure can cut rates by about five percent for all consumers, residential and commercial.
BRIDGE, NOT MEDIATOR
Jigar Shah: It sounds like you're a mediator. The value of a mediator is that the data center trusts you'll do a better job speaking utility, and the utility trusts you'll do a better job speaking data center. Is that right?
Amit Narayan: Mediator is a decent analogy, except I don't like the assumption that the two sides are necessarily fighting. That might be the case sometimes, but we're in a moment where both sides want to work with each other and just don't know how. So I think of us more as a bridge. Both sides have to do a few things differently, and if they do, they can meet in the middle, which is good for the utilities and the data centers.
AI IN THE SERVICE OF AI
Jamie Nolan: There's a major irony here. AI is creating the biggest surge in electricity demand in a generation, and the solution to finding the power for AI might actually be AI. You're using generative AI to solve AI's own infrastructure bottleneck. Is it as circular as it sounds? Do you personally believe AI will save us from some of the problems it created?
Amit Narayan: Absolutely. We can call it irony; I like to think of it as a virtuous cycle. That's the promise of AI. The more intelligent we become and the more we squeeze out of existing infrastructure, the more abundance it creates. The way out of the current situation is to become smarter about what we have and get more out of the system we've already deployed. And as we grow, that intelligence lets us deploy new capital far more efficiently too.
AN MRI FOR THE GRID
Jigar Shah: Where does this end up going? When you work for a data center client, are you mapping out the utility grid for them, handing them a map that says you could put a five-megawatt EV charging system here and fit another 185 kilowatts of manufacturing load there? Does this become a real-time map of load growth?
Amit Narayan: Not just load growth. There's the other side of the equation, around 2.6 terawatts of generation stuck waiting to connect. Right now these processes run in silos; when you study generation, you don't look at the load, and vice versa. This is really an MRI for the grid. If you know where the constraints are and when they happen, you can add load alongside the data centers, and free up power for other critical infrastructure, hospitals, manufacturing plants, electrified transportation. We need to look at the grid as the system it was designed to be. Once we understand it, we can double its capacity, and even then we'd only go from thirty percent utilization to sixty. It's not like climbing Mount Everest. But that would mean more than 300 gigawatts of capacity unlocked on existing infrastructure, which we can do in the next three to five years. There's literally no new technology that needs to be developed for that.
SOFTWARE PROBLEM OR TRUST PROBLEM
Jigar Shah: It's massive, and I agree. But is this really a software problem, or is it a contract-and-risk question? Is it really that utilities need to change their posture to unlock this three hundred gigawatts? I'm conservative and say a hundred gigawatts, but I'll take your number. Is it a risk posture, a bunch of people in the bowels of the utility saying we could never run the grid that way, it would break, we need all the things my grandfather taught me to do?
Amit Narayan: It's a technology issue in the sense that technology enables the unlock; we didn't have this compute power before. But it doesn't stop at technology. Ultimately it's a trust challenge. Every stakeholder in the grid ecosystem needs to feel comfortable that what we're doing is fair to them. Technology lets us create that trust and the contracts that go with it, so that when a data center gets power, it isn't unfairly taking it from another customer, and in return it brings the flexibility needed to keep the system reliable and potentially lower rates for everyone. The data centers have to trust it works for them. Ratepayers and their advocates have to believe it brings rates down and that we're not allocating capacity willy-nilly. Regulators have to be confident it's done transparently and neutrally. Every part of the world structures and regulates its energy markets differently, so we have to prove it one situation at a time. What gives me hope is that most utilities face the same challenge: they want to solve affordability, keep growing, and enable economic growth in their regions. The physics of the grid works the same way everywhere. I've deployed grid technologies in more than twenty countries, and it's the same physics, so if it works in Portland, it works elsewhere. In North America the NERC reliability rules are consistent, and everything we do is under current NERC and FERC regulation, so it's very transferable. But some stakeholders have to be brought along to make it truly scalable.
WHY IT'S VPPs' MOMENT
Jamie Nolan: Amit, VPPs are not new. Jigar and I have been banging the drum on virtual power plants for years. It feels like this is the moment, and we're starting to feel the scale and momentum. Why is it finally VPPs' time to shine?
Amit Narayan: Like both of you, I've been in the VPP business for more than ten years. For most of that time, VPPs were a solution looking for a problem. We promoted them as a way to add intermittent renewables or to hedge prices, so we wouldn't have a scenario where supply is constrained and demand isn't flexible. Those were good use cases, and we got fairly far, but the economic value was never high enough for data centers to take on the operational burden or perceived risk. We deployed about eight gigawatts of VPPs across twenty countries, and the only data centers that participated were Bitcoin mines when Bitcoin prices were in the toilet, because that was how they made money. But none of the cloud data centers.
Jigar Shah: It all comes down to Bitcoin, Amit. Didn't Crusoe start as a Bitcoin miner and now they're a big data center company?
Amit Narayan: Right, and that's great, because those Bitcoin companies were already forward-thinking about energy and flexibility. They're some of the most innovative companies out there, and they realized flexibility gives them a leg up. The real change now is that every megawatt you bring online creates twenty-five to thirty million dollars of new economic value for a data center. So the prize is big. When a data center is given the choice of waiting eight years and spending hundreds of millions on upgrades, or connecting right now if they add some flexibility, they choose the latter, because it's a no-brainer. In Portland we offered a flexible connection to six data centers. All six took it. They're all energized now, and they all want more. So I don't think there's any real debate about whether data centers want to be flexible.
CUTTING THE LINE OR CUTTING RATES
Jigar Shah: You keep coming back to trust. Are you seeing the broader public view your work as a way for data centers to cut the line and get capacity faster, or as a way to reduce rates and make the grid more affordable? I get your white paper, but are people believing it on the ground?
Amit Narayan: There's definitely more work to do on educating the public. Data centers can be done the right way and the wrong way, and we have many more examples of data centers coming online and raising rates than of what we're doing. It's an education and awareness issue, but the science is on our side and it isn't difficult. It's numerator versus denominator. If you increase the denominator and keep the numerator the same, the number goes down.
Jigar Shah: Exactly right. You want to limit investment in the numerator and sell more kilowatt-hours in the denominator.
NO PILOTS
Jamie Nolan: We've spent a lot of time in the last few months on grid utilization, and there's this data disconnect. Utilities claim they don't have the distribution-level data they need to optimize it. So it's refreshing to talk to someone who says, we're over here, we've got that for you, give us a call. I'm excited to see what happens with your company. You were named one of Fast Company's Most Innovative Companies of 2026, congratulations. But innovation in energy so often means piloting something for years and never scaling it, which is basically Jigar's personal mission in life to fix. What makes GridCARE different from every other promising grid tech company that got stuck in pilot mode?
Amit Narayan: We don't do pilots. My challenge to utilities is: bring me your toughest, most visible, most valuable problem, and we'll find you an answer in less than ninety days. We've done more than a gigawatt in the last six months, in places that were considered hopeless when we walked in, and we found more capacity than even we thought was possible. We're finding more still. At this point we're in more than a dozen markets and actively analyzing more than ten gigawatts of projects. Part of why we've been received so well is that we offer a fairly risk-free proposition. Our message to the utility is: give us a chance. All we need is data you already have, because you have to run these studies anyway, and in ninety days we'll find an answer. If we find more power, you're all heroes: you increase revenue, reduce rates, and your governors love you. If we can't, you have the satisfaction of knowing you tried everything. So what do you have to lose? Our bet is that because the grid runs at about thirty percent of capacity, we'll almost always find power, and so far that's been true. Same with data centers. They can stay in the queue and wait for space-based data centers, or give us a chance. It's a net-new voluntary option: if they get the power, they figure out how to pay us for the value we created. Otherwise nobody is worse off.
Jigar Shah: I feel like they want to go to space for other reasons.
Jamie Nolan: It feels like you're looking in the couch cushions and finding a million bucks.
Jigar Shah: Twenty-five million per megawatt.
Amit Narayan: It's like the change you find in a drawer when you open it, except here it's trillions of dollars. And it's a competitiveness issue that can literally define the course of humanity for the next hundred years. So the stakes are much bigger.
A GLOBAL PROBLEM
Jigar Shah: We talk a lot about the United States, which I love, but is this really a Western grid opportunity, Europe and the US? Or if you went to China, where it seems like you can build grids easily, would you find similar utilization and a similar opportunity?
Amit Narayan: Suboptimal grid use is a fairly universal problem. In places where building is easy, you can live with the inefficiency and keep building. But in the US, and a big part of Asia, the grid is constrained and building isn't easy. We're getting a lot of inbound from Southeast Asia, India, Japan, Australia, and most of Europe, Eastern and Western. In those places the grid is constrained and building new infrastructure takes time, whether it's supply chains, permits, or regulation. What you can do with the existing grid becomes extremely valuable, because it's already there and all you need to do is get intelligent and unlock it.
THE RIGHT QUESTION IS UTILIZATION
Jigar Shah: My final question, or statement. If you're one of the 36 governors with a race this year, and you just heard you can unlock three hundred thousand megawatts of capacity, why should you keep expanding the grid broadly and adding generation? Why do all the measures at once when you've got a silver bullet in GridCARE?
Amit Narayan: My message is that it's not one or the other. You want to optimize what you have.
Jigar Shah: It kind of is one or the other, though. People want affordability, and they'd love to put all the weight of the world on your shoulders, and you've got broad shoulders. The pressure I see is people arguing that if we focus on GridCARE, we're not keeping the pressure on new transmission lines, and we'll regret it one day.
Amit Narayan: I don't think that's the question. The question people should ask is utilization. If there's real demand and a new transmission line brings that demand sooner in a way that raises utilization, then I'm fine with it, that's the right decision. But if you build something and utilization goes from thirty percent to twenty-five, then by definition the other ratepayers are paying for it. It comes back to numerator versus denominator. So for any project, a new data center or a new transmission line, the right question is whether it raises the effective utilization of the grid. If it goes up, it's good. If it goes down, it's a cross-subsidy being handed to some people in the system.
Jigar Shah: A man after my own heart.
Amit Narayan: You're the inspiration. I hope we can make this the default.
Jigar Shah: It's wonderful to have you, Amit, and thank you for jumping back into the CEO fray. You have an excellent track record of delivery. Your previous companies were so successful, and you figured out how to get electric utilities to sign big contracts, which is what we need today. The trust gap you describe is real. Thanks for getting back in the fray.
Amit Narayan: Thank you for all the support, good wishes, vision, and inspiration. It's been a privilege to be on the show.
DEBRIEF
Jigar Shah: He has so many numbers. It's crazy. I always thought we were managing about fifty percent utilization of the grid, and he used thirty percent. He was basically glass-half-empty about how much of our existing grid we're using. But in some ways that's even better, because it means there's so much more capacity to unlock.
Jamie Nolan: One thing I've learned working on grid utilization this year is that a lot of these utilities and system operators don't know how much of their grid they're using. They truly don't have the data, and they don't have the tools internally to understand how much of their system they're using. It's crazy when you think about it. And here's a company that can come in and map that local grid. It's like finding a million dollars in your couch cushions. It's there the whole time, just underutilized.
Jigar Shah: At 650 megawatts, that's a billion, or more. I think it's fifty billion dollars now to build a thousand-megawatt data center. So reduce that to 650 megawatts and it's something on the order of thirty-five billion dollars of value. It's crazy.
Jamie Nolan: It's wild. I'm so happy these technologies are finally getting their due, and getting the big contracts and relationships with utilities where they can provide value to the utility, the end customer, whether a data center or a manufacturing plant, and everyday people through downward pressure on their utility rates.
Jigar Shah: The thing I don't understand is why it isn't just mandatory for all 160-plus investor-owned utilities to hire somebody like this by the end of the year.
Jamie Nolan: Maybe because their regulators don't know solutions like this exist, but there are so many. I think about the work the Utilize Coalition is doing, a group of service providers with different solutions to better use the grid we already have. And we're seeing some states act. Virginia recently passed a law to actually analyze how much capacity is available on their grid. A lot of these utilities don't even know their current grid utilization. Putting them in a posture where they have to prove they need the poles and wires before they build them seems like common sense.
Jigar Shah: Totally. It feels like a big unlock, and I'm glad to see folks starting to use it.
Jamie Nolan: It was wonderful talking to Amit and learning about his solution and all the hard data he has about the value. Let's hope more utilities get wise to this.
ASK JIGAR
Jamie Nolan: Welcome to Ask Jigar, our weekly segment where Jigar answers your questions about energy and honestly anything else you're wondering about. A quick reminder: anyone who sends in a question that Jigar answers on the show gets an Energy Empire hat. Our first question comes from Tommy Berbas. He asks: Jigar, what's your read on Freedom Forever's recent bankruptcy? They were one of the largest residential installers in the country. What does it signal for the retrofit residential market, and why is it so hard for retrofit residential companies to survive in the first place?
Jigar Shah: It's a great question, and one that's come to the fore in the last year because we've had so many bankruptcies, not just Freedom Forever but Sunnova and SunPower and over a hundred installers. The market they serve is existing residential homes. Someone knocks on your door and says, would you like solar panels on the roof and a battery in the garage? If you say yes, they file permits, do all the things, and a lot of them get financed, from people like Sunnova or SunPower in the past, and today Sunrun or Palmetto. Over time, we had low interest rates and a lot of competition between GoodLeap and Sunrun and Sunnova and others. They'd give people credit, milestone payments while you're installing, wiring more money into your account each time you hit a milestone. So installers started running leaner on cash. Then as things got worse for a company like Sunnova, they started paying late, net 30, net 60, net 90. And for a company like Freedom Forever, that's a problem: I'm running tight, I don't have extra money to pay salaries while you pay me late, and my balance sheet isn't strong enough to get a hundred-million-dollar loan to cover those losses. So the big problem is we ran everything too hot, and when higher interest rates and weaker solar financing companies hit, there was a domino effect that took down Freedom Forever.
Jamie Nolan: Our next question comes from Nikhil Vinod. He says: Jigar, here's a what-if. Say AI infrastructure turns out to be overbuilt and the energy it actually needs comes in well below the predictions. In that world, what becomes the main use for all the extra generation we'd have built?
Jigar Shah: It's a good question, and I think we're not quite at that stage yet. There are, let's call it, thirty-five one-gigawatt-plus data centers that have announced they're under construction. When you look at the actual data, most are way delayed. They were supposed to start construction in 2026, and most have slipped to 2027 or 2028. The bigger risk isn't that the load doesn't show up from AI. Lots of people are using AI, and there's going to be a lot of AI load. The real disruption is local versus central. There's training, which is training a new model, and right now that's maybe 500 megawatts of load. That's predicted to reach 4,000 megawatts by the end of the decade to train a model. Then there's inference, you using the Claude app or the ChatGPT app or building an agent, and a lot of those tasks can run in a five-megawatt data center near your home, or a 500-kilowatt data center in an abandoned office building, or a 100-kilowatt data center at the back of a telecom tower, or even a 50-kilowatt one like our friend Art Rao is experimenting with at Span.io in his partnership with NVIDIA and Pulte Homes. If we move inference to the edge and don't need these one-gigawatt data centers, that capacity could become excess. Right now we're not there, because we're so far behind in building. I think we'll figure out whether we can move data centers to the edge by the end of this year, and I think we'll succeed, and a lot of these one-gigawatt data centers go away.
Jamie Nolan: Next, Oluseyi Smith says: Jigar, I work in hydropower, and I've always wondered why it never gets the attention that wind, solar, nuclear, and storage do. Is it the price? The assumption that the good hydro resources are already tapped, which Canadian advocates say isn't true? Why doesn't hydro generate the same excitement, and is it doomed to follow coal toward obsolescence? I know that when we worked at LPO, we heard from the hydro industry that we weren't investing a lot in that technology. What do you think, Jigar?
Jigar Shah: I spent a lot of time on hydro while we were serving, and there are a couple of big problems. One is that most of the best sites in the US have been used. That's not true for Canada, but it is for the US. The sites available now are much smaller, and they're largely owned by the Army Corps of Engineers, so you need their permission to add power to their non-powered dams. And if you remember, we couldn't finance those out of the Loan Programs Office because of the federal nexus; we couldn't lend to something the Army Corps was part of. The other problem is that we have a lot of old hydro dams, and if you upgraded them with new software and equipment you could double or triple the energy they produce, but it takes about five years to get through the Federal Energy Regulatory Commission for approval. I know the Trump administration is trying to streamline that, but they're working more on nuclear than hydro. And the last piece is that the people who own hydro assets are sticks in the mud. They're electric utilities, or older owners who love the money that comes off them, and when you try to buy the assets to upgrade them, they make your life a living hell. So you end up saying, I'll work on geothermal and nuclear instead. Hydro needs a refresh, and we never did a hydro liftoff report, but we need to.