This year alone, the biggest tech companies plan to spend more than $600 billion on physical infrastructure — eclipsing the railroad boom, the interstate highway system, and the Apollo space program.

But are investors starting to flinch?

This week, we examine the negative market reaction to tech earnings. Is Wall Street reacting to the infrastructure bottlenecks that stand in the way of building at that scale? Or are they worried about the tech industry’s approach to solving them?

Then we turn to one of the boldest responses to those bottlenecks: space-based data centers. After SpaceX’s acquisition of xAI, Elon Musk says orbital computing powered by solar could be imminent. We unpack the arguments for and against space-based data centers.

Then we look at solar. Musk says Tesla plans to build 100 gigawatts of domestic solar manufacturing capacity. Tesla has launched a new panel and mounting system that it claims will reduce installation time by 30%. At the same time, a new poll from Trump’s chief pollster shows majority support for solar among GOP voters — especially when panels are made in America. Is there a vibe shift underway?

Ready to accelerate your career in clean energy? Yale’s Financing and Deploying Clean Energy Certificate is a fully online, 10-month program built for working professionals. It delivers real-world skills in clean energy policy, technology, project finance, and innovation — all in just five hours a week. Enroll here and use the discount code OpenCircuit26 on your application to save $500 on tuition. Applications close April 20, 2026.

Explore the new era of AI innovation in the fifth season of Where the Internet Lives, an award-winning podcast from Google and Latitude Studios. Follow and listen to Where the Internet Lives on Apple, Spotify, Google, or wherever you get your podcasts.

Join Latitude Media on April 13-14, in San Francisco for Transition-AI 2026, a two-day, in-person conference on the digital and energy infrastructure buildout needed to support AI load growth. Our podcast listeners get a 10% discount on this year’s conference using the code PODS10. ⁠Register today here⁠!

Transcript

Stephen Lacey: God, we have so much to cover this week. But you know where we have to start, with Bad Bunny climbing a utility pole during the Super Bowl halftime show. Did you guys see this?

Jigar Shah: I loved it.

Caroline Golin: Absolutely, yeah.

Jigar Shah: I loved it. I was getting so many text messages from all of my DOE friends, because we just did so much work in Puerto Rico. And so, it was great to see him sing that song, but also feature the blackout.

Caroline Golin: The blackout aside, it was such a happy performance. I mean, I was hard to watch that and not feel joyful and a little more positive about our future.

Jigar Shah: I’m sure some people were like, “That’s a weird pole to pole dance on.”

Stephen Lacey: Yeah. But people in our industry are so desperate for people to talk about energy issues that they will take it.

Caroline Golin: I mean, when has energy infrastructure ever made it to the Super Bowl halftime? I mean, if you don’t feel relevant after today, I don’t know.

Stephen Lacey: From Latitude Media, this is Open Circuit. For the last few decades, the tech industry has been largely a software industry. That is now completely flipped. This year alone, the biggest tech companies plan to spend more than $600 billion on physical infrastructure. That’s data centers, substations, generation, eclipsing the railroad boom, the interstate highway system, and the Apollo space program. But are investors starting to flinch? We’re going to take a look at tech industry earnings this week and what the market reaction tells us about the limits of that expansion.

Stephen Lacey: Then, limits? What limits? After SpaceX acquired xAI, Elon Musk is setting his sights on space-based data centers, and he wants Tesla to build a hundred gigawatts of domestic solar production to aid empowering them. And finally, a new poll shows once again that conservatives overwhelmingly like solar. With Musk talking a big game about US-made solar panels, and even Stephen Miller’s wife publicly hailing the technology, are we in the middle of another vibe shift on solar?

Stephen Lacey: Hey, everybody. Welcome to the show. I’m Steven Lacey. I’m the executive editor at Latitude Media and one of your co-hosts. Caroline Golin is my other co-host. She’s the CEO of Envision Energy Advisors. How are you today?

Caroline Golin: I’m well. I’m feeling better than last week, and I think we are all out of the clear from the flu in the family.

Stephen Lacey: Apparently, the Monday after the Super Bowl is when people call in sick to work the most, so you could have used that as your excuse.

Caroline Golin: I could have, I could have. But no, I can use my four children anytime I need to.

Stephen Lacey: Jigar Shah is our other co-host. He’s the co-managing partner of Multiplier. How are you, sir?

Jigar Shah: I’m good, I’m good. I think there’s a lot of public schools that just give the day off after the Super Bowl.

Stephen Lacey: Is that true? No.

Jigar Shah: Yeah, because they’re just tired of people not showing up to school the next day.

Stephen Lacey: Well, I think that next year’s Super Bowl times up with a federal holiday on Monday, so get ready.

Jigar Shah: Oh, is that President’s Day?

Stephen Lacey: Yeah, I think so.

Stephen Lacey: Well, let’s start this episode, which is going to cover a lot, with a look at the public markets, where the storylines around AI are continuing to evolve. The big story coming out of fourth quarter earnings is the crazy amount of CapEx that the top technology companies are planning for this year, around $650 billion. That’s a 60% increase over last year. And the battles between these companies used to be over software or engineering talent, and now they’re duking it out for land, electricians and power equipment. And at the start of the AI race, investors rewarded these companies, Google, Microsoft, Amazon, Meta, for their record spending. But in the wake of their fourth quarter earnings, those four companies lost $950 billion in value last week. Amazon and Microsoft in particular took a big hit. So are we testing the limits of their ability to build infrastructure?

Stephen Lacey: Caroline, let’s just get your reaction to the market signal, what kind of signal is this?

Caroline Golin: Well, I think it’s a recognition across the board that tech companies that are used to tech returns are not the same as infrastructure companies, and that while there used to be, for a very short amount of time, this focus on CapEx versus OpEx, I think, to evaluate a company’s health, now they’re starting to scrutinize a little more, and I think they’re trying to look at earnings per share. I don’t think they have the right metric yet in this space. I think that what’s going on, and you see what’s going on across the board, is that when we used to reward tech companies, who were never meant to be vertically integrated infrastructure companies, for spending more money on a business model that they fell into, that reacted in the market with a significant amount of bloat.

Caroline Golin: And then now, what you see happening is a question of, what’s a better way to do this? Is it public? Is it private capital? Should third party debt be responsible for this? Should it all be in=house? And there’s no real, I think, clarity, I have opinions on what it should be, but I don’t think there’s real clarity on the market about what is the best way for a tech company to transition into a space where they have been this reluctant infrastructure company, and what should we be scrutinizing, CapEx versus OpEx, or earnings per share, or what is the right metric in a long-term planning horizon where no one actually knows what the value is going to be?

Caroline Golin: So it’s clear as mud, but I think what I’m seeing, and I think what I think, that’s a great phrase, I think what we need to be seeing is stronger capital partners with sophistication for the tech companies to take the burden off of them becoming these infrastructure companies, which they were never designed to be, ever, ever, ever designed to be. But they fell into it by need because there wasn’t one.

Stephen Lacey: Yeah. I want to dig into that a little bit more, Caroline, like how should they finance this infrastructure development? But first to you, Jigar, what do you think is throwing off investors? Do you agree with Caroline’s take?

Jigar Shah: Yeah. I mean, I think that, frankly, the amount of hype that’s been allowed to run amuck without any clarity is pretty shocking to me. I mean, I think Amy Hood, the CFO of Microsoft, said, “We’re just short compute” in her quarterly call. She was like, “We could sell all the compute we have and we need more compute.” I think they’re deliberately fanning the flame of the hype cycle in a way that I think is not helpful.

Jigar Shah: For instance, if you look at how many gigawatts of data centers we need to meet the training requirements for these large language models, it’s not that large. It’s maybe 25,000 megawatts, and I think we’re fine building 25,000 megawatts of thousand-megawatt data centers. Then, when you move to inference, you can do all sorts of things. You could build five megawatt data centers, like Nvidia just rolled out with folks. You could build 100 kilowatt data centers with telecom towers, like Available Infrastructure just rolled out. I think Apple is going to create something where every one of your computers can actually just do inference and make you a couple of hundred dollars a year on the side by using your laptop to do that.

Jigar Shah: So I think that these models are changing so radically, and I find the lack of clarity from the hyperscalers are feeding this frenzy on purpose in ways that are shocking, in my opinion. For them to allow all of this misinformation to go around consistently, such that people have no idea whether we need 200 gigawatts of compute from crazy speculative powered land people or whether we need 25,000 megawatts of compute, is nuts.

Stephen Lacey: Well, I think they genuinely believe that they need unlimited compute.

Jigar Shah: But then they’re stupid, and then their stock price should go down if they’re stupid.

Stephen Lacey: I think they’re just doing what they have to do because there’s a commercial race behind it. Caroline, do you think that they’re fanning the flames in any way, or do they just fundamentally believe that they have to make these investments and they don’t even really know what the cap is because the race is so fierce?

Caroline Golin: I think they don’t know what the cap is because the race is so fierce. I think that they are agnostic to the cap, because this race will trigger another race and will trigger another race. I mean, we’ve unlocked this never-ending feeding story, and I don’t know where all the infrastructure goes, to Jigar’s point, I think it will manifest in many different build scenarios. But I don’t think that they are artificially inflating or deflating the amount of compute that is necessary.

Caroline Golin: I think what they are victims to is a misunderstanding of the supply chain and the restraints on infrastructure and build before they started this. I think that many of the leaders and the physicists and the engineers and the PhDs across all of these companies started this race without a clear understanding of what it took to build the infrastructure to win this race.

Caroline Golin: And so, sky was the limit and it was more of a physics question for them, and that is where I think you see a breakdown. I mean, I felt that breakdown within Google, which was there just wasn’t a clear understanding of how quickly we could get five, 10, 15 gigawatts, that didn’t translate. If anything, if anything, it would be in the tech company’s best interest right now to be very prescriptive and say it’s going to need less, because if it’s going to need less, then they’re going to be more efficient or they’re going to win the race quicker. But I think the fact that they’re leaving it open is because they believe one race will lead to the other.

Stephen Lacey: Which is why we’re talking about space-based data centers later.

Caroline Golin: Yes, of course, space-based data centers. To be honest, at Google, we were talking about five years ago.

Stephen Lacey: Yeah.

Jigar Shah: And I thought Elon said this. I mean, I thought Elon, in his podcast, said that all these people, who are basically software people, are suddenly learning about how hard it is to be a hardware person.

Stephen Lacey: For sure.

Caroline Golin: Absolutely, yeah.

Jigar Shah: And they were just making stuff up all day. And so, as a result, Microsoft still can’t build their data center in Wisconsin because everyone in Wisconsin hates them. They’re the ones who came out and said that they were going to be nice people and not use water and da, da, da, whatever else, and they still can’t build this stuff. I think it is in their best interest to stop basically fanning the flames of bullshit and actually start really being very specific around where they need a thousand megawatts of compute in one place and where they actually can do with five megawatt, 10 megawatt distributed compute, et cetera. And my sense is the reason they’re not answering the question is they haven’t spent five seconds thinking about it at the CEO and CFO level. They’re just like, “More compute, more compute.”

Caroline Golin: Well, I think chip technology is rapidly deconstructing and reconstructing itself, and the efficiency of chips is rapidly deconstructing and reconstructing itself. So on some level, you have a question of, to tell you how much compute you’re going to need, you need to have a line of sight to the efficiency of your chips, the volume of your chips, and there isn’t clarity around that either. I think anyone who is trying to say chip efficiency today or the energy draw of a token versus what it’s going to be in five years, there’s a still 30%, 40% swing there in the guess of efficiency. So you’re ballparking one for the other. And at the end of the day, this comes down to chips and the efficiency of that compute.

Caroline Golin: Now, I think we should go back though, because I think it’s an important point that the audience should recognize, which is that a decade ago, these tech companies had a handful of data centers. The only company of the large ones, and I don’t know Oracle’s business model, Jigar, so apologies, I’m going to include them in this one.

Jigar Shah: We can’t leave them out forever.

Caroline Golin: If someone from Oracle could please contact me and explain to me everything I’m saying wrong about them, I would appreciate that. But Amazon was the only one that was an actual logistics company. None of the rest of us had any experience in infrastructure or logistics. And so, we built all of our data centers as Snowflakes, every single one of them, and we didn’t build a ton. And the problem was also we had enough capitals to self-finance, and there was some third party capital infusions when you look at land or you look at some shells, but generally speaking, we were like a married couple going out and buying and flipping houses, where we were doing all the renovations ourself, and we were hiring smart people, we were hiring a ton of real estate folks, but there wasn’t a demonstrative amount of capital partnering with these tech companies or discipline partnering with these tech companies in the way that they went to market, the way that they built shells, and the way that they powered them.

Caroline Golin: And so, what you see happening now is that the third party space is going to have to enforce discipline on the tech companies, and you see that happening. And what’s interesting is the third party capital that’s coming to the table looks very similar to the third party capital we’ve used for a hundred years in the way we build ports and the way we build bridges and the way we build major hospital centers. And that model is now being transposed onto the tech industry, basically saying, “If you’re going to need $30, $40 billion to build a campus, then you’re going to need to raise that debt, and if you’re going to do that, it’s going to have to look this way.” And that’s new, and I think it’s really important that everyone appreciate that that is new for the tech industry.

Caroline Golin: So we got away with building 50, 60, 70 data centers, wildly using our own capital, every single build’s a snowflake, there was no external discipline, and there was no penalty for it in the market for doing that. Whether there should have been a penalty, I think, is a debate. But now what’s happening is that that amount of capital being used in-house to build this infrastructure, it’s not a wise way to do this, and they don’t have enough sophisticated parties out there that can actually understand power delivery, actually understand zoning and permitting and water and workforce and shell development and manufacturing and supply chain and everything to do it. So you’ve had this runaway capital spend with highly undisciplined infrastructure players who have an internal cost of capital of like 18%, 19% because they’re tech companies, and now you need to fit them into a model of external discipline.

Jigar Shah: It sounds like you’re agreeing with me, that they are literally stupid. I mean, the fact that the Trump administration from the White House’s podium is now saying that we need to impose ESG standards on the tech industry, they came from the White House and said, “We need to help you with water consumption.”

Stephen Lacey: Well, you’re going to have to step back and explain what you’re talking about here.

Jigar Shah: The White House tweeted out basically that we are imposing all of these standards now on the tech industry to make sure that they’re using energy efficiently, they’re using water efficiently, that they’re doing community engagement efficiently.

Stephen Lacey: But they’re doing all that stuff already.

Caroline Golin: They’ve been doing all that.

Jigar Shah: I’m just saying, the fact that the White House is now pro-ESG has meant that the CEOs and CFOs have not communicated properly what it is that they’re doing, and it is blowing back onto the White House so badly that they feel like they have to actually be pro-ESG now. This is how we’ve jumped the shark, and I think people are like, “Oh, it’s fine, Jigar.” It’s not fine.

Stephen Lacey: But I think we’re saying two separate things here. I mean, that’s a whole other question about how the tech companies are communicating this. I think what Caroline is asking is, do we have the right third party capital for building out these data centers?

Caroline Golin: Not yet, yeah. And my answer to that is it’s getting there. Actually, it’s an interesting story of what’s old is new. And the players that were able to say, “How did we build really big things quickly in the past? Let’s use that same model.” And the players who are able to do that are raising $20, $30 billion in a month. When you continue to do it the way we’ve been doing it, I think what the valuations are saying, what the market is saying, is, “Hey, we let you get away with this for a while, frankly because we didn’t know where you were going and you had the balance sheet to carry it.

Caroline Golin: But now, we see that there’s a few other players that are developing the balance sheet, developing the sophistication to do it,” and I will say, there aren’t enough players, “Maybe you should think about conforming and applying that type of discipline to the way you build.” I think the players that choose to do that will be rewarded in the market. I think the players that choose to continue to build your own adventure and data centers and everything’s a Snowflake and we’re going to vertically integrate it and do all of it ourselves will bleed capital. It’s just not what they’re designed to do.

Jigar Shah: Right. But I’m just saying that when Amy says, “We’re short compute and I’ll do whatever it takes to get more compute” on a quarterly conference call, she’s feeding the frenzy. When AWS basically says, “We don’t think we can actually do inference in a distributed way and that we need more 1,000-megawatt data centers to do inference,” and then Duke Energy comes out with a 15% rate increase in North Carolina, it blows back on AWS’s stock.

Jigar Shah: I just think in general that these tech companies, time and time again, have shown that they step in it until they hire the right consultants to go, “Oh, maybe we should take a day and actually understand how all of our constituents are hearing what we’re saying, and why don’t you craft a story that actually is truthful in terms of what we’re doing, but actually gives more insight to investors around how we’re doing this in a way that is consistent with communities, consistent with the affordability of electricity”? Consistent with all of these things that you’re saying, Caroline, that private capital is going to impose on the tech companies. Take a beat. That’s all that the stock market is saying. I don’t think they’re anti-AI. I think they’re saying, “Take a beat and figure out your talking points. Why don’t you talk to your communications consultants and come back to us with a coherent story?”

Caroline Golin: Hold on. I don’t think that the tech companies are out there saying, “We don’t want to cover our own costs.”

Jigar Shah: They’re saying, “We want a feeding frenzy.” They are absolutely saying that every single day with everything.

Stephen Lacey: Because they believe it’s a zero-sum game to build super intelligence and they need as much infrastructure and chips as possible.

Jigar Shah: Then they are not listening to their customers and their constituents. Their customers and constituents are saying, “Your zero-sum game feels like trampling on our rights, and we are not for that. That is why we are not going to let you build in Indiana, we are not going to let you build in Wisconsin, we are not going to let you build in all the other states.” We have six states now that are passing laws that are banning data centers at the state level. I don’t think that the hyperscalers are understanding how their callousness around the fact that we all need to win this race is affecting people’s daily lives.

Stephen Lacey: Caroline, I want to let you respond to that, but that does not feel like the market reaction. I think to your point earlier, investors are trying to figure out, are we investing in a software company that is basically like a sophisticated ad platform or are we investing in an infrastructure company?

Caroline Golin: Well, I think it’s a shove, a gentle shove, to say, “Do what you’re good at, and let’s actually let other market players take this lot.” But I think what Jigar is trying to say is that there’s a third pillar, fourth, fifth, I don’t know how many pillars we’ve named here, the political pillar, which cannot be overlooked, because the quickest way to stop the progression of AI is through the regulation of its infrastructure. If you can’t train because you can’t build, you can’t win the race. And so, the question then comes down is, how do you build this infrastructure in a way that is politically appeasing? And I’m not sure that anyone has the answer to that.

Caroline Golin: I think that most of the tech companies, not all the tech companies, and that only accounts for about 60% of the market, and even that isn’t all the vertically integrated hyperscaler-facing infrastructure development, but those that are hyperscaler-facing, they’re not going into communities saying, “We want to destroy your community.” I think what has happened, and Jigar, you’ve even made this point before, is that no one appreciated the volume of this, and so we just treated it like we treated any other manufacturing company, and we never put in place the fully legal, fully historic precedent, regulatory, jurisdictional rules to say, “You are going to cover your own costs.” There isn’t a single state in the country that needs to pass legislation to allow its utilities or its munis or its co-ops to ring fence data center costs and attribute them. You don’t need legislation to do that.

Jigar Shah: No, and I get that. I’m just saying that I think to say that this entire stock market correction is only about, “We don’t want you putting all this infrastructure on your balance sheet, figure out a way to use third party capital and figure out how to optimize your balance sheet” is misguided. All of these analysts are hearing about all of this pushback and they are reading the Trump Truth Social posts, they’re reading all of the stuff that’s coming on the other side, which you just said, this is another pillar, the political pillar. And I just think that the tech companies, time and time again, wait until everything becomes so loud and that the pushback is so heavy that then they’re like, “Oh, I guess maybe we should hire consultants and figure this stuff out and do this properly.”

Caroline Golin: Yeah. I mean, historically, their approach to anything has been to stay as much out of politics as they can. That is institutional and cultural.

Jigar Shah: Totally. But I think that is part of their stock market correction and I think it will be in the future. I think if people believe that MAGA and Bernie are getting together for a marriage against the tech companies, then the stock price will go down even further.

Caroline Golin: I think where these two are in conflict though is that if it is Google, if it is Microsoft, if it is Amazon, and there is the Google shingle from the beginning, so that is the vertically integrated approach, it is the Google team going out to meet with the county commissioners, to meet with the utilities, there is more accountability there than if it’s a third party. And so, where these two are in conflict for me is that Google has a brand reputation that it needs to maintain and that it can suffer in its other products and its other areas if it doesn’t maintain that brand recognition, same for Microsoft, same for Meta, same for AWS, same for Oracle.

Jigar Shah: It doesn’t seem like it’s actually the same for Meta. They don’t care about their brand at all.

Caroline Golin: Yeah, maybe not. But that exists. If I’m going representing Google, I’m representing big Google, I’m representing big Microsoft, that carries weight. But if you’re a third party and you’re going in there, you’re at the next sales cycle potentially. So where these two issues are in conflict is that, on the one hand, the market is saying, “Hey, we’ve been assessing you wrong for a while, we recognize that. We’d like you to be a little more disciplined and push this CapEx spend off your book and work with parties that actually know how to have disciplined infrastructure capital raises and do that in a more accessible manner.” I think part of this is, “How do we actually assess you now?”

Caroline Golin: At the same time to do that, you are also saying, “We are alleviating you from that front of the house responsibility in the political/social warfare of building a data center, and those two things are in conflict. And I think that I don’t know how that will play out, but until we have these capital providers also take on the same posture, political posture and responsibility posture, it’s going to get potentially worse before it gets better.

Jigar Shah: Well, and I think they’re getting tied together. I think that the third party data center companies are not-so-secretly telling local folks that the people who are going to buy our data center, once it becomes fully permitted and powered land, are a bake-off of those four or five players. So now, the fact that these large companies are not imposing their values onto the early part of the funnel also blows back on them.

Caroline Golin: That’s also really hard to do.

Jigar Shah: When you’re worth $5 trillion, figure it out, that’s all I’m saying. Look, I’m not suggesting this is easy, but I am suggesting that when you fan the flames by saying that we can’t find the compute, that is sending a signal to early-stage capital to give to powered land developers, who are then using surreptitiously named LLCs to get access to land and all this other stuff, they’re not as transparent as the hyperscalers are or some of the best developers are. I was talking to Jeff Bladen at Verrus, who is our old friend from Meta, and he shows up as Verrus everywhere he goes.

Caroline Golin: Yep, yep.

Jigar Shah: He does not hide who he is, and Verrus uses the highest possible standards in everything that they do, which I love. But not everyone is them. And then when people are like, “Well, who’s eventually going to pay for all of this stuff?” It’s the hyperscalers. That’s what they lead people to believe, whether it’s true or not.

Stephen Lacey: I want to wrap this segment up, because I want to make sure we have plenty of time to talk about space-based data centers, but one last question, which is, how are the constraints evolving? What are the biggest constraints currently? Is it political pushback? Is it equipment availability? Is it actually getting powered land? How is the set of constraints evolving?

Caroline Golin: The number one constraint is still power.

Stephen Lacey: It is.

Caroline Golin: It’s power. It’s power until 2029, 2030, yeah.

Jigar Shah: Yeah. I mean, I would say that I think that if we spend a little time being thoughtful about transformers, around locations of data centers, around how we want to be seen in front of communities, how we want to reduce electricity rates for consumers, which I think data centers should be able to reduce electricity rates for consumers, my sense is the next 10, 15, 25 gigawatts is going to be pretty straightforward to get done.

Jigar Shah: But right now, you have 300 gigawatts worth of powered land developers running around doing all this stuff and they’re being egged on, at least people believe they’re being egged on, by the statements in quarterly conference calls from these hyperscalers. I just think that we are in a place where people can be more thoughtful and more rational at the CEO and CFO level, and they have not been. I think that the people who work in the infrastructure side of these data center companies are so wonderful and rational and great to interact with at conferences, et cetera, but they’re not the ones on the conference call and they’re not the ones that are insinuating some of these signals that are allowing these powered land people to run amok.

Caroline Golin: The other nuance here and complexity that you could actually build a whole podcast about is just what is the risk profile or risk framework that each of these hyperscalers look through or employ when they’re building infrastructure, and it’s very different. It’s very, very different. Some lean heavier on environmental risk, others lean heavier on political risk, others lean heavier on actual contract risk and the way that they project finance. And there is not a unilateral belief system in place across 60%, 70%, let alone 100% of the market, around what is the greatest risk to build. And so, when you have that in place, you are going to get a hodgepodge of different approaches.

Caroline Golin: And I’ll close this out, while I think power is still the biggest risk, I would say power is the biggest risk, the second-biggest risk is raising capital and doing it in a way where contract terms are consistent and scalable across an entire fleet. And then, political risk is third, and that is the way that most of the hyperscalers are looking at it. And I think what Jigar is saying is that political risk needs to be first, and then that will maybe help alleviate the other two. But I think from the viewpoint of trade-offs, political is still the third risk.

Stephen Lacey: Well, when it gets too tough, you know what we can always do? Put the chips in a rocket ship and send them up into orbit.

Jigar Shah: To the moon, Alice, to the moon.

Stephen Lacey: Jigar, if you’re this fired up today, I’m itching to see what you’re going to say about space-based data centers. So before we talk about this, I actually have a little story to tell. Back in 2007, I was writing a lot of fiction and I wrote a lot of different scripts for my own amusement, they weren’t going anywhere. And one of them, which I never finished, by the way, was about a technology billionaire who was obsessed with going to space. This was back in 2007. And one of his missions went horribly wrong and created so much space junk that it trapped humans on Earth, and the character who was trying to make humans a multi-planetary species had to grapple with the fact that he trapped humanity on Earth forever. Is that scenario remotely-

Caroline Golin: That’s a great story. I mean, maybe you went into the wrong field. That’s a great story.

Stephen Lacey: I have no idea if that space junk scenario is remotely realistic. But with Elon now talking about a hundred gigawatts of data centers in orbit, which would collectively take up five times the size of Manhattan with solar panels, that scenario is starting to feel more real every day. I guess I should dust that script off.

Jigar Shah: You need to go call our friend, Ramez Naam. Isn’t he a science fiction writer?

Stephen Lacey: He’s a science fiction writer, yes.

Jigar Shah: You guys should co-author a story together.

Stephen Lacey: I love it.

Caroline Golin: I’m sure Marvel will take that, like Iron Man Reborn, it’ll be great.

Stephen Lacey: Just replace him with Tony Stark. Anyway, let’s get into Elon’s latest project, data centers in orbit. Over the past few months, we’ve seen serious conversations about putting data centers in space. In November, Google unveiled this idea called Project Suncatcher that detailed putting TPUs in space. And as you said, Caroline, Google had been talking about this for years before. Sam Altman and Jeff Bezos have both entertained the concept. Nvidia invested in a company working on it. And after SpaceX recently acquired xAI, Elon started openly talking about this concept powered by solar panels that he wants to make. So in typical Elon fashion, he claims that we’re not decades away, that they’re actually imminent, and he wants to put 100 gigawatts of capacity in the sky.

Stephen Lacey: So look, in theory, these orbital data centers offer near constant solar, no land constraints, no zoning boards. In practice, it comes with a lot of challenges, cooling, radiation, latency, launch costs. Who wants to take this first? What’s your initial reaction?

Jigar Shah: Well, I can start.

Caroline Golin: All you, Jigar.

Stephen Lacey: Yeah. I can feel something coming from you, Jigar. What do you think?

Jigar Shah: Look, we had a good conversation about geothermal last week, and I like the fact that Elon is constantly pushing something that is interesting and expansive and pushing the boundaries of human endeavor. So whether it’s self-driving cars or whether it’s electric vehicles or everybody’s going to be powered with solar or whatever it is, I like that, and I think retail investors like that, which is why Tesla’s stock price is way higher than Ford and GM’s stock price, even though Tesla’s car sales haven’t gone that well in the last 12 months. And so, I don’t fault Elon for pumping the stock, and now he’s pumping the SpaceX xAI stock that is going to go public and he needs it to go well for himself, so I get it.

Jigar Shah: I think the practical nature of putting data centers in space that have enough compute to matter is going to be difficult. I talked to a good friend of mine, who is putting long-term data storage in space and he’s really doing that, so think files that you don’t need to access every day, but files you need to store for in the future, like 30 years of emails or whatever, and he was like, “It’s actually better because of the sovereign political risk and all that stuff to put stuff up there and people are willing to pay extra for it.” And so, I think that might happen, which I think is great.

Jigar Shah: And so, I like the future endeavors, but some of the story just doesn’t hold together. For instance, you and I both know that if you are going to use solar panels in space, the payload costs are so high, even with the reduction in cost from SpaceX, that you would use advanced materials like gallium arsenide and other stuff. You wouldn’t use traditional crystalline silicon solar panels with 23% conversion efficiency for space. So parts of the story don’t hold together. Do I think that Elon’s really going to build 100 gigawatts of solar manufacturing? Yes, but I think it’s going to be for the United States to use it on residential rooftops or commercial rooftops or utility-scale. So I think that all of that stuff is real, but I don’t… I like believing in the future, and so I like having this conversation about space-based data centers. I just know from everything I’ve read that it’s highly impractical and decades away.

Stephen Lacey: Well, if I ever need to hide the documentation on my offshore bank accounts, I’ll talk to your guy who’s putting data storage in space.

Jigar Shah: It’s like a 10-day process though to retrieve the information, so you’re going to have to be patient.

Stephen Lacey: Caroline, what do you think of the concept?

Caroline Golin: As a concept, if we’re just talking about physics, it sort of makes sense. I think it will happen, I think it will happen, because there’s enough capital out there to try and to see. I think the real motivation is less about harnessing the sun’s power and avoiding the power crunch on Earth and more about getting out of the politics of completely transitioning this economy. And I think, and I said this before in the podcast, the quickest way to regulate the AI rise is through the regulation of its infrastructure, and it would be unclear how you would regulate the infrastructure in space and who would have control over what processing was done in space and what it was used to do.

Caroline Golin: And so, for me, I think the push is more a desire to leave the humanness of the economy and politics to be purely based on the rise of machines and artificial intelligence becoming the dominant force in economy. That freaks me out a little bit, to be honest.

Jigar Shah: Makes you want to sell the stock. It makes you want to short the stock. I mean, what the hell, man? How is that your feeling in life, is that you just want to be more libertarian and you’re like, “If I could avoid all human decision-making and governance processes…”

Stephen Lacey: A lot of people feel that way.

Caroline Golin: Well, going to back to what we said before was the tech companies historically have always tried to say, “We’re not going to break the law, but we’re also not going to be demonstratively engaged with it.” And the law has always been behind the tech industry, and the only way we’ve created the laws is by in post deciding what the tech industry did was illegal and then creating regulation for that. There was never a bar by which they moved up to, they just kept going and then we imposed the bar in hindsight. And I think that there are a lot of these leaders, and Elon is probably the most outspoken one, believes that that bar shouldn’t exist, that that’s the folly of humans is that they are political. And so, I think that that is the main driver.

Caroline Golin: When we first heard the pitch at Google, and this was almost five years ago, I did chuckle, but I wasn’t supposed to chuckle in the meeting. In fact, they were quite serious. And so, I think if you’re a company that has made its reputation and built its wealth on by doing the impossible, you’re going to try. So I do think it’ll happen, I do think companies will try. I just think the ultimate motivation has less to do with we can’t get enough power in the US and more to do with the fact that we’d like to get out of the humanity of politics in the way we change the economy. And I would argue that the one thing we aren’t really talking about is the materials that would be needed to do that and how you’re going to extract and refine and process those materials. That, to me, is the sleeping Achilles heel that no one, including us, are talking about enough, I think.

Stephen Lacey: I think the desire to remove data centers from the complicated decision-making on Earth and the belief that we need unlimited amounts of compute will inevitably push some of these data centers into space. Obviously, there’s higher latency in space, so we’re talking about batch processing and model training, not inference or other consumer applications, but certainly a lot of compute necessary for those applications. And then, there’s obviously just constraints in space. Space is a vacuum, so you don’t have traditional air or water cooling. The numbers are-

Caroline Golin: You don’t need cooling, that’s the thing, you don’t need cooling at the same level, right?

Stephen Lacey: But you need some cooling, right?

Caroline Golin: I mean, you need temperature control, but you don’t need cooling in the same way we need it when we’re building our data centers in Georgia.

Stephen Lacey: I mean, we’re talking about Elon, I think when he was on Dwarkesh Patel’s podcast talking about the 100 gigawatts of space data centers, he’s talking about thousands and thousands of rocket launches a year.

Caroline Golin: That’s a good business model for SpaceX.

Jigar Shah: I was going to say, for sure. It is a really good thing to be hyping right before the IPO.

Stephen Lacey: Yes, definitely, and if there’s any company that has a leg up on this, it would be SpaceX, of course. And then, there’s a big question about repair. I mean, Elon says we’ll have these modular services from robotic repair craft, but I don’t know, I haven’t seen anyone who really believes strongly that the repair services are cost-effective. And the big question is, if these chips last two years, how much repairing are you going to need to do? They burn through these chips so fast. I think it will happen.

Jigar Shah: I mean, it’d be great to have a robotic company that just fixes Samsung ice makers in refrigerators. I mean, that’s like $850 every time that breaks. You can make a lot of money just doing that.

Stephen Lacey: But I’m with you, Caroline, I think this will happen in some limited fashion, and I think that the insatiable demand for compute will get us there.

Jigar Shah: Yeah, no. I completely agree with Caroline that we will have somebody test the theory and put some compute in space, I’m not disagreeing with that. The notion that we’re going to have anything that resembles a critical mass of compute in space over the next, let’s call it, 15 years, is nuts. I mean, I don’t think that you will have any sort of real sizable compute in space within the next 15 years.

Caroline Golin: I mean, I think the macro question though we have to ask here is, if you’re going to space, it’s because, as Elon alluded to, you’re pursuing hundreds of gigawatts of compute on an annual basis. Why? That’s the macro question, why? And I think the motivation is that we don’t have to answer the question of why, which goes back to Jigar’s point, which is what’s happening politically is people are asking why, why do we need all this? There’s clear demand for some of the cloud-based options right now, but I think the macro question is, why do you need to do that? And if that question doesn’t get answered, I think it will happen.

Jigar Shah: You can pay me to do it. I mean, that’s what’s so crazy.

Caroline Golin: Yes. Of all the people, I will pay you, Jigar.

Stephen Lacey: We’ll send you up there.

Jigar Shah: I’m just saying, you could put a 200-kilowatt data center in my garage, I’ve got a 400-amp service from my utility, you could do that next week and I’ll rent you that compute next week.

Caroline Golin: But then, there’s private property rules.

Jigar Shah: But this is what I’m saying, is that instead of working with me to pay me or pay an apartment building or pay this group or that group, they’d rather go to space, which is what Caroline-

Caroline Golin: Because the vision around this is not human, that’s why. They’d rather not deal with-

Jigar Shah: I hear you, but I’m just saying, you’d rather spend a trillion dollars over here than help somebody make money on Earth, you’re like, “What is wrong with you?” And how much depravity do you have to be involved in to not want to help people on Earth?

Stephen Lacey: Well, you’ve just hit on the existential question around the race to space. Let’s talk about solar to wrap up the show, and we’ll look at another Elon project that is a little less ambitious down here on Earth, but it is still very, very ambitious. Elon recently launched, or Tesla launched, a new solar panel and racking system, and they announced their intentions to build a 100-gigawatt solar manufacturing line in the US. It’s a 420-watt panel. It’s got a mounting system that, according to Tesla, will cut installation time by 30%. Obviously, big claim in an industry where soft costs and labor constraints are the bigger bottlenecks than panels themselves. Check out our story from Maeve Allsup on Latitude Media, she’s got a great breakdown of the move and how it fits into Tesla’s history and master plan.

Stephen Lacey: Jigar, the solar manufacturing business is absolutely brutal. Tesla failed to make a solar roof product scalable. What do you think of Elon’s second go-around?

Jigar Shah: So I think the people who are running the energy division at Tesla have been crushing it. I think going from Megapack to Megablock has been a huge success story, where they have the transformer built into the Megapack. I think they’ve just released the Tesla Powerwall three-phase design, they’re starting in Germany. So I think when you think about the quality of the products coming out of the Tesla energy division, it is really breaking numbers every quarter, their sales are way up. So I think that team is really doing a great job.

Jigar Shah: That team put out a paper last August about what they thought would be required to get down to $2 a watt solar, and I think they’re executing on that plan, so that they had some hardware cost efficiencies, which is what this is a part of, they had some installation cost efficiencies. And then, there’s also the permit power stuff that Nick Josefowitz and others are doing around figuring out how we change laws locally to avoid permits altogether. That law passed in New Jersey, that law passed in Texas, Florida has some version of it, other states are trying to replicate it. And so, I think we are headed in that direction.

Jigar Shah: The one thing I find fascinating is that Tesla is the only company in our entire industry with a brand. No one’s ever heard of Sunrun or GoodLeap. If you just pulled a hundred voters, no one’s heard of any of the brands in the solar industry, so we clearly don’t spend money on marketing. And Tesla doesn’t either, famously, but people know who Tesla is. And so, the key to this whole thing is not making a solar panel that can be installed 30% faster, which is valuable, getting the cost of solar down is valuable, it’s how do we get away from the market push? Our industry doesn’t have any market pull. People don’t just call up some website and order solar. The only way we sell solar is some guy annoys you to death at your door and gets paid $5,000 to $10,000 in commissions to convince you to do solar.

Jigar Shah: We need a real marketing campaign, something that costs $100 million at least nationwide that says, “Your rates are now double what they were 10 years ago. You need to buy solar. That’s the only chance you have. You’re not going to put a fuel cell in your basement. You’re not going to run your house on a backup generator. So the only thing you can do to save money right now is solar on your roof and battery in your garage.” We need market pull so that one third of American households actually want to buy solar so that we can actually get to Australia. Building a hundred gigawatts of solar manufacturing in the United States doesn’t get you market pull. So while I love the fact that they’re doing it, totally agree with Elon that we should be building big things in this country again, I want to know what their marketing plan is. Are they going to spend $100 million creating market pull?

Caroline Golin: Well, I think he’s creating the market pull by saying we’re going to max out all the power and all the land-based data centers, which is going to, if we don’t do this correctly, create disruption for the residential customer, and that’s the incentive to put it on their roof.

Stephen Lacey: I mean, didn’t we see Tesla’s role in batteries creating that market pull? I mean, they needed a way to scale up batteries for electric vehicles to drop the costs and scale manufacturing. They created the Powerwall, and the Powerwall was a desirable product and a lot of consumers actually wanted it. And so, by their nature, they created a desirable product. Couldn’t they do the same thing in solar, where people finally say, “Oh wow, a Tesla solar panel, I want that”?

Caroline Golin: Well, it’s customer acquisition. So they already have two, three lines into the residential customer right now, and it’s a lot easier to purchase from the same vendor when you’ve already got the car and the battery. I mean, I think that’s what they’re banking on, right?

Jigar Shah: But they’re not doing it, that’s what I’m saying. All of their sales of Tesla cars do just go through Tesla.com. Almost none of their sales of Tesla Powerwalls or Tesla solar panels go through Tesla.com. The vast majority of their sales are B2B, through a local solar installer that annoys you by having somebody knock on your door and sell you a system, and then they sell you a Powerwall because it’s the best thing to install and it’s easier to configure and work out or whatever. But they’re not actually selling most of their solar through Tesla.com.

Stephen Lacey: Well, we’re also talking about Tesla as if this were five years ago too. I mean, Elon did a huge amount of brand damage with his work with DOGE and the administration and how he communicates on X and-

Caroline Golin: And the Cybertruck.

Stephen Lacey: And the Cybertruck. And there’s a real question about consumers who are likely to purchase or invest in solar, whether they want a Tesla product. My belief is that the brand damage is probably strong enough that it will hurt Tesla’s ability to create market pull through the brand.

Jigar Shah: I don’t know. I mean, I think we’re at like, what, 6% of rooftops have solar on them right now? I mean, we’re selling something on the order of 500,000 solar systems a year. My sense is that he could find 500,000 people who are MAGA-affiliated that wants to put solar on their roof. To me-

Caroline Golin: Yeah. I think I quibble making the jump that the only constituency, and I use that word specifically, that wants to have energy independence or be disassociated from rising costs because of AI are environmentalists. I don’t think that’s the case at all.

Stephen Lacey: And that brings us perfectly into the second part of this conversation, which I think is extremely important. We just saw this poll released from the chief pollster for Donald Trump’s campaigns that found a majority of GOP-aligned voters actually support utility-scale and rooftop solar. About 51% said they’re in favor. That jumps to 70% if the panels are made in the United States and not tied to China. And nearly two-thirds agreed that we need all forms of electricity generation, including solar, to lower power costs.

Stephen Lacey: This is not much of a surprise to anyone who’s followed this polling over the years. I was writing about this 15 years ago, I was writing about it during Trump’s first term, poll after poll after poll has showed this. What is most surprising perhaps is that the polling has stayed remarkably stable as partisanship has infiltrated everything and as the President himself has tried to paint the technology as woke. And so, it’s showing up in every single poll, national and local, from a variety of reputable pollsters. And then, we, of course, had Katie Miller, Stephen Miller’s wife, who was out on X tweeting about the poll and saying how solar is the future. I wonder, with Elon making this big investment, with still strong support, in spite of the President’s rhetorical attacks, what signal is this sending, Jigar, about how solar is going to fare politically in this country?

Jigar Shah: Yeah. I mean, I think you and I both know that solar is inevitable. Elon suggests that as well. We certainly have great stuff from Ember and other people talking about that as well. And so, solar is going to keep going up and to the right. But the big thing, I think, that people don’t understand is that there is no competition to solar. That’s something that people don’t really understand. If you’re running a Walmart store or you’re having a residential home and you just got rates that went up by 20%, what else are you going to do? You can do energy efficiency. You can maybe put a Bloom Box in your Walmart store, maybe.

Jigar Shah: I just think that in general, the only thing that you can do to actually get your own costs under control and to tell the utility that you don’t completely own me is to put solar panels on the roof and a battery in your garage. And so, there are a lot of people who are motivated by giving the utility the middle finger. There’s a lot of people who are motivated by independence. I mean, that’s how we won Georgia. That’s how Buddy McDonald was the vote in 2013 or whatever, when he said solar’s got to be allowed in Georgia, is because there were a lot of right wing conservative talk radio in Georgia that was like, “We want solar.”

Caroline Golin: It’s Bubba, not Buddy.

Jigar Shah: Bubba, sorry. Should I say that again?

Caroline Golin: No, it’s better that I correct you.

Stephen Lacey: Do you guys think the Trump administration will come around on solar, do you think the attacks will stop, seeing this polling and then facing real grid constraints as they get worse over the next few years?

Caroline Golin: Well, I hope so. Honestly, I think most utility CEOs out there have been pretty vocal about saying we need an all-of-the-above strategy, that it’s a very poor and shortsighted investment play to stack your portfolio with one or two technologies, and I think there are a few utilities in the country who’ve done that in the past and seen that tumble. And so, I think it’s a matter of whether you think Trump will respond to populace. I think he’s shown in the past, he will. And the other question is whether or not his friends or his constituents or who he wants to see succeed also will, and I think that that’s turning as well. I mean, it’s so strange to me that we ever attached moral authority to one electron versus the other, but I’m hopeful that the administration will ease up, because we have a lot of power that could be on the system right now if we could just get past this name-calling that we’re doing in the country.

Jigar Shah: I think you need to actually force Trump to unleash the permits from interior. I think that in general, I’ve been very disappointed by the tact that’s been taken by ACP and others around appeasement. And I understand why, they’ve got a lot of European companies who are on their board, and then you’ve got folks-

Stephen Lacey: And what is that strategy? What have they been doing?

Jigar Shah: It’s all trying to figure out how to convince people in backroom deals, figuring out how to get down to Mar-a-Lago to talk to people. I think that in general, there needs to be a narrative that takes hold that the President is deliberately raising everybody’s bills, that the President, on the one hand, is saying he wants a 50% reduction in people’s energy bills, as per the executive order he signed on day one, and then he’s taking every effort he can possibly do to raise people’s bills, and he needs to get tagged with that. And then, when the midterm elections occur, people need to say, on Fox News, the next day, “The reason we lost all these seats is because we are perceived as the party who wants higher bills.” And that, by the way, is already coming out in the polling. For the first time in 35 years, Democrats are being viewed as more reliable on keeping energy costs down than Republicans, the first time in 35 years.

Jigar Shah: But ACP is not reinforcing that message. The way this stuff works is you don’t just win, you have to take credit for the win and you have to pound him into the ground and you have to actually be feared by the other side. And I think we just, for a long time, have been like, “If we just stay quiet and we stay in the shadows, they will just do the right thing,” and I think we’re just too big for that strategy.

Stephen Lacey: Yeah. We’ve talked a lot about this. I don’t know. I can see why they would want to play the President’s game. He’s a deal-maker, you go and you cut deals, you keep your head down, you try to make him say a positive thing, and then let the advocates do the attacking for you or let the Democrats do the attacking for you. Simultaneously, the industry wants to be taken seriously, but they desperately want to seem nonpartisan, and the industry is generally nonpartisan.

Jigar Shah: I mean, I think API plays it much better. API says that they’re nonpartisan. They’ve got 10 Democrat congressmen and senators that kill anything that would regulate them in the House and the Senate. And during the four years of the Biden administration, their message was Biden administration has a war on oil and gas. Even though we had record amounts of oil and gas production, we had record amounts of LNG exports, we saved Ukraine with LNG exports, the Biden administration helped them get all those long-term contracts from Germany, et cetera. So it’s not like… I’m just saying that you want to be feared, that’s how you gain power, that’s how you control your destiny, and we are not doing that. We are basically saying, “The facts are on our side,” and that is not how you build power.

Stephen Lacey: Well, you are nothing if not consistent, Jigar. That marks the end of the show. Caroline Golin is the CEO of Envision Energy Advisors. Great to see you again.

Caroline Golin: Thanks for having me.

Stephen Lacey: Jigar Shah is the co-managing partner of Multiplier. Good to see you.

Jigar Shah: Always.

Stephen Lacey: When I have some data storage needs to send to space, I’ll reach out to put me in touch with your friend.

Caroline Golin: I’m telling you, you probably could feed your script into one of these AI platforms and generate the next Super Bowl commercial.

Stephen Lacey: Oh man. All right. That’s coming up on the next Open Circuit. Thanks everyone for being here. We are produced by Latitude Media. I am the co-founder and executive editor of Latitude Media. Jigar Shah and Caroline Golin are my co-hosts. Anne Bailey and Sean Marquand help us produce the show. Subscribe to us on YouTube. You can find our full video episodes and clips on YouTube, that’s the Latitude Media channel. And we will have episodes of Catalyst out soon there as well. I know a bunch of listeners to this show follow Catalyst. And of course, you can find transcripts and pod episodes at Latitude Media, plus stories to all the subjects that we cover. Thank you so much. We will catch you next week.

The post Are investors losing faith in the AI infrastructure frenzy? appeared first on Latitude Media.

via Latitude Media https://ift.tt/LnrN1SP