Inside OpenAI's Stargate Megafactory with Sam Altman | The Circuit

Speaker 0:00 [upbeat music] Stargate, put that name down in your books, 'cause I think you're gonna hear a lot about it in the future. A new American company that will invest $500 billion at least in AI infrastructure.

Sam Altman 0:16 The data centers are actually under construction. The first of them are un- under construction in Texas, the Abilene location, which is our first location.

Speaker 0:24 Abilene is a town on the western central plains of Texas. We kinda scratch and fight for everything good that comes our way.

Speaker 0:33 I had people reaching out, and they said, "Well, is the president gonna come see you?" And I said, "Well, he hasn't texted me yet, so..."

Emily Chang 0:38 [laughs]

Sam Altman 0:39 I'm thrilled we get to do this in the United States of America. I think this will be the most important project of this era.

Emily Chang 0:45 [upbeat music] Welcome to Abilene, Texas. This is the site of Project Stargate, a very mysterious, much talked about, much speculated about project that brings together OpenAI, SoftBank, Oracle. If it all comes together, building after building on this red dirt behind me will be filled with chips that'll power the AI revolution.

Speaker 1:10 What we're undergoing right now is sort of the largest infrastructure build in human history. It's gonna create new ways of doing business, it's gonna create new ways of invention, and it's gonna really drive humanity forward.

Speaker 1:20 I was in the real estate business. These buildings, these are [laughs] big, beautiful buildings. They're gonna employ a lot of people.

Emily Chang 1:26 How do you personally grapple with the pace of it all?

Sam Altman 1:30 It does feel like it's going very fast. Certainly think if I could, like, transport myself back three years ago, it would seem like unimaginable progress.

Masayoshi Son 1:39 This is amazing, explosive growth of intelligence. That's coming in our decade. Everything will change.

Emily Chang 1:48 When Stargate was first announced, there was a lot of skepticism. A lot of people thought it's just too big to be true. So this is your very first look. [upbeat music]

Emily Chang 2:04 Hello.

Speaker 2:06 Hey, Emily.

Emily Chang 2:06 Thank you so much for having us.

Speaker 2:08 Welcome to Abilene.

Emily Chang 2:09 It's so awesome to be here.

Speaker 2:10 Yeah.

Emily Chang 2:11 It's all happening.

Speaker 2:12 I know.

Emily Chang 2:13 How many people are on this site?

Speaker 2:15 Uh, today we have, uh, about 2,200 people on site.

Emily Chang 2:18 [upbeat music] The guy at the heart of the Stargate build in Abilene is Chase Lochmiller. He's the founder of Crusoe, the little-known data center startup overseeing this massive project.

Speaker 2:30 We started construction here, uh, last June, June of 2024. This was just a com- a complete barren field. There was-

Emily Chang 2:37 So there was nothing here in June?

Speaker 2:38 There were a bunch of mesquite trees and bushes and brush.

Emily Chang 2:42 Stargate is so mysterious. What actually is happening here?

Speaker 2:46 [laughs] What will eventually be here is eight separate buildings-

Emily Chang 2:50 Mm-hmm

Speaker 2:50 ... um, all containing very large clusters of, uh, fully interconnected GPUs.

Emily Chang 2:55 Yep.

Speaker 2:55 Um, we're gonna do a tour of some of that. This is gonna get us around the whole 1,200 acre campus.

Emily Chang 3:01 Hey, we're going on an adventure.

Speaker 3:03 It's like getting around Jurassic Park.

Emily Chang 3:04 [laughs] I love that.

Speaker 3:08 All right, let's go off-roading.

Speaker 3:10 Yeah.

Emily Chang 3:10 I feel like we're going on a data center safari.

Speaker 3:12 It is much like a data center safari.

Emily Chang 3:14 [laughs] [upbeat music] Crusoe's eight planned buildings will hold up to 400,000 chips, which would make this one of the world's largest known compute clusters. The partners behind Stargate, OpenAI, Oracle, and SoftBank, have committed $100 billion to the project so far, and claim they're prepared to put in five times that as they expand to multiple sites. This first site in Abilene has been dubbed Project Ludicrous, as in...

Sam Altman 3:41 Ludicrous speed. [gasps] Ludicrous speed.

Emily Chang 3:45 Which is appropriate. The whole thing is supposed to be done by mid 2026. Are you round the clock-

Speaker 3:50 Yeah

Emily Chang 3:50 ... 24 hours a day?

Speaker 3:51 We're, we're, we're round the clock. It is very much a 24/7 operation. We're trying to deliver on, you know, the fastest schedule that a 100 megawatt or greater data center's ever been built.

Emily Chang 4:00 [upbeat music] So why

Emily Chang 4:03 the ludicrous rush? Well, because Stargate is part of a much larger AI race. [upbeat music] Companies like Microsoft, Meta, Google, Amazon, and xAI are all building AI data centers of their own at sites across the US and beyond. Each company wants to have the most capable AI being used by the most people, and they've committed tens of billions of dollars to make it happen.

Speaker 4:29 There's no reason we won't scale up our computing 100,000 times in a few years.

Speaker 4:33 I'm good for my 80 billion.

Speaker 4:34 Right.

Speaker 4:34 [laughs] I'm going to spend $80 billion building out Azure.

Sam Altman 4:37 The downside of being behind is that you're out of position for, like, the most important technology for the next 10 to 15 years.

Emily Chang 4:44 But beyond just making better AI, many tech leaders want to be the first to create AGI, artificial general intelligence, a super smart agent that can think and learn just like a human, if not better. Large language models like ChatGPT have gotten more capable as the technology's evolved, so the hope is that adding an astronomical amount of compute will make them even more capable. Take me

Emily Chang 5:09 back to the beginning before Stargate. What did you start seeing that made you realize we need more compute, more power, more connectivity-

Sam Altman 5:16 Yeah

Emily Chang 5:16 ... that you were hitting a limit and needed to scale?

Sam Altman 5:19 We used to think a lot about the compute we would need to train the models, and what we didn't used to think about as much is how much people were gonna use these models.

Emily Chang 5:28 Mm-hmm.

Sam Altman 5:28 I mean, we did think about it. It just turned out people want to use the models much more than we imagined. And a couple of years ago, maybe after the launch of GPT-4 and ChatGPT, it really started to hit, like, "Oh, man, this is, this is, like, gone from a lot of compute to, like, the biggest infrastructure project in history." And out of that emerged Stargate.

Emily Chang 5:47 Why call it Stargate?

Sam Altman 5:49 It began as a code name, and sometimes code names stick.

Emily Chang 5:52 But it means something, right?

Sam Altman 5:53 Oh, um, the design of one of the very early layouts of a data center looked a little bit like Stargate from the-

Emily Chang 6:01 Mm-hmm

Sam Altman 6:01 ... from the show.

Emily Chang 6:02 How did you get to talking with Masayoshi Son and Larry Ellison, and how did you all come together?

Sam Altman 6:08 I did two kinda long trips around the world, and a lot of it was to talk to developers and governments, uh, but a lot of it was also to just really try to like get my head around the supply chain. On one of those trips, I met with Masa,

Sam Altman 6:22 and we got to speaking about, you know, what it would take to do compute at this kind of scale.

Emily Chang 6:26 Hi, Masa. It's so nice to see you. What conversations did you have with Sam that led to Stargate?

Masayoshi Son 6:33 Well, we have a, a shared vision of, uh, AGI coming soon, and we need compute.

Emily Chang 6:42 Of all the bets you could make, why this? Why Stargate?

Masayoshi Son 6:46 Well, I, I believe in the vision that, uh, this AI, the AGI, will change the mankind's life, uh, in every aspect. And we are lucky to see this enormous revolution, and I, I'd like to be part of that, and get excited. Last month, I came to, um, celebrate your winning, and promised that, uh, we would invest $100 billion. And you told me, "Oh, Masa, go for 200." [laughs]

Sam Altman 7:19 I do. [laughs]

Masayoshi Son 7:21 Now I came back with 500.

Emily Chang 7:23 Explain the math to us. How does it all add up to needing $500 billion?

Sam Altman 7:28 Well, that covers the capacity we think we need for the next few years, given our growth projections. The, the interesting question is if we had a... If we knew how to get a trillion dollars right now, which we don't, would we be able to deploy that profitably in the next few years? And I'm not sure about that, but I feel confident we can like make 500 billion of value back.

Emily Chang 7:48 All right, so what are we seeing?

Speaker 7:50 There's many forklifts, there's many cranes, there's many dump trucks, there's many excavators.

Emily Chang 7:55 New creatures emerge at every turn.

Speaker 7:57 [laughs]

Emily Chang 7:58 Are most of the workers from Abilene?

Speaker 8:00 Uh, we're a very large employer here in Abilene, but, uh, we, we have brought in, um, a lot of labor from outside the city and outside the state. We actually have workers from seven different states. I've never built anything [laughs] of this scale. It's one of the, uh, underappreciated aspects of AI, is actually just the physical infrastructure that's needed.

Emily Chang 8:24 This is awesome. I love it.

Speaker 8:26 We're building stuff.

Emily Chang 8:27 Yes.

Speaker 8:29 There's 155 vehicles that are moving dirt, moving earth right now.

Emily Chang 8:34 There's something so satisfying about seeing lots of dirt being moved.

Speaker 8:37 Yeah. Yeah, absolutely. If you look at that crane, like this red crane...

Emily Chang 8:43 Mm-hmm

Speaker 8:43 ... that's a 600-ton crane. Um-

Emily Chang 8:46 I have no... Like, how does that compare to other cranes?

Speaker 8:49 It's... It can lift a lot. It can, it can lift a lot of weight. I mean, 600 ton, that's like 100 elephants, right?

Emily Chang 8:56 This is actually happening around the world. Data centers are being built in Malaysia and Japan and France.

Speaker 9:01 I, I think the demand for compute infrastructure, it's not a US phenomenon, it is a global phenomenon. A lot of countries have actually decided that data centers are, are critical infrastructure that must be accelerated.

Emily Chang 9:14 So if, if hyperscalers are going to Malaysia or Japan, is that competition for you?

Speaker 9:18 Uh, we don't really see it as competition. We, we kinda see it as, uh, you know, just an overall growth of the sector. Uh, we think we have, you know, a lot of unique ways of doing it, and a lot of value to add in the, in the process. But, you know, pretty just excited by everything happening in the space.

Emily Chang 9:33 Lockmiller's path to leading the Stargate build has been circuitous. He's climbed five of the so-called Seven Summits, the tallest mountain on each continent, and worked in physics and finance. Tell me about your journey before Crusoe. You grew up in Denver. You started out using AI as a trader on Wall Street.

Speaker 9:52 Yeah. [laughs] We were big early adopters of deep learning, using these multi-layer neural networks to predict returns for, uh, certain securities. Um, and uh, you know, we found them to be fairly effective. I had high conviction that AI was gonna transform everything. It sort of always felt like this, this meta science.

Emily Chang 10:09 Lockmiller eventually left finance to get into crypto. Not in the get-rich-quick, hyping memecoins way, but in the hundreds of computers solving math problems way. His idea was to build Bitcoin mining operations near oil rigs, taking advantage of a quirk in the oil drilling process.

Speaker 10:27 When an oil company drills for oil, uh, one of the associated byproducts of that is natural gas, and when they don't have access to a pipeline, all of this associated gas is just being burned off on site. And so we had this concept that instead of trying to bring that gas to a market where you can sell it, we could bring a market to the gas. We could build these mobile and modular data centers, and bring them directly to the wellhead, and then use that power to power the data center.

Emily Chang 10:52 It was a good idea with unfortunate timing. The crypto market crashed. You then went from Bitcoin mining to AI.

Speaker 10:59 [laughs]

Emily Chang 11:00 It almost sounds like going from one hype cycle to another.

Speaker 11:02 Yeah. It, it may seem like that from the outside. Uh, but I, I'll say that, uh, you know, from very early on, we, we had planned to build, uh, an AI infrastructure business.

Emily Chang 11:13 So what do Bitcoin mining and AI data centers have in common?

Speaker 11:16 Uh, number one, they, they both consume a tremendous amount of energy. Uh, number two, they're both, uh, more location agnostic than traditional data centers. Northern Virginia is largely perceived to be sort of the center of the world for data centers today.

Emily Chang 11:33 That's data center 1.0, right?

Speaker 11:35 That's data center 1.0.

Emily Chang 11:37 Virginia's data center alley is where the first generation of cloud computing was built, close to the government agencies that created the original internet. Those data centers are full of CPUs, basic computer chips that power everyday things like social media, streaming, and the cloud. Stargate, on the other hand, is part of a new era of data centers, full of GPUs, the special advanced chips needed to power AI and do lots of tasks at once.

Emily Chang 12:05 So this is the inside of a data center?

Speaker 12:07 This is the inside of a data center under construction. [laughs]

Emily Chang 12:09 Okay.

Speaker 12:13 There are going to be...

Emily Chang 12:15 Very active construction.

Speaker 12:16 It is a very active construction site.

Sam Altman 12:18 So you can take them through one of these.

Speaker 12:20 Okay. Yeah, yeah.

Sam Altman 12:20 And then through the hallway.

Speaker 12:22 Uh, okay.

Emily Chang 12:25 People look like they were working very hard and fast. [laughs] There.

Sam Altman 12:29 People never love when I show up for data center visits.

Emily Chang 12:31 [laughs]

Sam Altman 12:32 Uh, no.

Emily Chang 12:33 Why?

Sam Altman 12:34 It's not my area of expertise.

Emily Chang 12:35 Yeah.

Sam Altman 12:35 And yet I have opinions, and so people are like, "Okay, you know?"

Emily Chang 12:37 Yeah.

Sam Altman 12:37 "Leave us alone."

Speaker 12:39 So we are gonna go to one of the GPU data halls, where, uh, they'll... there will be a very large cluster of, uh, Blackwell GPUs.

Emily Chang 12:46 Blackwell is the newest AI-optimized chip from NVIDIA. Once the eight buildings of Project Ludicrous are complete, each of them will hold up to 50,000 chips to be installed by Stargate partner Oracle. Wow.

Speaker 13:01 This is where the GPUs will go.

Emily Chang 13:03 This is where the GPUs are going.

Speaker 13:05 [laughs]

Emily Chang 13:05 I've been waiting for this.

Speaker 13:06 This is where the AI factory will be.

Emily Chang 13:08 All right. This is what an AI factory looks like.

Speaker 13:11 And it truly is like a factory, but the difference is the, the end product that gets manufactured here-

Emily Chang 13:17 Mm-hmm

Speaker 13:17 ... is intelligence.

Emily Chang 13:20 And the sooner this AI factory starts up, the better for

Emily Chang 13:23 OpenAI. In March, the company rolled out a new and improved image generator for ChatGPT, which quickly became almost too popular to handle.

Sam Altman 13:32 I've seen viral moments, but I have never seen anyone have to deal with an influx of usage like this.

Emily Chang 13:36 Mm-hmm. You added a million users in an hour or something like that?

Sam Altman 13:39 Yeah. I mean, more some hours, but yeah. It was, like un- unprecedented, wild.

Emily Chang 13:44 Mm-hmm.

Sam Altman 13:44 And also, like, making an image is not a... it's not exactly like a low compute task the way we-

Emily Chang 13:48 Mm-hmm

Sam Altman 13:48 ... we do it with the new image gen.

Emily Chang 13:50 You tweeted that the GPUs were melting. You tweeted-

Sam Altman 13:53 Yeah, I didn't realize people were gonna take that literally.

Emily Chang 13:55 [laughs] I, I, I mean, I get that it was a joke, but it's a bit-

Sam Altman 13:58 They're running very hot, but, like, the metal is not actually melting.

Emily Chang 14:01 [laughs]

Sam Altman 14:01 So, um, we had to do a lot of very unnatural things. We had to, you know, borrow compute capacity from research. We had to slow down some other features, 'cause it's not like we have, like, hundreds of thousands of GPUs sitting around just, like, spinning idly. If we had more GPUs, we would be more able to handle demand surges like this. Um-

Emily Chang 14:19 So it's that directly linked?

Sam Altman 14:20 Yeah. You know, more compute means we can give you more AI.

Speaker 14:25 So one of the things to just take note of, um, is just the amount of stuff, uh, happening overhead. You can kinda see these very large pipes that distribute cold water to, uh, to the individual chips to cool them down.

Emily Chang 14:40 The water is needed because GPUs run very hot, and keeping them cool typically involves an evaporation process that can consume millions of gallons of water per day. That's raised concerns about the sustainability of this AI boom. With Stargate, though, Crusoe says it's bypassed this problem using what's called a closed-loop system, which keeps the water permanently encased within pipes. You can put a million gallons in, and then you don't need more forever.

Speaker 15:10 Yeah. That's, that's essentially right. We, we basically fill it one time, and then we're done, and that's something we can easily coordinate with the city in terms of, you know, getting a million gallons of water one time. It's a far different ask than, you know, asking for a million gallons an hour or, uh, some, some other crazy request, which could be the case if you had an open-loop, uh, cooling architecture.

Emily Chang 15:31 So once this is all built, what's the grand vision?

Sam Altman 15:35 Give people tools to let them do whatever they're gonna do better. Um, people will use it and unleash their creative energy and make all sorts of stuff that you and I love, or some stuff that we don't, um, or stuff to just entertain themselves. Personally, the area I'm most excited about is AI for scientific discovery. You know, we're not there yet, but we're not far away. I think 2025 will be a world where we s- have agents do a lot of work, but work of the kind of work and things we already know how to do. I'm hopeful that 2026 will be a big year of, like, really, uh, new scientific progress.

Emily Chang 16:07 Hmm. But in order for that to happen, buildings full of chips aren't enough.

Emily Chang 16:14 You also need energy, lots and lots of energy. That could be a problem for the climate and a limiting factor in the growth of AI. Sam Altman's been hunting for more compute for a while now, and it seems like energy is a big constraint. Can you explain that? Like, what are the limits?

Speaker 16:32 Uh, let me sort of put this in perspective for you. If you look at a data center 20 years ago, an individual rack in the data center would typically be budgeted to consume between, like, two and four kilowatts. Today, with the latest and greatest, uh, configuration from NVIDIA, you know, we budget 130 kilowatts per rack. So you're talking about almost two orders of magnitude increase.

Emily Chang 16:53 To put it another way, asking ChatGPT a question uses about 10 times as much energy as a Google search.

Speaker 17:00 It is a complete transformation in terms of the amount of infrastructure required to support this workload, and so that has become the major bottleneck.

Emily Chang 17:10 So AI data centers can't be located just anywhere. You need a location with lots of available power, and West Texas happens to fit the bill.

Speaker 17:19 We came to Abilene, honestly, because it's a community that's very rich in low-cost and clean energy. A lot of wind development has happened in and around the Abilene region.

Emily Chang 17:29 Mm-hmm.

Speaker 17:30 And what they lack is actually demand for power. That created a massive opportunity for AI, which is so hungry for power in this moment. This is the first 200 megawatt substation. Uh, right over here, there's gonna be developed an incremental one gigawatt substation. So the total power capacity for the site will be a total of 1.2 gigawatts.

Emily Chang 17:53 All right, so we're talking bigger than a gigawatt.

Speaker 17:55 Bigger than a gigawatt, yeah. Everybody seems to want a gigawatt or more these days.

Emily Chang 18:00 If that sounds like a lot, well, it is, enough to power about 750,000 homes, 2,600 Tesla Model 3s, or 100 million LED light bulbs.

Speaker 18:12 There's a funny, uh, Back to the Future quote where he says, "1.21 jigawatts."

Masayoshi Son 18:17 1.21 jigawatts.

Emily Chang 18:19 [laughs]

Speaker 18:20 Uh, and because it's 1.2 gigawatts, it's kind of a funny...

Emily Chang 18:24 As more and more AI data centers come online in the next few years, they'll add many gigawatts of demand to an already overstressed grid. According to one estimate, data centers could consume more than 8% of America's electricity by 2035. And while the International Energy Agency predicts renewables will meet about half that demand, the rest will mostly come from coal and gas. Crusoe is even building its own gas power plant on site as a backup. Every major tech company has this net zero pledge by 2030. Are our AI ambitions running directly counter to our climate goals?

Speaker 19:01 It's complicated. [laughs] Um, my personal perspective is w- you know, none of those pledges are gonna be met by 2030. Um, and there's ways of kind of fudging it, where like, you know, you're buying a bunch of RECs or offsets. Uh, but the actual energy that's being used, there's no way it's gonna get to 100% carbon-free power production to power all of the AI infrastructure that's being developed. It's just, it's just, like, fundamentally not possible.

Emily Chang 19:25 So the tech companies are lying, or fudging it.

Speaker 19:28 [laughs] That's a, that's a good question. You're good. Uh [laughs] No, look, I, I mean, I think, I think, uh, a lot of these goals were ambitious and, and a lot of the goals, uh, predated the demand that people are experiencing from AI infrastructure. So I think AI has actually kind of changed their, uh, changed their path. Now, the optimist in me says that while it may get worse before it gets better, um, there is light at the end of the tunnel. Um, and I think the fact that, um, AI is a massive tool for invention, it's actually gonna accelerate the development a lot, of a lot of new technologies. The fact that we're building a lot of new, uh, infrastructure that requires a lot of energy, uh, actually provides this massive opportunity for AI to be, uh, one of the early adopters of new energy solutions, things like small modular reactors, geothermal, nuclear fusion. You know, the fact that, you know, AI has a massive demand for energy puts it in a position to actually intentionally decide what does the energy infrastructure look like that's gonna power it.

Emily Chang 20:32 For now, Stargate seems to have the resources

Emily Chang 20:35 it needs to accomplish its goals, but with huge investments come huge risks. Bloomberg's Brodie Ford has been following Stargate and the many ways it could go wrong or right for the Stargate partners and for the city of Abilene. Thanks for coming out.

Speaker 20:51 Yeah. Thank you. We got the cactus right behind you too.

Emily Chang 20:53 [laughs] I love it.

Speaker 20:55 Yeah.

Emily Chang 20:57 What's your impression of Stargate?

Speaker 20:59 It's happening, right? I mean, there's a lot of naysayers on the internet that nothing of it, about it's real, but it's clear, I mean, it is a massive project. The fact that they're talking about having, you know, a gigawatt of capacity live in about two years, it's an ambitious target, and seeing that in person kind of showed how frantic the whole thing feels.

Emily Chang 21:19 A gigawatt is huge.

Speaker 21:21 Oh, it's huge. It's comically big, right? I mean, nobody was saying these kind of numbers two years ago.

Emily Chang 21:25 Yeah. How does this compare to other big tech projects in other places?

Speaker 21:29 We've been tracking a lot of the big data center projects. I mean, data center spending has ramped in the last couple years, a past 200 billion in CapEx in the last year. I think two big projects that people are really watching right now is here in Abilene for Stargate and Elon's build-out in Memphis. Both have been thought to be the largest computing clusters at various moments, depending on who you ask, but I think what unites them is just the fact that they need so many chips, and it's the amount of labor coming into them.

Masayoshi Son 21:58 This is, to me, a very big thing, $500 billion, uh, Stargate project. I think it's gonna be something that's very special.

Emily Chang 22:06 The number.

Masayoshi Son 22:08 Yeah.

Emily Chang 22:08 $500 billion.

Masayoshi Son 22:09 [laughs]

Emily Chang 22:10 Is that for real?

Speaker 22:11 I don't think anybody thinks they're gonna hit 500 billion. Maybe I will eat my words.

Masayoshi Son 22:16 Now I came back...

Speaker 22:17 Softbank has a reputation for buying toward the top.

Emily Chang 22:19 [laughs]

Speaker 22:19 For [laughs] getting into projects once the hype cycle is already pretty advanced.

Masayoshi Son 22:25 I'm very, uh, uh, confident. Uh, we don't need $500 billion in one day. We go step by step, okay? So we are, we are very confident to make it happen.

Emily Chang 22:37 This is a lot of money, and, you know, not all of your bets have paid off. Where do you get your conviction?

Masayoshi Son 22:45 Well, uh, you know, sometime I get, uh, crazily overexcited, and, uh, I make mistake, like WeWork and, uh, you know, I have many scars. But when you have the conviction and the passion, you, you actually learn m- more from those mistakes and the scars, and that will make you stronger.

Emily Chang 23:10 Where's the money going? Like, who's the real winner here?

Speaker 23:14 So yeah, I think we've all seen Nvidia's stock go up [laughs] an insane number and said, "Whoa, how is that spending ramping so heavily?" It's these data centers, right? I mean, and what we saw today, 400,000 chips, that's conservatively $15, $20 billion.

Speaker 23:31 AI holds incredible promise for all of us.

Emily Chang 23:34 Let's talk about Oracle.

Speaker 23:35 Yes.

Emily Chang 23:36 Like, what is Oracle's role?

Speaker 23:38 For a long time, you know, the leaders in offering cloud infrastructure forever was, you know, Amazon, Microsoft, to a lesser extent Google. People weren't taking Oracle seriously for a long time, but with the AI boom, they have shown they're able to make these really vast data centers quickly. And so their role here is essentially putting in the servers, putting in the fiber optic cable, and getting that compute over to OpenAI. I mean, the underlying theme here is that OpenAI and companies like it need as much computing power as they can. They're looking for creative ways to, you know, conjure up those funds, and Oracle is a company that says, "Oh, God, you want to pay us to build out these data centers? Absolutely. Count us in."

Emily Chang 24:21 The sizes keep getting bigger.

Speaker 24:22 Yeah.

Emily Chang 24:23 And bigger. Is there a chance we're overbuilding? Do we need all this?

Speaker 24:28 God, every single day somebody has a different take, right?

Emily Chang 24:30 Yeah.

Speaker 24:30 Every single day, like we saw with DeepSeek, right? That, wait a second, do we really need all of this infrastructure?

Emily Chang 24:37 For those not obsessively following every twist and turn in the AI race, DeepSeek is a Chinese company whose own AI model went viral in January. Their AI performed at a similar level to leading US models, but used far fewer computing resources. This raised the question: are Stargate scale mega projects really necessary? DeepSeek appears to have found a more efficient way to power AI. Was that a moment of rethink for you, and are you doing anything differently now?

Sam Altman 25:06 I think the DeepSeek team is very talented and did a lot of good things. I don't think they've figured out something, like, way more efficient than what we figured out.

Emily Chang 25:12 But do you think there is a more efficient way to build AI?

Sam Altman 25:15 Oh, I'm sure. I'm su- we will... We have made incredible efficiency strides year over year, and I'm sure we'll keep doing that in the future.

Emily Chang 25:22 So if that's the case, why are you building all this?

Sam Altman 25:27 If we had an AI that we could offer at one tenth of the price of current AI, I think people would use it 20 times as much, and we would still need twice as much compute to satisfy the then current demand.

Emily Chang 25:38 Some companies seem less sure. Microsoft, for example, announced that it was pausing or canceling several data center builds in the US and abroad.

Speaker 25:47 A lot of folks worry that we aren't that far from seeing some kind of pullback, some kind of wall. It's kind of like how all the big tech companies were hiring every single person they could in 2020, 2021. Then at one point, they started firing people, and I think we're at this point where every time a lease gets canceled for a data center, people wonder, "Oh, did we hit this moment? Do we really need all of this infrastructure? Is there really a payoff in the end? Are folks gonna use these tools enough that it's gonna pay for all of this build-out?"

Emily Chang 26:17 How confident are you that OpenAI is going to be a financially sustainable and profitable enough company to justify all of this investment?

Sam Altman 26:26 It's looking like we're doing really well. I mean, like, we have to... We are definitely doing something unprecedented. Uh, you know, it seems like a... I, I, like, I feel confident in the bet. Doesn't mean something can't go wrong.

Emily Chang 26:40 What are the risks?

Sam Altman 26:41 I mean, maybe, you know, people stop wanting to, like, pay for AI services and then we have a difficult financial position.

Emily Chang 26:50 What do you say to the people out there who think this is a massive boondoggle?

Speaker 26:55 Uh, look, I mean, I think there's always gonna be doubters. Um, but I, I think the, again, I, the, the money piece, I think so long as AI is delivering value to society, the money will be there to build the infrastructure. I have no concerns around that. Um, even if it's not all 100% accounted for today, that money is there, and there's a lot of money that wants to invest in this category. What we're undergoing right now is sort of the largest infrastructure build in human history, right? This is like where we built the interstate high- highway system. You know, that enabled massive interstate commerce and transportation and catalyzed this massive boom in the US economy. What's happening in AI infrastructure is we're building the super highway system that's gonna enable modern intelligence systems. That's gonna create all sorts of new job categories that don't even exist today.

Emily Chang 27:47 Mm-hmm.

Speaker 27:47 It's gonna create new ways of doing business. It's gonna create new ways of invention. Um, and it's gonna really drive humanity forward.

Emily Chang 27:55 The data center boom isn't just affecting the AI business. There are local impacts, too. Across the country, places like Abilene are trying to attract data center projects, and the jobs and tax revenue that come with them.

Speaker 28:08 I make a dollar a year being mayor.

Emily Chang 28:10 Oh.

Speaker 28:11 So I've, I've gotta have another business.

Emily Chang 28:12 [laughs] So your side business is still the pest control?

Speaker 28:15 Uh-huh.

Emily Chang 28:16 Okay.

Speaker 28:16 Yeah, I'm a hired killer.

Emily Chang 28:17 [laughs]

Emily Chang 28:20 How would you describe Abilene?

Speaker 28:22 Well, Abilene, it's an old railroad town, you know, so we have lots of Western heritage. But that doesn't mean that we don't wanna grow and, and, you know, look forward to the future. We want our city to grow, and, and we look forward to these opportunities that are coming to us.

Emily Chang 28:35 What have you heard about the Stargate project?

Speaker 28:37 It, it could quite possibly be the largest AI data center in the world, and that excites me, and to know that we're gonna be a front runner in technology right here in West Texas.

Speaker 28:48 I've spent the last couple of months looking at some of these biggest projects and trying to see what documents we can find. How have these, you know, some of the world's largest, most powerful companies been interacting with these local municipalities? Largely what we have found is that these local towns are willing to forgo quite a bit of tax revenue, quite a bit of control, to be able to bring these big companies in.

Emily Chang 29:11 What is Abilene giving up to have Stargate here?

Speaker 29:14 So what Abilene agreed to with Stargate, with this data center, is 85% of tax revenue off the property tax, wiped.

Emily Chang 29:22 So they're, they're, they're losing out on 85%.

Speaker 29:25 Yeah. I mean, I'm sure they would say, and the developers would say, that you're still getting a lot of property tax revenue, and that's probably true, right? But these are pretty tough deals being struck, right? I mean, I think a lot of these smaller towns did not take place in the tech boom over the last 20 years. All of those jobs went to the coasts, right? And I think there's a lot of local city leaders who want to be able to say, "We're a part of this."

Emily Chang 29:50 Mm-hmm.

Speaker 29:51 And they're willing to strike some pretty extreme deals to make it happen.

Emily Chang 29:54 You're giving up 85% in, in tax revenue.

Speaker 29:58 Yeah. We're getting 15% of billions of dollars, so that's the way I look at it, right? It, it is a large tax increase. We're talking, uh, millions of dollars on an annual basis. It's gonna help us to pave streets, to hire police officers, to do the things cities do.

Emily Chang 30:11 When you were negotiating, did you feel like there was a risk they could go elsewhere? Like, were you worried about-

Speaker 30:18 Um-

Emily Chang 30:18 ... not winning the deal?

Speaker 30:19 Whenever you're attracting a prospect, yes, you always worry about that. But at the same time, if it wasn't gonna work for both parties, we weren't gonna lay down and just get run over by these people either.

Emily Chang 30:30 How many jobs do you think Stargate will really create?

Speaker 30:34 Well, I'll tell you what I've heard. [laughs]

Emily Chang 30:36 [laughs]

Speaker 30:36 So somewhere between 4 and 1,200. There's gonna be security. There's gonna be maintenance. You know, they're gonna have grass. People are gonna mow grass out there, so there's gonna be all kinds of jobs that go with something that large, too, other than just the technical jobs.

Speaker 30:50 We think about tech investments and tech campuses as kind of these thousands of high-paying jobs that are really kind of boosting local economy. Once data centers are completed, it's often not a lot of full-time workers.

Emily Chang 31:03 Mm-hmm. How many people does it really take to run a data center? Like, I keep thinking about that scene in Silicon Valley, and there's one guy in that data center.

Speaker 31:11 [laughs] There's really no difference between day and night down here, so it makes things easy.

Sam Altman 31:17 Um, there are gonna be quite a few people here. I know you're looking for a specific number from me-

Emily Chang 31:23 I know

Sam Altman 31:23 ... but I can't give it to you.

Emily Chang 31:26 What if they don't create as many jobs as you hope for?

Speaker 31:29 You know, I don't know a lot about, uh, AI centers or how many people they employ. I mean, I'm gonna be honest about that. But, you know, they're still building the facility. They're still putting mil- billions of dollars into the economy, and so we just trust them. You know, we have to have a little trust with these partnerships.

Emily Chang 31:48 Mm-hmm. We're in downtown Abilene right now. It's really windy. I can see why this is a good place for wind power. Ah. We just wanted to talk to people to see what they think about Stargate, if they've heard of it. Um, so we're gonna go try to find some folks and not blow away. So have you heard of Stargate?

Speaker 32:06 I have.

Speaker 32:07 Until recently, I had not.

Emily Chang 32:09 Have you heard of Stargate?

Speaker 32:10 No.

Speaker 32:10 It was announced by Trump, and that's, that's kind of really all that I've heard.

Speaker 32:15 I know that it has to do with AI and is supposed to bring maybe a lot of people to Abilene.

Speaker 32:19 It's gonna be a big s- uh, server farm-

Emily Chang 32:22 Yeah

Speaker 32:22 ... and, and employ quite a few people.

Emily Chang 32:24 What do you think about that?

Speaker 32:25 Computers are a good thing. I'm a computer guy. Did that in the Army.

Emily Chang 32:28 Mm-hmm.

Speaker 32:29 I invented computers.

Emily Chang 32:30 Oh, you invented computers?

Speaker 32:31 Yeah.

Emily Chang 32:32 What do you hope it brings to Abilene?

Speaker 32:34 I would love to see more people here, and especially, like, it would help local businesses.

Speaker 32:39 For Abilene, for West Texas, I think it creates a, a, a huge stream of jobs that aren't readily available. Meth and, uh, alcohol addiction is huge in this area because one of the only jobs available is oil and gas. And so this adds something different and, and hopefully breaks some of those cycles.

Emily Chang 32:54 Do you have any concerns that, like, about, like, how much power they're gonna use or water and...

Speaker 32:59 You know, I know that AI takes up a lot of power, um, to use, but I'm not too sure what exactly it'll do to Abilene. I, I feel like I'm not concerned on that side of things.

Emily Chang 33:09 It's gonna use a lot of power. Does that concern you?

Speaker 33:12 That's all good. What's wrong with that?

Emily Chang 33:14 Well, it might put some stress on the Abilene grid.

Speaker 33:18 Oh, they need to increase the grid, then.

Emily Chang 33:20 What do you think about AI?

Speaker 33:21 AI has terrified me for 20 years. I think there are so many questions that we don't have answers to when it comes to AI. I think we're biting off way more than we can chew.

Emily Chang 33:33 Do you use AI at all?

Speaker 33:35 I actually do use AI, yeah. Our professors at ACU, uh, sometimes recommend that we use it. I think it's a good thing. I think it can be used two ways, either, uh, creatively, uh, or kind of a crutch.

Emily Chang 33:47 So how do you feel about the fact that Abilene's gonna be the site of this big development?

Speaker 33:51 I love, I love jobs coming to Abilene, but I, I'm not excited about it. Where are we going, and how do you even fight it? The big picture to me is AI's gonna take over, and then where are we as humans? You know, I just, it, it terrifies me. It honestly just terrifies me.

Emily Chang 34:15 Well, mixed reviews. There's definitely a lot of questions. We'll have to see how the project pans out, and it's gonna happen over years, so it could be a while before we know for sure.

Emily Chang 34:29 There is this lofty promise that AI data centers are gonna create thousands and thousands of jobs. Meantime, AI is destroying jobs elsewhere.

Sam Altman 34:37 Yeah.

Emily Chang 34:37 And I feel like even, you know, there is serious anxiety out there. People are scared.

Sam Altman 34:42 Totally.

Emily Chang 34:43 E- even among the best engineers and technologists, people are scared. What do you say to those people?

Sam Altman 34:48 AI is for sure going to change a lot of jobs, totally take some jobs away, create a bunch of new ones.

Emily Chang 34:54 Mm-hmm.

Sam Altman 34:55 This is, like, the kind of... This is what happens with technology. And in fact, I, I think if you look at the history of the world, like, that's just been happening for a long time. The thing I think the world is not ready for, like, people have maybe abstractly thought, like, "Okay, it's gonna be a better programmer than me. It's gonna be, you know, better at customer support and whatever." I don't think the world has really had the humanoid robots moment yet. Like, the first day you're, like, walking down the street, and there's, like, seven robots that walk past you doing things or whatever, it's gonna feel very sci-fi, and I don't think that's very far away from, like, a visceral, like, "Oh, man, this is gonna do a lot of things that people used to do." So, so yeah, it's coming, and we, we have always tried to just be super honest about what we think the impact may be, realizing that we'll be wrong on a lot of details.

Emily Chang 35:39 Lately, predicting the future of AI seems harder than usual.

Speaker 35:42 [cheering]

Speaker 35:44 In a few moments, I will sign a historic executive order instituting reciprocal tariffs.

Emily Chang 35:52 President Trump's tariffs and the increasing volatility of geopolitics have created potential challenges for the AI infrastructure boom and shaken up the race

Emily Chang 36:01 to dominate AI between the US and China.

Speaker 36:04 Technology faces rising costs.

Speaker 36:06 Building these data centers just got considerably more expensive.

Speaker 36:09 It is very difficult to immediately decouple. Uncertainty is bad for business. Instead of moving forward, people have war rooms about, "Oh, what's tomorrow gonna look like?"

Emily Chang 36:24 Former diplomat Anya Manuel and Bloomberg's Sherine Ghafari have been closely watching how this could impact projects like Stargate. How much does the AI and data center supply chain depend on China?

Speaker 36:35 Yeah, so a lot of the materials that are used to build data centers, a lot of those materials actually come from China, right? Aluminum, steel, China's a major manufacturer. That could be an area of impact on raising construction center costs. Now, on the chip side, that's a much bigger, higher cost for these companies, right? Um, and that, there's a lot more uncertainty.

Speaker 36:56 The most sophisticated chips are manufactured in Taiwan, Korea, Japan, often brought to China to be integrated into all of the electronics that are in all of our pockets. This is a really complex, totally international supply chain, and I think the takeaway for everyone is you cannot just break the international trading system, and we cannot be a country that lives in autarky. We're always gonna depend on others.

Emily Chang 37:22 What does Stargate mean for the US-China tech race?

Speaker 37:26 I think it's one critical step in so many steps. When you look at the race that's on between the US and China, we're in a real race. The Chinese are giving us a run for their money.

Speaker 37:37 China's right behind us. I mean, they're v- we're very, very close. 50% of the world's AI researchers are Chinese.

Speaker 37:43 That doesn't mean we can't collaborate with the Chinese, and it certainly means we should keep talking to them, but where you used to have some level of trust and some level of communication, now you don't have it, and I worry that the tariffs, not directly, but indirectly, are making us lose that kind of trust.

Emily Chang 38:01 Now that we're in this position, there's obviously this tension between the US and China. Is this an opportunity to do it a different way?

Speaker 38:08 Yes.

Emily Chang 38:08 To build data centers a different way.

Speaker 38:10 I think that's exactly right. This is an opportunity to do it differently. What I would like to see is that more advanced manufacturing of the most sophisticated chips happens here, and I think TSMC is doing some of that. Others, the CHIPS Act really helped. Also, going back to Stargate, those data centers and all of the AI training and the AI inference shouldn't just be happening in the United States. It should be happening in friendly allied countries as well. Iceland, Greenland, the Nordics that have lots of geothermal power should be a part of it. Europe should be a part of it. Japan, our friends in Asia, Singapore. There's really an opportunity here to have a community of like-minded countries that are doing this together.

Emily Chang 38:52 Is President Trump alienating many countries that would be our allies in this endeavor?

Speaker 38:58 Yes, some of our friends and allies feel a little bit alienated. I hope, ultimately, that won't be the case. I think increasingly, the administration will understand how important our partners are and how much they add to us, and so I actually believe we'll come out okay in the end.

Emily Chang 39:15 [gentle music] There are still many questions about where the business of AI is headed. The expense of building and running these data centers makes it tough to turn a profit. OpenAI, for example, ran a $5 billion loss in 2024, and there's no guarantee that simply adding huge amounts of compute will make AI models smart and reliable enough to be truly transformative. So as solid and substantial as Stargate might seem, it's surrounded by uncertainty about how smart AI can get, which bets will pay off, and what kind of future we're creating. The tech industry goes through booms and busts.

Speaker 39:58 Mm-hmm.

Emily Chang 39:58 AI goes through winters and

Emily Chang 40:00 springs. How long is this boom gonna last?

Speaker 40:03 I think this boom has quite a bit of legs, um, to it. You just sort of look at the long-term investments that the biggest companies in the world are, are making in this, and I think, uh, you know, even with things coming out, like DeepSeek, uh, you know, I think e- everybody was pretty unanimous in terms of we're staying the course. What's required for AI to scale is a lot more infrastructure, um, and there's a bigger risk to under-investment than over-investment.

Emily Chang 40:29 If ChatGPT were to discover a, a Stargate, a futuristic wormhole that we could all travel through, what's on the other side?

Sam Altman 40:38 If ChatGPT could, like, transport us to the future?

Emily Chang 40:40 Yeah.

Sam Altman 40:41 I mean, no, no, no one knows, right? Do I think I could have sat here in 1905 and told you what we were about to discover in physics-

Emily Chang 40:51 Mm

Sam Altman 40:52 ... and that 40 years later we would, like, have an atomic bomb?

Emily Chang 40:54 Mm.

Sam Altman 40:55 Definitely not.

Emily Chang 40:56 Mm-hmm.

Sam Altman 40:56 Uh, I, and I, I think I am way too self-aware of my own limitations to sit here and try to say I can, like, tell you what's on the other side of that wormhole.

Emily Chang 41:04 Mm-hmm.

Sam Altman 41:05 I have no idea.

Emily Chang 41:06 But net, net, good, yes?

Sam Altman 41:08 Up and down. Not better in every way, but yeah, I think n- you know, net better.

Emily Chang 41:11 Up and to the right.

Sam Altman 41:11 Yeah, up and to the right.

Emily Chang 41:12 Okay.

Sam Altman 41:12 With some choppiness. [gentle music]