B (3:44)

Chapter 6 collateral damage. Just two years after creating Xai, Elon Musk had proven many of the experts wrong. He was no longer a have not with Colossus. He now really did command one of the most powerful data centers on Earth. Colossus was training Grok, Elon's steadily improving chatbot. And this year XAI absorbed X elon's social media platform and Grok was integrated into x dot com. You could begin to see a grander vision emerging. XAI would be the brain running the rest of Elon's operations. As Grok grew more intelligent, it would one day power his optimus robots, support operations at SpaceX, be the driver in self driving Teslas. That was the dream. But for all of that to happen, Grok would need much more computer. And so Xai announced it was scaling up yet again now to a million GPUs. The company made a new purchase. Another piece of land in Memphis. This one a warehouse larger than the oven factory. It would be anointed Colossus too. Elon, of course, was not the only mogul trying to move at hyperspeed. OpenAI's Sam Altman announced Stargate, which would also scale up to a million GPUs. Mark Zuckerberg was also talking about his giant supercomputer, Prometheus, followed immediately by an even bigger project, Hyperion. Watching all this, you had to wonder, as the spend got bigger and the energy needs ballooned, did the contestants in this race even know what the finish line was? This September, I got one answer. On stage at a conference, the hosts of the all in podcast interviewed Elon.

C (5:32)

You're scaling up your supercluster in Colossus in Memphis. Two now. Yeah.

D (5:38)

But there was a second one.

C (5:39)

Yeah.

B (5:40)

Could you give us an update? And Elon was asked the question, at what point in the GPU frenzy might enough be enough?

C (5:47)

Is there a point of diminishing returns? Or, like, how much more compute?

B (5:51)

If you throw twice as much compute.

C (5:53)

At it, do you get a 10% better model? Do you get 100% better model? Like, is it log linear?

B (6:00)

Great question. Elon responds. More GPUs will always mean better AI. Like, say, for argument's sake, like, 10x.

D (6:07)

More compute will double the intelligence.

B (6:09)

Maybe that's.

C (6:09)

That might be a rough rule of thumb.

B (6:12)

And for him, any increase in intelligence obviously, is worth the investment. And then he says, I think we'll.

D (6:21)

See intelligence continue to scale all the way up to where, you know, most of the power of the sun is harnessed for compute, and then ultimately, most of the power of the galaxy.

B (6:33)

Most of the power of the sun, and ultimately most of the power of the galaxy. That phrase stuck with me because it crystallized an essential truth about this AI race. For Elon, and maybe for the other CEOs he's racing, there is no finish line. It's not that we reach AGI and then we just get to stop. There is no limit to intelligence, which means there's no limit to the number of GPUs we will need. This partly explains the ravenous hunger driving this AI fight, its infinite quality. And to feed that hunger, tech companies are doing a lot of borrowing. Take Xai, for instance. Last year, the company raised $12 billion in equity, the kind of money we're used to seeing in startups. This year, XAI raised another 10 billion. But this money was different. Half of it was debt, meaning XAI took out a $5 billion loan, a loan it has to pay back with interest, like a mortgage. I talked to Miles Krupa, a reporter who covers AI finance for the Information, and he said that right now, there is a lot of people lining up to lend big money for data centers.

D (7:45)

It's sort of the biggest thing happening on Wall Street, I would say. You know, you see firms like Blackstone basically saying they're all in on data centers, and so they're owning data center companies, and then they're lending to data center companies. You see banks sort of tripping over themselves to reorganize themselves internally to attract data center companies that want to go out and raise a ton of money.

B (8:10)

I see. And how does that work? Like. Like refreshing my memory from the 2008 financial crisis, where people who are debt investors want to get in on debt because it's a nice way to get regular payments.

D (8:24)

Yeah, so that's, that's the promise here, is that these debt investors will get a regular payment on their interest. And, you know, given the fairly risky nature of this, probably a fairly high interest payment.

B (8:38)

That firm that Miles mentioned, Blackstone, it's one of the major lenders to US Data center projects. One of the ways Blackstone makes money is by buying up all sorts of debt. Credit card debt, car payments. And the company has latitude to make risky bets because it's private, so it's not regulated the way a bank would be. And the total amount of data center debt in this market is staggering. Analysts at Morgan Stanley say that by 2028 it could exceed $1 trillion. All of it fueling a construction boom the likes of which we have not seen in a very long time. So data center debt, a hot commodity. Eager investors think they can buy a piece of debt, say in a data center like Colossus 2. They can make money off those regular payments. Especially if GROK lives up to the hype and begins throwing off revenue. But what about the risk? Remember, XAI is a startup and right now has very little revenue relative to its debt, which you would expect might worry all of those investors lining up to buy the debt. What if things go south and the company defaults? But some of the investors that Miles has talked to believe that the loan is actually not that risky because no matter what happens, it's backed by a very valuable asset, the GPUs.

D (9:57)

Nvidia GPUs talking to people who are doing these deals. They're very quick to point out, you know, we've been doing asset backed finance for a long time. This is just another version of asset backed finance. These chips are highly valuable. A lot of people will want them. With that kind of asset, why can't you lend against that?

B (10:17)

The argument here is that if in the worst case, XAI defaults and at least the investors will have those chips, they're valuable collateral. But there may be a problem with this collateral.

E (10:29)

The trouble is that these chips have about a two to three year lifespan before a new generation comes along and you're faced with rapid obsolescence.

B (10:40)

Paul Kadrosky is an investor and a consultant to investors. He's the first person who pointed out to me the speed with which these Nvidia chips become outdated. Remember, all this started with just two Nvidia chips wired together by academics years ago. Those were much older chips, GeForce GTX 580s to be precise. You wouldn't use a 580 today, even to play a new video game. Fast forward to the cluster that trained ChatGPT that used a much more advanced chip, probably the A100. Just two years later, Colossus 1 would use an even more advanced chip, the H100. And then a year after Colossus 2, no surprise, will reportedly use the latest chip, the Blackwell. That is a lot of advancement from Nvidia in just a few years.

E (11:29)

So it's very different from prior waves where we've done this. If you think back to the 19th century and railroads. If I put a bunch of rail in the ground and I didn't use it for five years and then came back, there was no one saying, whoop, that's obsolete. That'll never work. Right? It still works. Same thing was true with the fiber optic build out in the 2000s period. So this is very different where you have expensive, fast depreciating chips at the very core of one of the biggest capital spending bubbles in the history of modern economies.

B (12:02)

There's a reason Paul is talking about these two historic infrastructure bubbles, railroad and fiber. It's because they're the two precedents that AI boosters often point to. I've heard them do it a lot, the founders and investors on podcasts I've listened to, they'll say, even if we're overbuilding data centers, it's okay because we've overbuilt useful infrastructure in the past, we always end up using it. Think of Loudon, how Buddy Reiser ultimately used all that fiber in the ground. The thing about that fiber, it's just glass strands in a plastic jacket. It's so resilient that much of what was laid in the ground in the 90s is still functional today. But will that be true of the AI data centers we're building? Will we be able to use them as is, many years from now? Here's where we are right now. Billions upon billions of dollars of debt have been taken out, much of that debt backed by Nvidia chips as collateral. And now everyone is trying to figure out exactly how the value of those chips goes down over time. I asked Anissa Gardizzi, the data center reporter, you're at a data center conference right now. You're in Vegas. You're talking to me from a hotel. How big of a topic is this? Like, how big of a question is this something people are talking about? This is the main thing that people are here talking about. No one really has an answer, which is, I think, just a reflection of where we are in this cycle, some of the projects are getting so big that it's almost hard for people in the industry to really understand what' going on. Miles, the finance reporter, has been experiencing that same degree of uncertainty. He told me he's gotten very different ranges from companies. When he asks how long do Nvidia chips hold their value? The answer from one AI cloud company was six years. Although that company, Core Weave, they've borrowed a lot of money to buy chips, so you might expect them to say that. Others give a more conservative answer.

D (14:06)

The major clouds, Google, Amazon, Microsoft, are expecting more in the sort of three to five year range, which, I mean, three years is pretty different from five, and then three years is pretty different from six. And then there's even a startup called Grok with a Q, not to be confused with Elon Musk's Grok with a.

B (14:27)

K. That's a very unfortunate name.

D (14:29)

Yeah, whose CEO Jonathan Ross, is basically saying it's one to two years. The, the useful life of these chips is basically as long as they can last until Nvidia comes out with the next best, greatest thing. So it's really kind of up in the air, I would say.

B (14:49)

We reached out to some of the companies making these assumptions. Nvidia declined to comment. Core Weave, the more optimistic of the bunch, told us that their assumptions are grounded in their ability to, quote, recontract older processors at high value, meaning even when the chips aren't the latest model, they'll still get used for something. There's such a huge demand for compute, and yet if you zoom out and look at the kind of funding deals that are happening these days, what you notice is that many of them have become more circular, more baroque as companies try to get these depreciating chips off of their balance sheets. A lot of companies are doing it OpenAI Meta and this year, Xai. Xai needed billions of dollars worth of Nvidia chips for Colossus too. But the company did not directly borrow the money to buy the chips. Instead, a longtime ally of Elon stepped in. Antonio Garcias, who runs a private equity firm. He is raising $20 billion, the majority of which is debt. He's putting it into a separate company set up just for this deal. That company is buying the Nvidia chips and then leasing those chips to Xai. If your head isn't spinning already, Nvidia is one of the investors in the company that has been set up to buy those Nvidia chips. It's all very circular, and Anissa says That these kinds of deals are designed for a very simple purpose, to get as many chips as possible as fast as possible. Based on what I can tell from sources in the industry, this whole idea of leasing the chips is a way to unlock more capital and make it easier for investors to lend money for this asset. But that's not an actual solution. That's an accounting trick. Yes, not much is changing besides how things are being accounted for. Chapter seven. Does this make sense? Can this make sense? A bubble almost always has at its core at least one widely shared and flawed assumption. For instance, in the housing bubble of the 2000s, people wrongly believed that houses were being valued accurately and that housing prices would always keep going up. Many smart people created lots of complicated debt instruments so they could profit off those housing prices. Instead, they almost took down the entire banking system. And communities everywhere were devastated, including Loudoun, including Memphis. After the crash in 2008, regulation was passed to make it harder for banks to make those kinds of risky loans. But these days, those loans are instead being made by very large, very opaque firms like Blackstone. As one longtime finance editor said to me, people are definitely doing stupid stuff in AI right now. Some companies will go bust, but the more important question is not whether there's a tech bubble, but whether it's a debt bubble. Meaning has all this AI excitement caused people not just to value some stocks too highly, but to actually lend money that won't be repaid? And if so, when that bubble pops, will it affect the world outside of AI? Will it affect the rest of us? Paul Kadrosky thinks the answer is yes.

E (18:28)

It's very, very difficult to avoid being exposed to what's happening because it has found its way into all these different assets, ranging from life insurance to private equity, well, to the S&P 500, which is now dominated by largely AI centric companies, to REITs. Very conservative. You thought real estate income trusts that are, you know, owning commercial real estate, like apartment buildings and office blocks? Well, over the last four years, data centers have gone from a 0% share of REIT ETFs to. To somewhere between 10. And at the high end, I've seen 18%.

F (19:03)

Wow.

E (19:03)

Of the assets of a REIT. So, weirdly enough, you're making a conservative investment, no different than being a conservative life insurance policyholder. What do you know? You're actually investing in data centers.

B (19:14)

It's so wild. It's so wild that it's. That it's. That it's everywhere. Like, it's everywhere.

C (19:24)

No, no, no.

E (19:25)

It's It's, I, I, it's, it's a terrible analogy to use, but I say it has literally metastasized through the economy. It's very hard to escape. Leaving aside that AI itself can be hugely valuable and useful, technology has mutated into this real estate bubble around these things called data centers, which many have the same features as the stuff that caused the financial crisis. Hidden debt, syndicated financings, synthetic instruments, where I roll up the rents from a bunch of different data centers and then securitize those. All the kinds of stuff that should make your head spin in circles if you paid any attention during the financial crisis are happening again.

B (20:10)

So that is the bear case from Paul Kodroski. Other folks I've talked to, like Miles Krupa, are much more cautious. Miles says he doesn't know that we're in a debt bubble per se, but certainly believes that there are pockets of bubble. And Xai concerns him. When he looks at Xai, he sees a company that's taken on more debt and more risk than most. I asked him if he could justify all of this from Elon Musk's perspective.

D (20:37)

I think the simplest explanation is basically, we are deep into the biggest technology arms race in the world, and Elon Musk sees no option, other than getting these chips as fast as he can to build what basically amounts to a digital God that some of the smartest people in tech think has infinite commercial potential. I mean, it's, you kind of have to go through those logical steps to justify the kinds of things they're doing, the lengths they're going to, to try to get their hands on these chips.

B (21:18)

It's funny, I can hear you as a business reporter, Miles, like, straining to explain to me, like, the logic of this. You're like, well, you know, the revenue that would come from the digital God obviously would pay for all of these lease agreements.

D (21:33)

I'm a skeptical business reporter, but, I mean, I see the rationale, I see the potential payout, but it really does take a kind of leap of imagination. And I think AI in a way that almost no other technology has before, it has kind of like captured the imagination of technology leaders and gotten people to do things that they probably wouldn't do otherwise. In pursuit of this kind of commercial nirvana, I don't really know how else to put it.

B (22:16)

Commercial nirvana, for me, is a very helpful phrase. There is something almost religious about all this. Our tech moguls have developed faith, faith and a vision for tomorrow where intelligence knows no limit, where we will one day harness the power of the sun to feed a machine God that we have created. Created. If you believe that, all this does make sense. But what if you don't? What if you don't buy the vision and one day a billionaire shows up in your city with plans to hastily build a juggernaut. After The Break Memphis vs. Colossus.

C (23:01)

It.

A (23:30)

This episode of Search Engine is brought to you in part by Bombas. This time of year, it is sensory overload everywhere, but one feeling we're all still chasing is cozy. And Bombas has the socks, slippers, tees, basically everything you need to get there. Honestly, there's something therapeutic about putting on a fresh pair of socks, and Bombas has figured out how to bottle that feeling. Their slippers have sink in cushioning, their tees have the perfect weight to them, and that comfort keeps going after the first wear. And when it comes to gifting, Bombas has answers for everyone on your list. Running socks for the marathoner, Baby socks that actually stay on tiny feet. They've got something for everybody, and this season they're really stepping up their slipper and slide game with new styles and materials. What I really like about their mission is that every time you purchase, Bombas donates one pair to someone facing homelessness. So when you're cozy, someone else gets to be too. Head over to bombus.com engine and use code engine for 20% off your first purchase. That's B O M B A S.com Enjin code Enjin @ checkout this episode of Search Engine is brought to you in part by Rocket Money. You might think you have a solid handle on your budget, but if your bank account isn't reflecting that, something's off. That's where Rocket Money comes in. Rocket Money is a personal finance app that helps find and cancel your unwanted subscriptions, monitors your spending, and helps lower your bills so you can grow your savings. It shows you all your expenses in one place, including subscriptions you forgot about. If you see a subscription you no longer want, Rocket Money will help you cancel it. You can even automatically create custom budgets based on your past spending. Rocket Money helped me spot a couple of subscriptions that I had completely forgotten about, and I even used it to negotiate lower bills for myself, which was pretty fantastic. Rocket Money has saved users over $2.5 billion, including over $880 million in canceled subscriptions alone. Their 10 million members save up to $740 a year when they use all of the app's premium features. Cancel your unwanted subscriptions and reach your financial goals faster with Rocket Money. Go to RocketMoney.com search today. That's RocketMoney.com search.

B (25:42)

Are my AirPods giving me cancer? What is cancer? Why are we having fewer babies? What do I do if I see a bear? And does size really matter in government? Welcome to Smart Girl, Dumb Questions. I'm your host, Naima Reza, and in this new show, I ask simple questions to big thinkers. Like I asked Neil DeGrasse Tyson whether it's a problem to believe in horoscopes only if you wanted to become head of NASA, but otherwise there are plenty of jobs for you in the world. Or I asked Mark Cuban if billionaires can save us, what it's like to be so rich and why everyone has an opinion on everything these days. Everybody's so sure, Mark.

C (26:17)

So sure.

B (26:17)

Yeah, I'm surrounded by sure people. And I feel like, am I a moron? Because, like, I'm not sure what the answer is. A lot of times that's probably true that I'm a moron. I have to change the name of the show, right? Smart Girl, Dumb Questions. It's a curiosity party. And you know you're curious, tune in on Apple, Spotify, YouTube, or wherever you get your podcasts. Chapter eight, Power Struggle. So what are you seeing in your viewfinder?

C (27:00)

I mean, it just zooms in. It just zooms in. So if you want to put it to your.

B (27:03)

Oh, wow, this thing is super.

C (27:05)

It's just for golf. Yeah, it's a little golf thing.

B (27:08)

I got to do a little bit of spying on Colossus 1. When I was in Memphis, I was standing outside the fenced perimeter with an investigative reporter named Sam Hardiman. He works at the Daily Memphian, loves golf, hence the rangefinder, which he's been using this past year when he visits the data center to watch the build out.

C (27:25)

Yeah, and so I'll count some cars in the parking lot. I'll just see how many people are here. So what we're looking at here, this crane? Yeah, this crane here is building another substation. So this will be when XAF fully connects to the grid and TBA needs a full blown.

B (27:38)

Sam's been keeping an eye on this place since before it had anything to do with Elon Musk. He visited back when it was just a factory in the Electrolux era. The outside still looks the same, he says. A single story, gray block, almost a million square feet. Though these days you can see an XAI flag billowing above it all. It's just funny to me because standing here, I'm like, I've been to data centers. To be honest, this doesn't look like a data center. It looks kind of like. Like a dusty old factory.

C (28:05)

Yeah, I mean, it's a factory. We're here next to an old soybean field. There's a steel factory couple hundred yards to our left. This used to actually. They're letting this one be fallow this year. But this last year was all corn.

D (28:19)

Just to our left, like we would.

C (28:20)

Have been corn would have been right here.

B (28:22)

We move our rangefinder to take a look at the most exciting feature of this otherwise unremarkable site, the subject of all the controversy. Right in front of us on the other side of the fence is a cluster of shiny silver smokestacks which are right now producing a very loud mechanical roar. The turbines.

C (28:45)

So those are the turbines, and you can see the one on the right there. At one point, they all looked like the one on the right.

B (28:50)

For the group of Memphians who wanted to shut Colossus down, boot Elon Musk out of their beloved city, these turbines would become maybe their best shot. Here's what happened last year. As you know, Elon is trying to get Colossus up and running at a pace that every everybody told him was impossible. You cannot just build a supercomputer from scratch in a few months. But Elon was determined. He'd already raised $6 billion. He poured a big chunk of it into the latest Nvidia chips and other AI hardware. They figure out all the tricky cabling, the water cooling. But the thing that ends up being the hitch is power. Electricity. The cluster is ready for training. With a connection to the grid, that substation is not and would not be ready for many months to come. Maybe in another scenario, in a world where the race to build a machine God wasn't looming over it all, in a world where billions of dollars were not at stake, Xai might have waited. Instead, the company found a workaround. It brought in dozens of these shiny silver natural gas turbines, essentially mobile generators, the kind that you would use during a natural disaster like Hurricane Katrina. Except standing here, one thing is crystal clear. Calling these things mobile generators is very funny because each one is the size of a house.

C (30:13)

You know, people can't see, but they're giant things. I mean, they're, you know, probably 30.

B (30:17)

They all look like little. Like mini factories with little factory stack.

C (30:22)

I mean, this is essentially a mini power plant, what we're looking at right now. And I mean, just to be thought.

D (30:27)

Of that way, conceptually, for people Because.

C (30:29)

I think people, when they think turbines, they think the thing that's connected to your house.

B (30:34)

So these turbines are giant. And at one point there were as many as 35 of them, each one emitting while it ran an unknown amount of pollutants into the local air. Sam Hardiman has done a lot of reporting on these turbines. He says the questions were raised when they were first brought in about whether they were legal or safe. But the fight around them really got going months later this January. Local politicians and community members felt that they had not been consulted about this enormous Elon Musk AI project. They were not blown away by the cool cabling or by Grok, the world's most popular anti woke chatbot. What they saw were the turbines and what they wondered about was their emissions. So those concerned Memphians started filling the seats at county government meetings, many of them from the Boxtown area. Boxtown being the neighborhood in the shadow of Colossus and new at 5, the XAI debate is heating up over just how safe the computer company Elon Musk has built in Memphis really is.

D (31:38)

Some people are not happy about the.

B (31:39)

Environmental impact of the facility. In Southwest, a group of Boxtown residents and local organizations are calling out city leaders, urging them to take a closer look into what they say is a violation of the law. By April of 2025, XAI had run its quote unquote, temporary mobile turbines for almost a full year. The company would need a permanent permit to keep using them past that one year mark. That permit was at the discretion of the county's health department. Department. The health department is now gathering comments from the public on XAI and will hold a community meeting April 25 to hear from the public and let them know if there is reason to worry. I don't know if there's been a time in American history where the public has had so little faith in the tech moguls, such mistrust of the visionaries. The question hovering in Memphis then it wasn't, do we trust Elon Musk? The question was, given that many of us don't, do we get a say in what he can do in our town? The public meeting took place in Fairleigh High School. Sam Hardiman was there.

C (32:49)

It was. It was a hot day. You know, it was like just Memphis starts to really heat up in about April. It really starts. Oven starts to get cooking. And the people were standing in line for hours to get in because there was a lot of security, like a tremendous amount of security for a public meeting. There had been A rally outside before.

B (33:11)

All right, let's march in. Sign up for public comment. Let's show them who we are.

C (33:18)

Area. There was national media there from all over. I mean, from. From Europe even. I think German news stations were there. Like it was. There was a lot.

B (33:25)

I mean, why do you think that was? I mean, do you think it's AI? Do you think it's Musk?

C (33:29)

It's a combination of both, right? It was a. The richest man in the world, prominent Trump supporter, you know, unpopular president in a city. And you're probably in precincts that voted, you know, 90, 10 for Harris, right? And then you have this narrative of Elon Musk is using illegal turbines to poison people. And so it was a hot button issue.

B (33:48)

So before we get started, the purpose of tonight is to. To make sure that we record your public comments. This is a public hearing. Everyone is gathered in the school auditorium. Hundreds of people. The head of the county's health department is leading this meeting. She brings on Elon Musk, deputy for the Colossus project. The closest thing they'll get to having Elon in the room. Representative up from the applicant XAI. Mr. Brent Mayo, Tacoma. And have a few words with the.

C (34:28)

I think Brent Mayo. I barely got in the room by the time Brent Mayo was speaking. He spoke at the beginning of the meeting. You read some talking points. Largely couldn't hear him because people were just booing him.

E (34:38)

As a leader in future technology, we.

C (34:41)

Will continue to provide economic growth for the city of office and national. We look forward to being the second largest contributor to tax dollars, providing tens.

B (34:53)

Of millions of dollars in local tax revenue. It won't come to us. Brent Mayo tries to get his talking points out. Like he mentions tax revenue, which this is true. XAI will pay $25 million in taxes in just its first year, 3 million of which are promised to Boxtown and the neighborhoods closest to Colossus. The room's not interested. What you hear instead is people shouting, do you think we're dumb?

C (35:17)

And then he left through the side door. No, you stop killing us.

B (35:29)

And then the people get to speak. It's a mix of politicians, activists, residents. We here because we demand that the Shelby County Health Department denied XAI's permit for 15 gas turbines that will pollute our community.

A (35:45)

That is always.

B (35:47)

If we don't protest now and try to do something now, then nothing will ever be done. Elon Musk, we do not want here in. No, we can't say any plainness. We want. Do not want him in nothing. The Argument in this room is that Elon Musk needs to be evicted from Memphis because of all of xai's air pollution. From one perspective, this doesn't entirely make sense. Memphis is an industrial city with a lot of companies generating emissions. Take FedEx. FedEx operates nearly 400 daily cargo flights out of its hub in Memphis. Planes, terrible for the atmosphere. But those flights, those emissions, which have been happening for years, they are not causing protests or town meetings. Of course, there is a very big difference. People like FedEx, they understand why package delivery is valuable. The company provides lots of jobs. Xai, on the other hand, likely won't bring many jobs to this community. And tax revenue, like a supercomputer, is just a very abstract thing. I think about how differently things had gone in Loudoun county, how differently data centers had been received there, and how differently they'd been unfurled, not overnight, not with a name like a sea monster. What we call a new technology shapes how we understand it. And people in that room apparently did not see Colossus as a gift. I asked the man who helped summon Colossus here, Ted Townsend. AI is new, but data center projects are not. Did XAI somehow err in moving as quickly as it did? What I understand is that other companies like Microsoft, Amazon, Google, that have a lot of experience doing data centers, have never used mobile generators in the way that was used in the XAI site. And I've heard people say, well, that's why Microsoft wouldn't have done it this way. That it is, in a way, a byproduct of how fast XAI was moving and how fast the city allowed it to move.

C (38:07)

Yeah, yeah. Well, you know, I can't speak for other companies and what their timelines are. I know XAI was aggressive, and if they could be aggressive and hit milestones doing so by the letter of the law, then that's what they did. And AI is a race. XAI came into the race late, admittedly. Elon has said that. So they knew that they had to build very quickly. And that's why I think you've seen the variance in this project versus others across the country.

B (38:40)

Absolutely. And I think, I mean, it's your first big data center project. It's their first big data center project. I'm curious if you. You're not the boss of Elon Musk. Nobody is. But if you were doing it, you are doing it again with Colossus, too. I'm curious, what would you do differently or how would you ask the company to do it differently?

C (39:03)

At this point, I wouldn't. I think they have built out responsibly. I think that the investments that they've made in our infrastructure and protecting our resources is second to none. I don't see anything that needs to be done differently.

B (39:20)

So that is Ted Townsend's view, and it seems to match the view of the Memphis government. A few months after that emotional meeting at Fairleigh High School, XAI got its permit. Colossus would stay, so would the turbines.

C (39:38)

And so now here we are pulling up the Colossus 2.

B (39:44)

That's the former warehouse. A huge white. Garrett and I drove with Sam Hardiman to the site of xai's newer, bigger data center, the one that's scaling up to a million GPUs. It's about 20 minutes south of Colossus 1, right on the border between Tennessee and Mississippi. It's still under construction. There's just, like, big mounds of dirt and diggers.

C (40:04)

Yeah, we'll just turn around up here.

B (40:07)

Sam brought us here to show us the lesson that Elon Musk and Xai seem to have taken away from their big turbine fight. Although it's probably not the lesson the community would have hoped for.

C (40:17)

As you can see, more turbines right there. Big old stacks over there.

B (40:21)

Yeah. So wild. See turbines. Colossus 2 will use dozens more mobile turbines, but this time, XAI placed them very carefully just on the other side of the state line, about a mile south of Memphis in Mississippi.

A (40:37)

And so tell me where the state line is.

C (40:39)

It looks on the map like we're.

A (40:40)

Right on top of it.

C (40:41)

We're on top of it. It's right there. Oh, wow.

B (40:43)

That's Mississippi. That's what?

C (40:45)

Mississippi.

B (40:46)

So this road is the dividing line between.

C (40:48)

It's right there. I think the stakes are in there.

A (40:50)

We're in Tennessee, but If we drove 10ft to our left, if I.

C (40:53)

We literally threw a baseball out the window, you would throw a ball in Mississippi. It's simply like the Greater Memphis Chamber at one point issued a news release, said there would be no new mobile turbines in Shelby County. Well, that's easy when DeSoto County, Mississippi is right next door.

B (41:06)

How convenient.

C (41:07)

Very convenient.

B (41:09)

We drive on. Chapter 9 Boxtown. The neighborhood is just a short drive from the Original Colossus. Colossus 1. On our way there, we pass by the wastewater treatment plant that XAI is building. On our right is a big natural gas plant and then a sewage treatment plant. And then we take a turn into this shockingly green verdant zone. It's taken us less than 10 minutes to drive from the center of industry to. To this quiet residential area.

C (41:43)

This will be Boxtown that we're coming up. Yeah, Boxtown Road. Those will be the closest houses. Texas AI.

B (41:50)

It's more sparse compared to other parts of Memphis. We've driven through a few pockets of houses and then green woods. Yeah. But it doesn't seem like a lot of people live here.

C (41:59)

The census tract data would show you there are not a lot of people here.

B (42:03)

About 3,000 people live in Boxtown, which is a historically black neighborhood. A century and a half ago, black Americans who'd just been freed from slavery built homes here out of wood from old railroad cars. There we go. Okay. Here. Hello. Hello. Perfect. Garrett and I returned to Boxtown the next day. We wanted to talk to the people who lived there. Can I just have you say your name?

F (42:29)

Parady. P A R H A D Y l. Bird.

E (42:33)

Jefferson Jr. Ray.

C (42:35)

Ray.

F (42:36)

That's. Call me Ray Ray.

B (42:37)

It was a Thursday afternoon. Pretty quiet. But we walked up to a few people who were just standing outside their homes. Have you been here a long time?

E (42:45)

All my life.

B (42:46)

Born right over there in the home right over here. So that was in 1961.

F (42:50)

I've been here a long time.

A (42:52)

Really?

F (42:52)

See, I was raised up. Down through here. Yeah.

B (42:55)

Ray Lewis worked as a sanitation worker for 44 years, but he's retired now. When we saw him, he was sitting on his front porch on a chair and just watching truck after truck go by.

F (43:07)

You just hear these trucks right now, but after a while, when they get through doing what they're doing, it be so quiet down here, you can't hear a peeing fall.

B (43:15)

Do you like that?

F (43:15)

Yeah. Who, me?

A (43:17)

Love it, love it, love it.

B (43:20)

The people I spoke to that day had a lot of pride in Box Down. It was like a slice of country inside the big city. We're. We're country folks.

E (43:29)

I hate to lose that sense of peace.

B (43:36)

Albert, who's also retired, he used to work for the city's legal department. I talked to him and his brother Clarence in their backyard while their other brother was doing work with a tractor. The lawn up. And he's older, just clearing off some stuff. So the lawn is in wonderful shape. I want to say. It's a large area. A lot of lawn. Keep up. The two of them knew about the Xai Project. Albert had been to a community meeting about it. But they had not been involved with the protests. Their feeling was air quality had been a problem for years around here. This fight felt a little political. The thing that was most on Elbert's mind was actually not a data center. It was garbage.

E (44:14)

We've been dealing with the problems associated with dumping of a humongous number of tires and all that, you know.

B (44:25)

How long ago was this? The hundreds of tires, Boy, what this been?

E (44:28)

Months.

B (44:29)

Just months, months ago. Go right across the track up here and you may still see a big stacked pile of tires.

D (44:35)

It was a boat on the street.

B (44:36)

Just the other day, somebody dumped there. You know, we asked for cameras, everything. Now here. And we're not getting it. The brothers had no idea who was dumping tires and boats in their neighborhood. They'd even found stolen cars left burning in the middle of the street. Albert said he called local officials who told him that they would deal with it, but they didn't. And so the brothers cleaned it themselves with their own tractors. Ray Lewis, on the other hand, the man on the porch, he had a lot more to say about Xai and Elon Musk.

F (45:12)

Really, truly, I've been hearing a whole lot about Elon Musk. Elon Musk. You know, how many million dollars is being spent in Elon Musk? But see, the point is this. How many by then putting this here in Memphis, right? How many folks they gonna hire to work with this company?

B (45:32)

Ray's question. How many people in Boxtown were actually gonna get jobs here?

F (45:38)

Because if they just gonna stick it up there and don't hire nobody, then how in the world you gonna get rid of the crime? No job. That's what makes crimes. Don't get me wrong, you know, crime is high. But now you gonna throw up a building out here in a black neighborhood and then don't hire anybody on that job. It's just like pouring milk in a paper sack with a hole in it. You know what I'm saying?

B (46:06)

See?

F (46:08)

And then what I'm saying is just like this here, you know, waste makes warm.

B (46:11)

It's like pouring milk in a paper sack with a hole in it. There are a lot of problems in this world. Even more problems than can be solved by all the supercomputers on our horizon. And the thing about many of our most important problems is that they are not exciting. And yet they need to be solved by very smart people who are not yet bored by these mortal scale dilemmas. No one I talked to in Boxtown was awaiting the new machine. God, they just hoped when it was all over, there might be some new jobs, Less crime, maybe. Colossus 3 in its idle time between finding the meaning of life or locating aliens could finally just do something about the pilot.

F (47:00)

Okay. So that's all I got to say.

B (47:03)

That's really helpful, sir.

A (47:04)

Thank you. All right.

F (47:05)

Thank you.

A (47:12)

Reporter Shruti Pinamani. She's also Search Engine's editor. This episode is brought to you in part by Odoo. Running a business is hard enough, so why make it harder? With a dozen different apps that don't talk to each other. One for sales, another for inventory, a separate one for accounting. Before you know it, you're drowning in software instead of growing your business. That's where Odoo comes in. Odoo is the only business software that you'll ever need. It's an all in one, fully integrated platform that handles everything. CRM, accounting, inventory, E commerce, hr and more. No more app overload, no more juggling logins. Just one seamless system that makes work easier. And the best part, Odoo replaces multiple expensive platforms for a fraction of the cost. It's built to grow with your business, whether you're just starting out or already scaling up. Plus, it's easy to use, customizable, and designed to streamline every process so you can focus on what really matters running your business. Thousands of business have already made the switch. Why not you try Odoo for free at odoo.com, that's o-o o.com this episode of Search Engine is brought to you in part by Perfectly Snug. You ever notice how a bad night of sleep just wrecks your whole day? I used to wake up drenched in sweat, kicking off the covers, turning the thermostat down, then arguing with my partner because they were freezing. That all changed when I got the Perfectly snug Smart topper. It's like air conditioning for your bed. Not air conditioning, but the same kind of cooling comfort. And honestly, it made my old mattress feel brand new. If you deal with night sweats, hot flashes, or just sleep hot, this thing is a total game changer. And my favorite part? The dual zone controls so my partner can stay warm while I stay cool. You can pay with FSA or HSA funds. It's super easy to set up. Literally drop it on, plug it in, done. And there's a 30 night risk free trial with free shipping. Both ways. Seriously better. Sleep changes everything. Black Friday is their biggest sale of the year. Save up to $330 with code BLACK Friday. Visit perfectlysnug.com. This episode of Search Engine is brought to you in part by Quints. Cold mornings, holiday chaos and all the errands I just want my wardrobe to be simple with pieces that look sharp and feel amazing and that I will actually wear. Which is why I love Quint's this season. Quince makes it easy. $50 Mongolian cashmere sweaters that feel like a little everyday luxury and wool coats that are stylish, durable and made to last. Their denim fits perfectly and is insanely comfortable, all without the high end price tag. By working directly with ethical factories and top artisans, Quince cuts out the middlemen so you get premium quality without the premium markup. I recommend their cashmere sweaters. They hold up beautifully, wash after wash. But it's not just clothing. Quince has thoughtful gifts for home, bath, kitchen and travel. Give and get timeless holiday staples that last this season with quince. Go to quince.com searchengine for free shipping on your order and 365 day returns. Now available in Canada too. That's Q-U-I-N-E.com search engine free shipping and 365 day returns. Quints.com search engine. Search Engine is a presentation of Odyssey. It was created by me, PJ Vogt and Shruti Pinamaneni. Garrett Graham, who's actually from Memphis, is our Senior producer. Emily Voltaire is our Associate producer. Theme, Original composition and Mixing by Armin Bazarian Fact Checking this Week by Mary Mathis Special thanks this week to Ken Brown, Sam and Margaret Graham, Asa Fitch, Alex Bloomberg and the Listening Group. Our Executive producer is Leah Rhys Dennis. Thanks to the rest of the team at Odyssey Rob Morandi, Craig Cox, Eric Donnelly, Colin Gaynor, Maureen Curran, Josefina Francis, Kurt Courtney and Hilary Shaw. If you'd like to support our show, get ad free episodes, zero reruns and give us the financial support to make more ambitious and longer investigations such as this one which take travel time. Resources Please consider signing up for Incognito Mode. You can join Incognito Mode at Search Engine Show. Please follow and listen to Search Engine wherever you get your podcasts. Thanks for listening. We'll be back soon with a new episode.

D (51:49)

Sa.