Data Centers: Can't Live With Em, Can't Live Without Em

0:00This is an iHeart Podcast. Guaranteed Human.

0:31on the iHeartRadio app or wherever you get your podcasts. Welcome to Stuff You Should Know, a production of iHeartRadio.

0:45Hey, and welcome to the podcast. I'm Josh, and there's Chuck, and Jerry's here too, and we've got our pocket protectors and tape on the bridge of our glasses. Hey. And this is Stuff You Should Know. Nice work. Did you just hear something? No. Oh, weird. I heard you say nice work and then stop abruptly. Well, I stopped abruptly because I thought I heard a little digital glitch. Oh, no, I didn't hear anything. I might be losing my mind then. You know, I'm curious whether we end up editing this out or not. Any other podcast on the planet would edit that out without even thinking about it. But there's like a 50 % chance it'll stay in with us.

1:27I mean, this is why we didn't get that Golden Globe nomination. That's right. This kind of classic stuff you should know, unprofessionality. That's exactly right. That and the Italian accent, the mispronunciations, there's a whole laundry list. Yeah. That's okay, Chuck. I think we're golden regardless. Agreed. So we're talking about data centers, which I had a very, very rough idea about, but actually, no, I knew that they existed, essentially, and that they were becoming a problem with the rise of AI. That was about it. How about you? Are you a data center filiac? No, you know me. I'm not super technology-minded, so I don't know a lot about this stuff.

2:12I remember walking by our server room back in the day when we were at Ponce City Market and seeing our colleague Izzy in there hard at work. Yeah. And when that door was unlocked and open, hearing the whir of the servers and the cooling machines. And, you know, that's on a smaller scale, that's a data center. Absolutely. 100 % that's a data center. It was also a great place to curl up and take a nap in the middle of the day. The warmth of the server. Yeah. And the whir puts you right to sleep. Yeah. So, yeah, that definitely counts as a data server. If you have like one of those little home networking setups in like a closet in your house, data center.

2:54Sure. Technically, the PC is a data center. Anywhere you can store and access data, that's technically a data center. And you're like, well, that's stupid. Why did you even say that, Josh? That's quibbling. That's quotidian. Shut up and get on with data centers. Whoa, whoa, whoa. First of all, don't use the S word. And then secondly. Or the Q word. Yes. Wait, what did I say that's the Q word? Quotidian. Oh. No, that's KW. Right. So the reason that I bring that up, though, Chuck, is because that technically is part of the progression of data centers. Yeah. Probably goes without saying, but it's evolved along with computing.

3:34And as computing's kind of gotten bigger and bigger, the need to store and access more data has gotten bigger. So much so, Chuck, that just wrap your head around this one. In 2024, just over a year ago. Yeah. We used 150 zettabytes of data. That's what we consumed. And consuming is anything from making a video and uploading it to TikTok or putting a post up on Instagram. It's browsing a website. It's buying a song from iTunes. It's doing web analytics. It's buying something with your American Express card. All of that is data consumption. And we consumed 150 zettabytes of data in 2024. Yeah, I don't even know how many Big Macs that is.

4:21I know you name dropped a lot of brands. It should be like the movies where every time you even just say like buy something on your Amex, the bank account grows by like$10. I agree wholeheartedly. I agree that Amex should do that. Amex, there's 20 bucks. I'll split it with you. Wow, you just bought lunch in 1997. That's right. So just real quick, a zettabyte, Chuck, is a trillion gigabytes. Okay, wow. So we consumed 150 trillion gigabytes. Wow. That was 2024. That's worldwide, right? Yes, yeah, that's worldwide. In 2010, we consumed two zettabytes. Jeez. Yeah, so it's growing exponentially, which means that data centers are growing exponentially.

5:05and now they're about to just blow up, like truffle up, essentially, from, you know, this kind of calm, like, plateau that they'd reach that's about to just go and hyperdrive. Yeah, and actively is, and we're going to get to some startling statistics later on in the episode, but Kyle helped us out with this, our rider over in the U.K. He did a fantastic job. He did a really good job, and there's going to be some U.K.-specific things in here because Kyle's always keen, as they say, to throw that stuff in there. Yeah, for sure. Kyle likes to pepper those in. Yeah, of course. And he's not barred from doing so.

5:43We allow it. That's right. So since Kyle, you know, is frequenter of the Wayback Machine, as are all the wonderful riders that we use, they all have the keys to the car, essentially. He jumped in the Wayback Machine to sort of give us a little bit of a timeline on data centers and a bit on, you know, mainframes and PCs. He also left all of his used tea bags in there, too. I don't know if you noticed. Oh, it's fine. You know, you can throw those back in some hot water and they do just a little bit weaker tea. Well, if you put like five of them together, it's like one. Yeah. And Kyle, I mean, that thing was full.

6:18It really was. He drinks a lot of tea. He really does. So if you want to talk about the earliest data centers that you could kind of call maybe a data center, they were, you know, computers. They were electronic computers. Most of this stuff that we're going to talk about early on was military in use. And as you'll see, even the first, when we talk about the UK one, that was supposedly not military, they even loaned it to the military, which was kind of interesting. But these things were built with, you know, state-of-the-art technology at the time, which meant vacuum tubes and, you know, manual switches and plugs and things like that.

6:54And the first thing that we can really talk about as the first programmable electric digital computer was the Colossus. And as we'll see, Elon Musk has now stolen that for his own purposes, that name. Probably because of this, I would imagine. But it was at Bletchley Park, of course, during World War II. And they were trying to, you know, crack into Hitler's messages at the time. And these things were huge. And kind of, to me, the thing that stood out about Colossus, which is a neat little factoid, is that where Colossus was at Bletchley Park, at Block H, it is now the National Museum of Computing.

7:31I want to go to that so bad. When we do that UK-Europe tour next year, we got to go to that together, okay? Okay. Are we doing that next year? That's what we were talking about. All right. We kind of already have promised it. We have to now. We're locked in the punch. Why is my voice so high then? I don't know. Are you practicing for the Alps? Yeah, that's right. No, that'd be a lot of fun. I'd love to go to that. So that was Colossus. Another one about the same time was the ENIAC, Electrical, Numerical, Integrator, and Computer. So that's a quality acronym. Yeah. And it was the first general purpose electronic computer.

8:10And here's the thing. This is technically not data storage yet. It's data processing. Right. But these things, Colossus, ENIAC, you walked up to them and you said, what's the trajectory of this missile if I fire it from here? And ENIAC would go, beep, bop, boop, boop, and then say like whatever a trajectory is described in. Sure. Or Colossus, you'd be like, what is Hilter saying here to Goebbels? And Colossus would say, Hilter is saying that he's a big fan of Goebbels' work, but he's suspicious that the rest of the world doesn't like either of them. And that was it. After that, you'd be like, hey, what was the last answer?

8:51And they'd be like, what's an answer? Yeah, you got to just tell me what's going on with this Hiltor business. You don't remember from our art mysteries of the art world that How Stuff Works already go. Oh, did it say Hiltor? The title of that section was, did Hiltor do these paintings? Oh, my God. That's a deep cut. I did not remember that. Yes. And I think it still says that on that listicle. Yeah. Yeah, I can only hope. It's got to be Hilter forever. All right. So we go into mainframes at this point. And this is like the 1950s, basically, when companies could actually have their own computer.

9:28It wasn't just the military. These were the old punch card computers. And they were called mainframes. It wasn't made up for this term. Mainframes were originally described or describing like what you would house telecommunication equipment and maybe some other sciencey stuff. But it was referencing literally the cabinets that held this technology. and it became known as just, you know, it kind of took over when the computer world started using it as computer only. Yeah, but again, this is like you're a company and this is where you store and process all of your data. And it's in this one room, but it's not going anywhere else.

10:02It's not for anybody else. And you have to physically be in the room to get your answer, process whatever data you're looking for. When the PC came along and then the Macintosh came along, They took that thing and just made it very small so you could put it on all of your employees' desks. And now they had, like I was saying before, their own little data center right there. So if you said, like, hey, what's the, I need to know the Q4 reports, they'd say, go to Debbie's desk. Debbie's the one who's got that on her computer. And you would go over there and say, Debbie, what's the Q4 report? And Debbie would give it to you, right?

10:40There was no connectivity, but you could still, like, do a lot more stuff than you could when you had a mainframe. Yeah, for sure. Which makes mainframes feel, like, really outdated, but it turns out they're, like, totally still in use today. Oh, yeah, absolutely. I do want to jump back in time a little bit because I promised talk of lending the military, basically, your equipment. And that's what happened in 1951. There was a T-Shop chain in the U.K. I don't know if it's still around. Lions, L-Y-O-N-S. And they were the very first company in the world that used a mainframe. It was called the LEO, the L-E-O.

11:16And it was, you know, like what you would think. They handled like payroll and stock management and stuff like that. But there wasn't a lot for it to do at a T-Shop chain, except for those couple of things. And so they calculated missile trajectories like you were talking about for the Ministry of Defense. Exactly. And that actually kind of helped establish like a, I guess, a pay schedule, how people charged for data centers to come. It was like you would charge them for the time that they used it or you could lease it for a month. And that really started to come around when IBM got in the game.

11:55They became like the mainframe leader in the 50s, the early 50s. I think they had a unit that you could lease for$16 ,000 per month. That's in 1952 money. And then as the processors got better and smaller and faster, that price came down dramatically. And then finally in the 60s, they released the IBM System 360, which not only got Apollo 11 to the moon and back, it appears in an episode of Mad Men, apparently. Oh, really? Yeah. You didn't see that, though, right? No, I never did. I just saw a reference to it on the internet. And you knew it was a show. Yeah. But you should look up pictures of it.

12:42It's like those giant burnt orange cabinets with reel-to-reel magnetic tape. It's just— They're cool to look at, yeah. Yeah, I remember we've referenced the movie War Games from our childhood in the 80s a lot, and the Whopper from War Games. That was, you know, at that age, to see the Whopper in action and to see Matthew, I almost said Matthew Modine, Matthew Broderick hanging up his handheld telephone receiver onto a modem to talk to the school computer. It was mind-blowing. Yeah, that phone and the modem made just a really big impression on me. Yeah, and a cool sound. Beep-bop-boop-boop. Should we take a break?

13:22Wait, let me talk about mainframes today, though, because I just want to give a little, I don't know if a shout-out's the right term, but they are still around because they're so reliable, because they're so secure. You can make it so that there's information on those things that you, again, have to be physically present in the room to access. You can put all sorts of different layers of security. So if you're like Visa or you're a healthcare company or you're the Census Bureau, you're probably still using a mainframe because you're protecting information as tightly as you can. But those things are also super fast and can hold huge volumes of computation at once.

14:02Yeah. Or a non-Golden Globe nominated podcast. Yeah. We've got our own mainframe. Yeah. We've got our IBM 360. That's right. What year was that from again? The 360? 64. Yeah, yeah. That's the one. I was just making sure we didn't have the 65 because that was notoriously buggy. All right. We can take that break now and we're going to jump out of the Wayback Machine and venture into the modern world right after this. Beep bop boop boop.

14:53Thanks. Hey, just finished drawing up that quick one page business plan for you. Here's the link. But there was no link. There was no business plan. It's not his fault. I hadn't programmed Kyle to be able to do that yet. My name is Evan Ratliff. I decided to create Kyle, my AI co-founder, after hearing a lot of stuff like this from OpenAI CEO, Sam Altman. There's this betting pool for the first year that there's a one person billion dollar company, which would have been like unimaginable without AI and now will happen. I got to thinking, could I be that one person? I'd made AI agents before for my award-winning podcast, Shell Game.

15:28This season on Shell Game, I'm trying to build a real company with a real product run by fake people. Oh, hey, Evan. Good to have you join us. I found some really interesting data on adoption rates for AI agents in small to medium businesses. Listen to Shell Game on the iHeartRadio app or wherever you get your podcasts.

16:00All right. So we are out of the Wayback Machine. We're making that, we're combining all those old tea bags and making some still somewhat weaker tea. It really works though. Yeah, it's not too bad. It's a combination of Earl Grey and chamomile, all kinds of fun stuff, but not too bad. And now we're going to talk a little bit about when things started to ramp up because it kind of happened in fits and starts. And one of the biggest, I guess, would it be a fit or a start was the internet. Because once the internet came along, every business in the world started using it. And so all of a sudden, you had to have a lot more data storage and bigger data centers and bigger server rooms in your companies, which was, you know, a pretty good thing at the time.

16:44After the dot-com bust, there were a lot of casualties of that growth. But then things kind The ship kind of righted itself. Yeah, and because it was accessible to basically every business now, you didn't have to buy a mainframe. You could lease space on someone else's mainframe, like you're the Ministry of Defense or something all of a sudden. So that led to this huge proliferation. That gave the foundation for web commerce, e-commerce. That's what they used to call it. That's an old-timey term now. But it created the ability for e-commerce to start and flourish. So this data center's scaling up to meet the needs of the Internet and then to kind of give people all sorts of new space and room to come up with new stuff.

17:31That's where the digital economy came from right there. Yeah. All of a sudden you could hop on Webvan and order a sack of groceries. I have a friend who was all in about that. Webvan? Yeah. Yeah, I think I had a friend who was pretty heavily invested in WebMap. But clearly it was ahead of its time. I mean, let's see. It's Postmates. Well, there's several of them now that have succeeded. We'll name them. We'll get$10 each. Well, you just got$10 for Postmates. You've got to split that with me, you know. I will. All right. Cloud computing was the next big jump. When cloud computing came around in the early 2000s, Or do they call that the early aughts?

18:17I do. Okay. I thought I'd heard that come from your mouth. But that is when, you know, that was the real game changer because things were still, I mean, when cloud computing came along, people thought of it, if they didn't look too hard into it, they thought it was just, you know, floating up in the ether somewhere. Yeah. It's still being stored on stuff. It's just not being stored locally. So all of a sudden, things were just going somewhere else for someone else to worry about all that storage. And more importantly, they could link everything together and store a lot of stuff from a bunch of different people.

18:52Right. So now you have data centers not just available to somebody like a huge bank or something like that. Or a British tea shop. Then a bank, yeah, or a tea shop. And then to e-commerce businesses. is now it's available to you and me. So it's really hard to remember back because the world has changed so much. But Chuck, like 2008, 2009, they were giving us like VPN little things that like you could go home and work and like it would never work. I never understood how to make it work. But that was like the very beginning of how you could take your work home with you and work from home and do things remotely like we can now.

19:36Like it's nothing. But this led to the rise of businesses like like Dropbox. Right. So Dropbox goes to Amazon Web Services and say, hey, we want to buy a bunch of your cloud. Right. Which means that they're going to use a bunch of like different servers and different data centers all over the place. And then Dropbox turns around to you and says, hey, if you give me$19.95 a month, you can have one terabyte of data. Right. You can consume one terabyte of data, right? And then hopefully you don't use all of that. So they don't have to pay Amazon Web Services for stuff they didn't use. But you're paying that$19.95 a month rather, whether you use that whole terabyte or not.

20:17It's a pretty smart business model. Would not exist at all if the cloud didn't exist. Yeah. It's funny. As you were talking about 2008 and how quaint that is now, that's the year we started the show. I know. I know. Isn't that crazy to think about? It really is. But imagine like like working from home at that time. It was just you didn't. You couldn't. No, it was kind of great. You went home and you homed. That's exactly right. That was a big difference. Yeah, I remember. But these data centers have now come together in such a big way now that the largest ones are called Hyperscale. And they host more than 5 ,000 servers.

20:54Like servers. Not individual person's data. It's incredible how big they've gotten. And we'll go over some of the kind of square footage and then later talk about the elephant in the room, which is energy and water usage. But Google's first data center was built in 2006, just but two years before Stuff You Should Know launched. And this was in Oregon. And they are still expanding that thing beyond 1.3 million square feet. Meanwhile, in China, they're like, hold my tea, I guess, because China Telecom has a 10.7 million square foot data center in Inner Mongolia. It's 250 acres. that is a warehouse full of whirring servers heating up and being cool.

21:43Yeah, which is a big problem for any data center, it turns out. But the whole expansion, this jump starting in 2017, thanks to cloud computing, because again, cloud computing just means all your stuff isn't on one server and one data center. It's broken up into pieces and spread all over the place. That's the cloud. That's basically it. even though it's way more advanced and intricate than that. That's like all you really need to know for the purposes of this episode, right? It led to a huge jump, a huge need in data centers, and it also expanded all the stuff we can do now. And COVID actually gave it another bump.

22:20Yeah. Made building a data center a very economically attractive thing to do if you had the money because remote working finally, finally established itself as like, no, we're doing this. Stop calling us back to the office. Yeah, which is what they're doing now. I know. I know. I hope it doesn't work because I remember when all of that started and everybody was so nervous, like management was all so nervous that people were just going to totally like mess around and everything. It just, it didn't happen. I don't know anybody who's even been like gotten a talking to, let alone been fired for just messing around at home.

22:59As a matter of fact, like you were saying, it just makes you work more. Yeah. I mean, do you know how many times in our old offices I would see Jonathan Strickland just wandering aimlessly through the office chatting with people? Yes, I do, because he would chat with me a lot. He did. He did that in front of God and everybody, as they say, right there in the office. So I can't imagine what happened with him at home. Yeah, like it was a sorority mixer or something. We love Strickland. He's still around everyone, by the way. He retired from tech stuff, but he's still with the company, which is great.

23:31We love Strickland. All right, so now we're on to AI data centers. And that was the, I mean, to call it a game changer seems quaint compared to the rise of cloud computing and everything. Because it is off to the races in a way that seemingly cannot be stopped. The genie has left the bottle, as they say, starting in 2022 when chat GPT was released by OpenAI. All of a sudden, the need for these data centers became exponentially greater in size and the speed at which they need these things built. Because AI requires a ton of computing power to operate. So much so that they don't even use the standard, what's called the compute machine.

24:18So compute is like all of the processing power, the networking, all of that stuff. And traditionally with a computer that's done on a CPU, right? That's how all of this gets done, right? Everything else is infrastructure. The CPU is doing all of the work. Those are so, like they still work. Like most data centers are running on CPUs. For AI, just not fast enough. They use GPUs, graphic processing units, which are associated with video games for most people, right? You need a good graphics card to run your video game, I guess. But they, the reason that for AI data centers that they use GPUs is because they're really good at parallel processing.

25:02They can run a bunch of different operations at once. So you're like, cool, you just throw a GPU in a data center and you can run an AI. No, you need hundreds of thousands of these things strung together. And instead of like a CPU running like a couple of servers or something like that at a data center, all of them are strung together to form one giant supercomputer that the AI operates on. Yeah, like ChatGPT itself was trained on 20 ,000 of these GPUs. a GPU, you know, the sort of the biggest name in the game. There's a couple, but the biggest one obviously is the NVIDIA, but the NVIDIA H100.

25:42That is the standard right now. If you look this thing up, it fits in your hand. It's not like some gigantic thing. 20 ,000 of them linked together or 100 ,000 of them linked together. Who knows how many, you know, hundreds of thousands are eventually going to be linked together to end the world. Right. That's where all the power comes from, like you were saying. But it's you know, it's just a little rectangular handheld thing that's like, oh, that looks like something that maybe came out of a computer. And NVIDIA is what are their stock jumped over a couple of years, like 900 percent over 2023 and 2024?