The Nostalgic Nerds Podcast
The Nostalgic Nerds Podcast, where we take a deep dive into geek culture, tech evolution, and the impact of the past on today’s digital world.
The Nostalgic Nerds Podcast
S1E15 - Bubbles Build the Big Stuff: How Overhyped Tech Leaves Us Infrastructure We Rely On
Come for the canals and railroads. Stay for the part where Marc and Renee admit AI might just be the next overbuilt mess we rely on.
In this episode, we explore the historical patterns of infrastructure bubbles, revealing how emerging technologies often spark both excitement and economic pitfalls. We analyze past transformative advancements, such as canals and railways, highlighting how they initially fueled industrial growth but ultimately led to speculative investment and significant losses for latecomers.
The Infrastructure bubble pattern has repeated throughout history and we end with a discussion on the latest bubble that hasn't quite burst yet...AI...cue impending doom music.
Notes - Carlota Perez “Technological Revolutions and Financial Capital: The Dynamics of Bubbles and Golden Ages”
https://en.wikipedia.org/wiki/Optical_Carrier_transmission_rates OC-768 is 40 Gb/s
Join Renee and Marc as they discuss tech topics with a view on their nostalgic pasts in tech that help them understand today's challenges and tomorrow's potential.
email us at nostalgicnerdspodcast@gmail.com
Hey, everyone. Welcome back to the Nostalgic Nerds podcast. I'm Renee.
Marc:And I'm Marc. And today we're talking about bubbles. Not the fun kinds, the ones in the bath, the kind where everyone gets a little too excited about new technology.
Renee:Historically, whenever something big shows up, people suddenly decide the laws of economics have stopped applying. Canals, railways, electricity, telecom, the early Internet, EVs. The pattern repeats
Marc:And now we're hearing that same pitch in AI the voice that says this time we have the math trust us which is the moment yeah that's.
Renee:The moment you should start checking the exits honestly
Marc:So here's an interesting twist a lot of those giant bubbles the really messy ones the ones that wiped out entire fortunes they built the infrastructure the world ended up relying on The.
Renee:Money gets spent, whether it's spent wisely or not, and decades later we're all living on top of whatever survived.
Marc:So today we're asking whether AI is on the same track.
Renee:And if it is, is that actually a bad thing?
Marc:I don't know. Do you think it's a bad thing?
Renee:You know, I don't. I think this is the laws of physics when it comes to adopting something new.
Marc:Yeah. I mean, I'll have to put it in the show notes because this came up as a topic on LinkedIn. So I shouldn't pump LinkedIn because it's like really bad. But I'll put the book, I just ordered it into the show notes, you know, basically saying, hey, you know what? Bubbles have to burst. And they're essential. So, yeah, it should be good.
Renee:So, OK, but let's start with something older than railways or telegraphs, canals, because canals were not just ditches with boats. They were the first real attempt to hack geography.
Marc:I really like ditches with boats.
Renee:Okay, so like Indianapolis, like cutting right through downtown is a giant dish with boats. Like that's how they used to move stuff from one end of town to the other.
Marc:We're talking about canals. That's here in the UK is a big deal. And I actually live like two minutes from a canal and a real working lock and canal boats. And you can see them lined up, which is really cool. But the technology was brilliant in the simplicity. Basically, water reduces friction so dramatically that one horse pulling a large barge on a canal could move 30 tons. And it's the first big leap in industrial logistics.
Renee:And those locks you love so much were the real magic, right? Precision engineering in wood and stone, using gravity and controlled water flow to lift and lower boats. For the 1700s, this was cutting-edge mechanical innovation.
Marc:So It's hard to believe, right, that, I mean, from one sense, like Archimedes Principle, and it's been known for hundreds of years, and, you know, everybody gets it. But, like, wood and stone and big giant levers and water raising and water dropping, it's like, like, this is, it is pretty amazing.
Renee:Pretty amazing. Again, everything we've ever done before, we've done again. Like, I don't. Yeah, I know. Yeah, every time.
Marc:Okay, so I don't know. We'll talk about it tonight, but there's this cool, I guess it's a lock. It's called the Falkirk Millennium something or other. You have to check it out. It's this big, giant wheel, and the boats drive into the wheel, and it spins around. It's like a big Ferris wheel for boats. I have seen it.
Renee:I have seen it. Yeah.
Marc:Anyways, early canals make serious money. The Bridgewater Canal is a sort of classic example. It slashed coal prices in Manchester, so it paid for itself and became proof that industry and infrastructure could transform each other.
Renee:Which created the first wave of winners, early investors who backed canals between major industrial centers, industrialists who suddenly had cheaper inputs, and cities that happened to sit on the right routes.
Marc:And of course, the mania are, you know, like after that, you know, some good winners. Then the mania arrives. Dozens of companies are approved. Investors pile into schemes that exist mainly because someone drew a straight line on a map and assumed the law, the land would cooperate.
Renee:Isn't that how we used to talk about data cabling, though? We'd be like, it's only six miles as the crow flies, because that's what you'd have to qualify it that way. It was like, as the crow, you know it's four miles, but it's really nine and a half. Stop it. We have to go through central offices. Right, right. And so, but late investors get crushed, right? Companies collapse halfway through construction. Some canal beds never see a drop of water. It's, and I think this is the most important thing. Any bubble I'm in early, I make money. I'm in the middle. It's too much girth. I'm never going to make any money, but I'm part of the hype. I'm riding the ride. I'm having And a really good time, even though probably going to end up broke.
Marc:Yeah. Yeah, the hype, the hype, the hype factor. But, you know, even through this chaos, the canals that worked changed Britain. They connected cities, they lowered transport costs, and kick-started the first big wave of industrial expansion. So think about, okay, like we think about boats and moving, like shipping containers and boats. It's like that's like that's not canals. Right. But it's it's how big stuff gets moved from place to place. And before canals, it's horses and wagons like like so there's no cars.
Renee:Right. Yeah. There's no rails.
Marc:I know. Yeah. So so big upgrade, massive upgrade, huge impact, big mania because of it.
Renee:And in the record, you see the familiar pattern. We've seen our whole lives. You and I were part of the dot-com boom, right? Early winners, late losers, and consolidators buying the remnants. I mean, let's not forget, at some point, Amazon owned e-toys. At some point, that actually happened.
Marc:Oh, that's right.
Renee:A template that keeps repeating over and over again.
Marc:Yeah. Gosh, I didn't even think about e-toys before.
Renee:Yeah, I mean, that ends up a tab on Amazon at some point.
Marc:Oh, I remember. Oh, my gosh, the crappy tab design.
Renee:Yeah, right? Remember that? And so, like, it was a store. Amazon had a toy store now, and it was eToys because they bought them up in the crash.
Marc:Oh, yeah, I remember that. I do remember that. Oh, yeah. Boy, I wish I could go back and look at it. Oh, I guess I could with archive. Oh, yeah, go to the Wayback Machine.
Renee:Check it out.
Marc:The Wayback Machine. Check it out. The tabs. Oh, the tabs were bad.
Renee:And the privacy policy for Google when it first started. It's so like sweet and kind and happy.
Marc:Okay. Railways. Yeah, let's talk about railways. So once canals proved you could reshape a country with infrastructure, railways arrived and made everything faster, 10 times faster.
Renee:That engineering leap was huge. Iron rails, steam locomotives, manufactured wheels, signaling systems. It was like introducing jet engines into the world of rowboats. Wait, can I just tell you one more thing? I actually did a presentation on this talking about... The history of shadow technology. And I said that actually in the rail industry, there was people made their own signaling tools and they weren't the same. Yeah, yeah, yeah. They would just stand out on the road with like a green lantern and be like, what the hell does that mean? Right. And then they would do all this crazy stuff. Like they would make their own engines. They would put, they would make their own locomotives and stick them on the track hook and maybe they wouldn't run into another one. Like the deregulated life that railways were before we actually said, this is infrastructure and we're going to put rules in place. Up until that point, man, it was a free-for-all and it was crazy. It's crazy.
Marc:Oh, man. Well, I guess that's where the train wreck phrase came from, right?
Renee:Right. And that was the thing you did at the county fairs is they would put two steam trains on a track and run them into each other. Right? And you're like, that's rad. Like, of course we would do that.
Marc:I mean, you know, we got demolition derbies. You got demolition train tracks.
Renee:Right. I guess. Pretty amazing.
Marc:The early lines between major industrial hubs made real money. And I don't think that's like, you know, that's obvious, right? Enough that ordinary people, shopkeepers, clerks actually started buying railway shares. And this is where it becomes not so obvious.
Renee:Right. This is when we all go in on NVIDIA when it's like $300 a share.
Marc:Oh, yep. Exactly. Yeah. Right. Well, did you see did you see last week? It was they had decent report, you know, or I guess you could interpret it as a decent report. It spiked it. But I think it dropped again.
Renee:It did today. Like today. Yeah. It's super volatile. No one can make up their mind whether or not they want to get on the train. Right. And Parliament approved new railways in an unbelievable way. Pace again it's not a ton of regs so go ahead do whatever you want more than 9 000 miles of proposed track in a short span it was railway mania and it was taking over the world for sure yeah
Marc:I think the uk was an early adopter you know for this canal mania but also the railway mania and there's a whole we probably should do a railway episode once because there's a whole like you know, gauges and, you know, widths and interoperability. And that would be kind of a cool... I'm someone who.
Renee:Clocks a lot of hours on Amtrak. I'm more than happy to have that conversation with you.
Marc:We should do that. My dad was big into model railways, so...
Renee:My grandfather worked on the railroad. My dad worked for...
Marc:Yeah, exactly.
Renee:The railcars at Port Authority.
Marc:Yeah, yeah, yeah.
Renee:Murphy's come from that whole line of rail workers.
Marc:Let's see. The early winters were obvious, right? Manchester to Liverpool, London to Birmingham, the UK in particular, the arteries of commerce, right? You also have obvious corridors in the U.S. As well, the Transcontinental Railway and that.
Renee:Yeah, but late-stage investors didn't do as well. Many companies underestimated engineering challenges, overestimated traffic, and mispriced land acquisition. Because remember, if you want to drive that train through the middle of Arkansas, you've got to buy all that land that cuts through the middle of Arkansas, right? Right. Yeah. And imagine you thought, oh, I'm going to do, I don't know, wherever the track is to Boise. And it turns out there's not a ton of people in Boise. There's no one to buy tickets and there's nowhere to go. It's like building the train to Deadwood. Why would you bother?
Marc:Why would you do that?
Renee:People did, though. People, you know, it's again, this is why we can't have nice things, because someone tried to sell somebody, you know, stock in the railroads. Like, now you're like, I'm going to put a railroad everywhere, buy my stock, right? Like, it's just.
Marc:Yeah.
Renee:Yeah, yeah, yeah.
Marc:There's the monorail episode of The Simpsons, right?
Renee:Right, yeah. Why do we need a monorail in a small urban area that is already serviced by, you know, public transportation? Like, explain that to me. And he's like, shut up, little girl.
Marc:Yeah. So a huge number of railway companies go bankrupt, a third of them by some counts in the UK. But the interesting thing is that the winners at the end were not always the early investors. So this was actually, you know, sort of a different scenario where, you know, early doesn't always mean, you know, tons of cash. The consolidators, the ones who bought the bankrupt lines cheaply became the giants.
Renee:Keep holding out. That's going to be Ford. Like Ford's just going to wait till like Tesla implodes on itself and then they're just going to go buy it. Like, like, why wouldn't you like it seems like they have pretty good engineering, like or you wait for Waymo to Waymo to like like go sideways somehow because of bad investments. And then you swoop in and buy it up. I mean, in my mind, like you're not the barbarian at the gate. You're just the guy who's going through the dump looking for, you know, the thing that you think is cool. And if that's a sofa, by God, you're taking it. Right. Yeah. Yeah. Yeah.
Marc:I mean, like especially with rails and you see it in the, you know, I didn't talk about it here, but in London in particular, the underground transport is happening at the same time as all of the sort of commerce transport. And each one of these little places, the Bakerloo and this line and that line, they were different rail companies. And you don't have a unified, successful system until they're all unified. They're all consolidated under one operator with the same track gauges and the same standards and the same signaling and the same infrastructure to support it.
Renee:And then the rails that they built, even the ones that ruined companies, became the foundation of the modern transport network. Factories moved to junctions. Cities grew along the lines. Markets unified the country. And I guess that's it, right? That's it in a nutshell. If you ask why do you need a bubble, you need a bubble for this. Like we can't turn it into reliable infrastructure if we don't fund the building of it, the overbuilding of it, and then the consolidation of it. And then with that, no matter what bubble we're talking about, with that, we've created infrastructure. And I can't, number one, somebody has to fund it. Number two, it's almost never going to be the government, although they'll help you do some of it. And three, you know, just while you're funding it, like right now, if you have a ChatGPT subscription and you're paying $19 a month, don't buy their stock. You're doing your part. You are doing your part You don't need to own Another thing from them
Marc:There I said it There you go I think we'll talk about chat She can see an open AI later But like I mean, come on, you know, there's some bad numbers behind this stuff. Right?
Renee:Like, you're doing your part just by buying it. Just leave it at that.
Marc:Exactly. There you go. Demand, you know, for Rails, it was real. But the investment behavior was not. That mix that creates these bubbles, it also accelerates the infrastructure. If you think about it, like, If you don't have these accelerated periods of growth, then the time that it takes to actually implement something that will be useful as a bit of infrastructure is maybe it won't even happen, right? Maybe it takes too long or maybe it won't actually be implemented at all. Like, if you think about cars and rail, like, there's enough separation that perhaps rail would have always been a good thing and a good piece of infrastructure. But, you know, think about the delay of rail infrastructure in the states and then the development of highway infrastructure in the states and how that shifted, you know, from a place like, you know, Europe or Britain, you know, were very dependent upon rail to the states were very dependent upon cars.
Renee:Yeah.
Marc:You know, and and that was a huge infrastructure play. We didn't I didn't put it in here with the notes and stuff. But, you know, think about the infrastructure around cars and the highway system. I mean, Eisenhower basically said, build it.
Renee:Yeah. Right. Yeah. I think you see a couple of times in history where people are like, it's time. Like, like we wouldn't have cars today if the government hadn't built gas stations. I mean, they're the ones who did that. Ford was too busy building cars to build the gas stations that they were going to rely on. So, yeah, and the government felt like, OK, well, what can we do? We can aid with the infrastructure. That's what we'll do. Gas stations. We'll do that. We'll subsidize that. Right. But the next shift isn't about moving stuff. It's about moving information. And the telegraph separates communication from transportation for the first time. Right. Because up until then, it's the Pony Express or your stuff's being delivered by train. You're like a letter from home is you're writing them every day so that you can get them every day because you can't make a telegraph call. Right.
Marc:Yeah, yeah. Railways brought speed for goods. Telegraphy brought speed for thoughts, messages, signals traveling through the copper at a fraction of light speed. I looked this up. It's about 0.6 light speed through copper. Suddenly news moves in minutes instead of days.
Renee:Early telegraph companies had every advantage, right? They controlled the long-distance communications. They set the prices. They were the clear early winners.
Marc:We haven't talked about this. I can't remember which episode it was. Was it the ethics one where we talked about the Western Union and the control of information and how it shifted the 1876 U.S. election, right? So the speculation around telegraphy kicks in and small companies start to spring up everywhere. Some build lines that work and others build lines that actually never function at all.
Renee:The losers are predictable, though. Underfunded companies, poorly surveyed lines, installations that use cheap material or bad insulation, systems that could not maintain signal integrity. So while everybody wants to give you a calling card, very few could connect that call.
Marc:But like some of the earlier bubbles, useful parts actually survive. The industry consolidates. The strong operators absorb everything. Western Union becomes the dominant inheritor in this telegraphy system.
Renee:And society moves into a new footing. Financial markets synchronize. News becomes faster. Railways coordinate schedules to avoid collisions. The world becomes a smaller and more predictable place.
Marc:Do you ever see those early stock tickers? Have you ever seen those?
Renee:Oh, yeah.
Marc:The ribbon. Yeah, the ribbon. And they're underneath the little glass dome.
Renee:Yeah, right.
Marc:Yeah. That ran off of telegraphy.
Renee:All day long. And it started from the spool to a trash can because you only cared about it as it was coming by.
Marc:But all of that is built on top of the bubble that burst and then left behind copper skeletons that, you know, essentially are kind of too valuable to abandon themselves.
Renee:No, I mean, when I think about the telegraph, I feel like, you know, This is what happens. It got us used to stuff fast. I have information fast, right? Once we were hooked on that, that dopamine rush of I know fast, everything that comes after that is in keeping with I want it fast. I want it fast. I want more efficiency. I want it faster. I want more efficiency. I want it faster. It doesn't matter what it is. I don't want to buy oil, put it in a lamp and light the lamp and keep an eye on it to make sure it doesn't burn down the house. I'm going to flip a switch, baby. I want electricity, right? So now we get to the part where the story starts to sound familiar to us because the telecom boom in the late 90s is where you and I enter the workforce. We come into this just as, oh man, voiceover IP is starting to come around. We're going to avoid certain networking technologies because others are taking its place. so it's point-to-point networks are gone, frame relay comes in. Yeah, we're part of the story now, like this is where it happens for us.
Marc:Yeah, I think, you know, fiber, and so I worked for this test and measurement company, and they were, they built up all of the equipment that basically tested all these fiber technologies, but, It's like this irresistible thing. Copper has these physical limits on interference and bandwidth. This is why I looked up the speed of light through copper versus the speed of light through fiber. And actually, it's not that much different. It's like 0.6 to 0.65 or so in copper. And at max, it's like 0.7 or something like that through fiber. So not a lot of difference between them. But it's a width. It's a width issue. But fiber solves these limits. You know, it's not the speed limit, it's the width limit. With medium and low loss and huge spectral potential eventually.
Renee:I mean, if you ever worked in an office, if you're listening to my voice and you've ever worked in an office, you probably had a multiplexer that was bringing in phone lines, right? And so those phone lines would come into a gigantic digital switch, but they were analog phone lines. So we would bring in, let's say I had an office full of maybe 100 people. I would bring in 22 lines because I know 22 lines. At no point are 22 people all at the same time going to pick up the phone. Like it's just not going to happen that way. And if it does, we bring in another trunk. But like that's how we used to think about bandwidth too, right? You would use multiplexers. So let's say I need five terabytes or I need five megabytes of throughput in my data. Well, now I got to put in four T1s, right? And I'm going to put them through a multiplexer so I can pull out as much data bandwidth as I have, forgetting the fact that I'm going to have overhead on the data, right? So we really did have a problem. Like you think today, like, oh, I have 5G. Yeah, good for you. Do you think life was that easy back in 1992? No, I'm here to tell you it was not that easy. I had an antenna on my phone that I had to pull up when I used it. Don't talk to me about it. Do you remember your cabling dude,
Marc:Gordon? I love Gordon. Gordon's like my idol. And he would write his little notes on his thumbnails, you know? That guy, I knew exactly how many lines to bring in and, you know, where to line them up. Are you talking about the teamster.
Renee:Gordon? That guy?
Marc:Yeah, yeah. The gray overalls.
Renee:Yeah, yeah, yeah. Okay, we're talking about the same guy.
Marc:Yeah, that guy's awesome. Yeah.
Renee:And as many, for every Gordon you met, you never met a Gladys. I met one woman that did that kind of work in the 90s. One. And she definitely needed a manicure. Like, her hands were so terrible. Like, I'm like, oh, girl, you're never going to get a date like this, right? Like, I felt really bad about it. Anyway, people described fiber as fast, but the real breakthrough was in the bandwidth. You could push enormous amounts of data through it without worrying about electromagnetic noise or attenuation, right? Because if you put a signal through copper, eventually you're going to have to boost that signal. Otherwise, it's going to degrade. We didn't have the same problem. And we could go miles and miles and miles and miles and miles on fiber where we could go feet, maybe hundreds of feet on copper.
Marc:Yeah. What was, what, Cat 5 is, what, 100 meters or 100 feet? Yeah, something like that. I don't know. It's something like that.
Renee:And Cat 5 was as good as it got, dude. Back then it was like Cat 5 was.
Marc:5e. Ooh, 5e.
Renee:You're talking to someone who used to custom make her own cables, dude.
Marc:I know. I know. I know. I did too. I did, too. Remember, like, Cat 3. Like, when Cat 5 came out, it was like, oh.
Renee:And the oldie but goodie RJ11 for your phone.
Marc:Oh, RJ11. Yeah. RJ11. Well, you could just run the RJ11 off the pair, the brown pair and the twisted pair. Or, yeah, off the Ethernet. So anyways, everyone assumes demand is going to keep rising. So companies laid fiber under oceans, across continents, through cities, billions spent on capacity they believed would be used any day now. And like the WorldCom CEO was saying things like, capacity is doubling every month, you know, or I don't remember what he said every month.
Renee:And Ron, we're going to commoditize it all, sell it at low rates. Like, yeah, it was really crazy talk. That that i remember that time specifically because of enron enron being like you don't use data at night why don't you sell it to a different part of the world where they're where they are using it and i just kept thinking like how what like that just makes what you do a scam because like now i just know it's all a scam because if this is what you're doing with oil and gas like this is clearly not going to work out like commodity
Marc:Commodity trading on data packets like what.
Renee:It was it was the nuttiest thing i'd ever heard but that's just But that's the bubble, right? This is the bubble. It's the telecom bubble that we're talking about. And everyone wanted in on it, including gas trading platforms.
Marc:I know. Quest. So I was talking to my daughter today. And I was like, oh, yeah, we're going to talk about this. We recorded tonight. And you might know some of these companies. And she's like, I don't know any of these companies. And I named a bunch. Level 3, Global Crossing, MCI, WorldCom.
Renee:You're taking me way back.
Marc:Yeah. Quest. BBN Planet. Like Sprint. And I explained to her, do you know what Sprint means? And she's like, yeah, Sprint. You know, you run or it's a Sprint in two weeks and you're a developer. And I was like, no, it's Southern Pacific Railway Internet.
Renee:Because they would run the cables along the railway.
Marc:There you go. There you go. Back to railways. Back to railways. Yeah.
Renee:Yeah. So early winners included backbone providers and high-traffic corridors and companies that owned rights of ways others needed, right? Yeah. Well, yeah, Sprint. You know, MCI would tell you, you know, we leased to everybody else. Even today, I would say that T-Mobile isn't really – it's a leased network.
Marc:Oh, yeah, yeah.
Renee:Yeah, T-Mobile is a marketing thing that sells you phones on a network that doesn't belong to them.
Marc:Yeah.
Renee:They just take less of a margin than the people who own the network, right? Yeah. So, like, they sold capacity to the first generation of internet businesses. Yeah.
Marc:Yeah.
Renee:Were you an MCI customer? I was an MCI customer.
Marc:No. No. No. The West Coast. West Coast was, like, Pac-Bell and then SBC. And then SBC got sucked in.
Renee:So, MCI used to do this crazy stuff where they'd say to you, outsource your routers and switches to us. and it's managed services, right? And we'll do that. So News Corp decided that's what they wanted to do. They were going to take that out of the hands of the IT people and they were just put it right before MCI goes under, they're going to put it in the hands of MCI, right? And so we go through this big issue of getting all of those routers and passwords and everything into the hands of MCI so they can continually monitor that stuff. And it made you wonder like, you know, they were so desperate for income. You never see it that way in the bubble though, right? You never in the bubble think to yourself, Why do you want this business? Aren't you a long-distance company? Like, why do you want this business, this managed services business? And you just realize they weren't making any money. They needed to do something. Yeah.
Marc:I mean, do you remember UUNet? You know? UUNet.
Renee:That sounds familiar.
Marc:You see, like, these companies, like BBN Planet, UUNet, like, these companies were the very first backbone providers. And they just don't even exist anymore. They're gone. They're gone.
Renee:Yeah, they're gone. Yeah.
Marc:Anyways, the mania spreads, right? Firms build these giant redundant routes. Some ran parallel networks down the same highway, competing to be the first destination. So I was at a Quest data center once in Burbank. And I can't remember. Like, they were like, oh, we've got 16 pairs of OC, which is optical carrier, 768, which at the time was like, yeah, which is the widest of the wide.
Renee:Right? Do you drive a truck down that?
Marc:Bandwidth yeah i can't i i think if i recall correctly 768 is like 40 gigabits something like that it might even be more than that maybe 192 is 40 but i mean it's big like six 16 redundant pairs and and they're like well yeah we could carry the whole at the time and this was in the was like 2002 like well we could we could carry the entire traffic of the internet on on half of this this router right here and it's like talk about freaking overbuilt buddies and.
Renee:We got more efficient like that's what is that moore's law that says like the longer this goes on the more efficient we get which brings me to why the data center bubble's eventually going to burst too because we're just not going to need this much square footage in raised flooring yeah we just don't need it late investors suffers when the revenue never arrived many networks sat dark entire companies collapsed world calm global crossing level three massive massive failures
Marc:Yeah okay and then and then so people are laying fire like left and right and then comes dwdm dense wavelength division multiplexing okay so basically lasers are light, And fiber is glass and you shine the light down the glass and, you know, you think, oh, well, I'm shining a light down it and I get one laser. Well, guess what? Dense wavelength division multiplexing means that you can shine lots and lots of lights. You know, I think it's millions. It went from wavelength division multiplexing was like 16 to 16.7 million or something like that. I didn't do enough research on it because it was like old Marc days, so I didn't go back into it. But essentially, you overbuild, and then the technology shifts, and the capacity suddenly— And now you're really overbuild.
Renee:Now you're really, really overbuild.
Marc:Totally.
Renee:Right? We don't need more fiber. We just need better optics now. The infrastructure's done. And those of us who paid crazy amounts of money to get it— Yeah. We funded that. We funded that.
Marc:Yeah. We funded that. Yeah. Well, and the government paid for it. And then, well, anyway, so that's a whole other.
Renee:Right? So you didn't need more fiber. You just needed better optics. And that's how we were going to grow going forward.
Marc:Yeah. So overbuild, you know, this overbuild perspective, it makes, you know, even worse from an investment perspective. But from a societal perspective, it created a massive surplus that ends up powering cloud computing, streaming, mobile apps, remote work, IoT, everything that comes after. Yeah, you're welcome.
Renee:You're welcome, millennials. If it wasn't for the fact that we were overpaying and overengineering, you wouldn't be staring at your phones wondering if Twitter's real. You're welcome.
Marc:I mean, we're still living on the surplus. We're still living on the surplus. Yeah.
Renee:So the winners in the long run were not the companies that built the fiber. The winners were the companies that arrived later and used the excess capacity. Don't lose sight of that. It's not the ones who started it. It's the ones that come at the end and say, yeah, you know what? That is a good idea. Here's what I want to build on it. Yeah. On it. Because it's infrastructure.
Marc:The capture of value, right? That's the concept. It's like, okay, you spent all this money, billions upon billions of putting it into the infrastructure. The guys that built the infrastructure didn't capture the value that it generated, right?
Renee:Facebook, Google, Amazon. They were just thinking, I got to move packets. How fast can I do it? God bless them. Bell Labs. God bless you guys.
Marc:There you go.
Renee:Right? Which brings us to the moment you've all been waiting for, I'm sure. The conversation about AI. the latest attempt to build infrastructure for something fundamentally new.
Marc:Yeah. So the scale on this is enormous. Data center spending is exploded. And when I say exploded, it's, I mean, we talked about this in our very first episode. It's orders of magnitude bigger. Companies are competing to buy GPUs at prices that would have seemed unrealistic five years ago. Huge demand, low supply, massive payouts.
Renee:And the early winners are exactly who you'd expect. NVIDIA, the makers of the tools required to run the workloads, they're the shovel sellers in a gold rush.
Marc:Yeah, the early labs are also benefiting here. And, you know, Google has built their own chips, right? Because, of course, why wouldn't they? They built their own servers in the 90s instead of buying, right? Buying servers, they build custom servers. Companies that train the large models before compute scarcity hit gain, a kind of first mover advantage.
Renee:But then there's a sense of panic buying. Companies build new data centers before they fully understand their own demand curves. Model training projects that cost billions without a clear monetization path. I mean, it really, AI feels like a bubble because it feels a little like the Wild West out there. It's like, we're not sure what it's going to do, but it's promising. Like, that's it. That's what you guys are investing in. I'm not sure what it's going to do, but it's going to be promising. Let me tell you what it's going to do. It is going to be the infrastructure of the digital transformation that we're about to make. And yeah, who's going to consolidate it? Who's going to own NVIDIA in the end? That's the question.
Marc:Yeah. Yeah. I mean, that is a good question. I don't know, man. I don't know. I'm not sure. I'm not sure there is going to be like this kind of the chip thing. Yeah, I don't know. There might be. There might be. I mean, certainly the big players are.
Renee:I feel like every time we have a talk about OpenAI or Claude or Gemini or Gronk, like whatever is your model du jour, right? Like whatever that is, that's all going to become one thing. Like it's going to become one thing. It's going to be owned by someone, right? And then we're going to probably Verizon, if I'm being fair, or Oracle. And then we'll turn around and say that's the foundation that we're building on. But I'm going to warn all of you who think you're going to buy all that up and be the last man standing when it comes to the large language model, you and only you will own the regulatory landscape for that. So be careful what you wish for, because it will be regulated. Because if there's one thing that comes with the commoditization or the platforming of kind of an infrastructure, it's the fact that regulation soon follows.
Marc:Yeah i know you're right i think you're right to call out that risk right the late entrants are at the highest risk there and the firms that are planning to build their own models from scratch i think you know this is why would you go it's just it's like data was like 10 years ago or 15 years ago right we're gonna build a big data thing it's like why would you do that why would you do that now because you're just gonna go like just use google's data tools in the cloud because they're better than what you can build off the shelf. So teams that are investing in racks of GPUs with no assurance those models will ever generate revenue are in trouble. Cloud providers are taking on enormous capex to build this stuff out.
Renee:And again, we as corporate America who have adopted the cloud as a better way to run the infrastructure of our own corporate entities, we are paying for that. We are paying for that. Oh, and by the way, you, the person who's listening to my voice and your electric bill went up by 25%, you too are paying for that because the data centers are pulling electricity from a grid that you're being charged for. So yeah, the question becomes whether we're building ahead of real demand or simply hoping that the demand appears because we've committed too much money to back out gracefully. I don't know. When has that ever worked? Like, like dot com, like, dude, maybe maybe we shouldn't have one more store where you can buy stupid stuff that no one cares about. Nope. We need three more of them, because if there's not a lot of them, then we won't have the traction to make it normalized. Right. So in a lot of ways, again, that ballooning in the middle, Like, I swear to God, we need it. We need it. Otherwise, when it bursts and the puddle that's left over that everybody's trying to like drink out of, like that is like that's the end. But that middle thing in between where we're paying for our usage and we're paying for access and it's a crazy amount of money. And, you know, all of that is required if we're going to move. If we're going to evolve in technology, all of that is required.
Marc:I think you're right you know and it mirrors this the telecom bubble but I think it mirrors the the e-com bubble as well right the dot-com bubble and, You remember, like, cooking.com version one, right? We had, what, 15 web servers?
Renee:Right, and three backend servers. Like, it was almost nothing.
Marc:Yeah, I mean, so there's this, you know, three sets of paired, you know, databases or whatever. Yeah. And then, you know, when we got those racks and racks and racks of one-use servers, I think we built, like, 200 web servers.
Renee:Uh-huh.
Marc:You know, and that infrastructure literally kept that site going for years before they got acquired.
Renee:Yeah, I mean, Christmas, like any of those really important holidays, like we never had performance issues with them. And just so we're clear, we would turn most of them down so we weren't spending a ton of money on. So like low periods, we would just shut off half the servers and walk away. And like we would only run the site from the other half of the servers. But yeah, I mean, we were ready for it. And that's the other thing, right? Like when we talk about the cloud and why you bother going there, it's like there's something a web developer never has to think about again is the fact that I might not have enough CPU power to actually have 10,000 consecutive users. Like we used to have to do that math and build out that infrastructure. And now they just dynamically expand into that space and retract from it when they don't need it anymore and they don't do anything. Like you guys better never complain ever, ever, ever, ever. Like, I'm on to you guys now. I am on to you. Like, you don't even know how bad it was. You don't know what it was like to be laughed at because you were using front pages extensions. Like, just don't even go there.
Marc:Don't go there. You know, drawing the parallels to this, you know, I think it makes sense. But there's a bit that's like it's sort of stuck in my head and it's a bit. Like, it worries me a bit. Like, compute depreciates faster than fiber, right? You can't stick a GPU in the ground and use it 20 years later. That's the thing that worries me, right? And I was talking to somebody the other day. It was like, okay, yeah, there's going to be an AI bubble and it's going to burst. Okay. But the infrastructure that's laid out here, can you use it in five years' time? you know will the gpus.
Renee:That's just it no yeah like like getting past quantum is going to change everything right like just just look at what quantum is going to do to hardware and then ask yourself like oh man should we be building on that like you know what i mean like yeah yeah
Marc:But quantum quantum is not good at algebraic tensor math right so this so it's a fundamentally different use case. So I don't, yeah. But do you.
Renee:Think we'll have the compute power? So even at that, the data center space that I have and the amount of energy I'm pulling, and I'll go back to when we were kids and we were responsible for buying tons of hardware. God only knows why they would put that in the hands of kids, but they did. And, you know, we would sit there and say, you would hear it from Cisco, shout out to Cisco, over and over and over and over again. Like, you know, why do I need to replace this Whatever it is, switch, router, firewall, whatever. And it's exactly the same. I know how it works. It hasn't broken down yet. It doesn't have any performance issues. It's only three years old. Why do I let it off the budget and take a new one? And they would look you right in the eye and say, because this one is 35% more efficient in power. And you're like, because if you got enough of those, that matters because that starts to cut thousands of dollars off your data center bill. So like you would invest in that just because the total ROI of that was so much cheaper. So that's where we're headed. So when I say we're building all this square footage, these hundreds and hundreds and hundreds of data centers that are popping up all over Texas, southern states, like this is how they think they're going to make money. They're going to end up with things that are 30 times the size of a Walmart and just as empty.
Marc:Yeah, I know. I know.
Renee:And you gave them the infrastructure. You gave them the tax breaks. You gave them the square footing. You probably built the building for them. And then you put the electric bills on the backs of the community around them. For what? For what? They're going to be mostly empty in five years.
Marc:Bubble or no bubble.
Renee:They're going to be mostly empty in five years.
Marc:No, I'm with you. I'm with you. I just, I think that's the. That's the thing, right? You know, how much compute, and it feels almost a bit like ATM versus Ethernet too, right? Yeah. You know, ATM was built and it's more efficient as a transport mechanism than Ethernet ever is or was, right? But it can't handle, ATM is, it can't handle peak, right? Right. It can't handle bursts. So Ethernet, you know, wins out because we're a bursty, peaky, you know, infrastructure. You know, the things humans do demand lots of peaks and valleys. And it feels like, you know, AI development is sort of in that space where the peak usage is when you're doing these big, you know, model development and training. But you don't do that every day. You know, that's not happening all the time. Um, so, so, so.
Renee:You know, then you'll have the conversation with like a research firm, you know, I'm, I'm a research analyst. And so, you know, I've, I spent a lot of time in these places. And one of the last places I was at, they were saying like, we're not going to have our own model because the, the cost of training the model is so great that it doesn't make any sense for us to do that. Like we can be part of other models. We can try for a small language model. We can look at how we can leverage it against our own data sets, but we'll, we're never going to build our own model because it's cost prohibitive. And then you think to yourself, but you're a research firm. Like if anybody is suited for this kind of thing, it's you. Like, I don't understand. Then you take it, right? But this is what it comes down to yet. The incentive structure remains, right? So compute depreciates faster than fiber, right? I can't stick GPU in the ground, like you said, and use it 20 years later. Yet, the incentive structure remains. Nobody wants to be the company that has underbuilt if a breakthrough happens. And fear drives investment more than rational market sizing. If we don't invest in this, we're never going to get to Mars. You're not going to Mars anyway. I hate to break it to you.
Marc:Well, I mean, don't let, you know, Musk hear you say that, but... Yeah, I agree.
Renee:You ain't going.
Marc:But this is why it feels like another infrastructure bubble forming, right? Not because AI is fraudulent, but because the scale of the bet is extraordinary compared to the revenue it currently produces. In order for OpenAI to meet the numbers and make sense, their revenue has to hit like $78 billion or some stupid number.
Renee:Yeah, see? It's crazy. It's crazy. Yes. And when you start doing the math like that, you start to think to yourself, there's no way on God's green earth that's going to happen. And there's four of you vying for that market. There's no way. Because that means a total addressable market for you all to exist is in the trillions. And it's just it's not there.
Marc:It's not. It's not.
Renee:No. More people are interested in playing video games and using AI. Like and if the AI happens to use the video game, so be it. But that's not what this is about. Anyway, if there's a bubble, the real story ends up being who inherits the remains. Every cycle has companies that quietly win the long game. You want to hear who won the long game in dot com? You guys use it all the time. Every day it's Amazon. I remember. Yeah. When they just sold books. Yeah. I remember that.
Marc:Okay. My kids didn't even know that that's what it was.
Renee:That's all it was was a bookstore.
Marc:Bookstore. And you know what? Do you remember when you bought on Amazon because it was cheaper to buy books on Amazon?
Renee:Yes. Because he would buy them in bulk and stick them in his garage and then sell them back to you. Yeah.
Marc:But if you think about it, okay, so I know a little bit about books distribution and publishing and all of that.
Renee:Right.
Marc:And at the time, you know, the markup on books across the industry is 100%. So if you flip the book over and you look at that and it says 1095, then the cost to the seller was half of that. Exactly. Like that's how it always worked. And so Amazon buys them, you know, in the same distribution channel everybody else does. But they don't have the same costs. And so what do they do? They cut the price. Below, you know, that, that number that's on, on there. Everybody's up in arms. Everybody freaks out. But what happens? Retail bookstores close. Yep.
Renee:No more borders.
Marc:There you go.
Renee:Your kids have never been into borders.
Marc:That's right. But, but then what happens over time is, oh, the price of the book goes back to the normal, you know, sticker price and their margins are higher.
Renee:And their margins are higher. So, yeah. So if you guys are wondering who, who won, who came out on top but at the same time i don't use just amazon there's a bunch of sites i use so it in becoming infrastructure it changed the way we work it continues to push the boundaries we expect free shipping we expect stuff delivered same day we expect they keep pushing those boundaries for us to keep us accepting the idea that that's how we should shop i think like yeah that is like that they quietly win the long game and the next thing you know that's infrastructure yeah
Marc:Yeah, yeah. So, railways still use right-of-way, originally surveyed in the mania from ages past, right? Western Union inherited that telegraph infrastructure. The modern telecom landscape sits on top of fiber routes laid by companies that no longer exist.
Renee:Level 3's remains sit inside Lumen. Oh, remember Lumen? That takes me back, too. That's what came after Bell Labs. Sprint Spectrum ended up inside T-Mobile. Google, Amazon, Meta, and Netflix built their businesses on cheap long-haul fiber, brought after bankrupt estates. So, like, yes, right? Like, that infrastructure, that failed idea of commerce, right? What we would call today an abject failure, MCI, Level 3, all of that. No, it just became so commoditized and so much infrastructure that it's not the point anymore. Now it's how do I create value on top of something that's cheap and plentiful? So I'll ask you, how are you going to build something on top of something that's ultimately going to come cheap and plentiful? AI and models.
Marc:Yeah, that's right. Yeah. Well, I mean, I am building something. Oh, there you go. There you go. But if this AI bubble pops, the most likely winners are not going to be the late stage model builders. It's going to be the companies that make use of the compute that survives, right? Value capture. Who is going to take advantage of the infrastructure that exists?
Renee:So just like the cloud wave, the cloud wave lived on fiber someone else overbuilt, the AI wave may live on compute someone else overspent on. So, you know, those NVIDIA chips that everybody keeps building out those giant data centers for, one day, Marc and I are going to own a corner of that data center. We're going to pay 29 cents a day for it. And on that, we are going to build genius. But that's what it comes down to, right? Like, that's what I'm waiting for. I'm waiting for everybody who is, one more time for my public service announcement, this is Renee Murphy telling you, if you pay for a model, you've done everything you need to do to support that model financially. Please don't invest. because you're not going to be the one that's going to be able to consolidate it on the backside when it bursts. So don't, don't just don't, don't be part of that. Yeah.
Marc:Don't be part of that. But also, you know, unfortunately I think because, because it's wall street is crazy. Um, and they act irrationally. I mean, there's, we talked about this, you know, off, off script last time, just off, you know, B B roll or whatever it was. There's derivatives on these infrastructure plays. There's massive bets, essentially, on all of these different hyperscalers. This is what scares me. It's like, no matter what happens, yeah, okay, don't invest. Don't be a late investor. But because so much of the financial markets are tied up in these huge bets, that we're just going to be in another 2008 all over again. And that scares me.
Renee:I think it's interesting how much money is being traded back and forth across seven companies, right? And then in the end, are those two seven, are they too big to fail? Is Oracle too big to fail? Is Facebook too big to fail? Like, I hope not. I hope they get bought in a fire sale by somebody else. I don't even care if it's Verizon. And I'm not particularly nice to Verizon on any given day. I don't care if it's them. Like, I hope not. I hope NVIDIA is not too big to fail. I hope, you know, at no point should we be thinking about it that way. Like, AI is going to do what AI is going to do. The value is going to be built on top of that infrastructure one day. It's just not here yet. It's not here yet.
Marc:Yeah, we'll see. We'll see. So, you know, no matter what, society ends up with more infrastructure. The question is who pays for it and who's going to benefit. So that's what we're hoping.
Renee:And as always, the answer is rarely ever the same group, right? That's right. Right. Yeah. Yeah.
Marc:Yeah. This is like, just think about that. Like. From zero to like major player, open AI, like it didn't exist. And then all of a sudden it existed and shifted the whole cultural conversation.
Renee:So here's what I'd say. The government had AI probably for 25 years. Definitely IBM had Watson for at least 20 years, right? And so while they're letting Watson, you know, go off and play chess grandmasters to prove it was smart, they were keeping it out of the hands of you and me right like for me open ai was the democratization of something that had been around a really long time and we were finally getting access to it and so i think that's part of the reason why it took off so fast we hadn't had new technology in such a long time we hadn't been able to play with that stuff in a way that was really good we were chomping at the bit to get past really really hard analytics like this is easy analytics like we had to get past the I need a data scientist degree to actually use analytics like we're past that now and so for me like maybe that's why I think this bubble it's still gonna burst but maybe it doesn't burst as fast right because like in on one hand that democratization showed you and me the world of the possible and I'm not so sure you or I are willing to let that go I'm not I just wanted to see it turn into infrastructure I just wanted to I want it to be over I want the bubble to burst I I want it to be bought up. I want there to be one model, and it belongs to somebody who already bills me, and I'm good. Netflix, buy it. Anybody who already bills me, I'll pay you more money for it. We'll call it a day.
Marc:It's really interesting to see, because I was playing with this, my LinkedIn BS detector thing, and so I used Anthropic to help me with that. And some with the code, but some with the actual model piece itself. And in order for the model to respond, you know, I have to basically, you know, throw it into Anthropics API and get a response. And I got to pay for that, you know, and it's not cheap. It's like 10 to 15 cents. Per per call and i'm like well that's man that the economics are dumb then you know that's like that's expensive for basically for me to like put it i would have to pay 10 to 15 cents on every single linkedin thing that it wants to analyze and that's like whoa wait a minute that's expensive.
Renee:So maybe there's an opportunity for government to step in and say you don't get to make money that way right like at 15 cents a call like that's not how a digital world should work essentially right every time i pick it up and get a dial tone i owe you money like so if if you are going to call it infrastructure and it's going to be commoditized like that's the first thing that has to happen it needs to go away and usually how that happens think about what we did to batteries so for a long time elon's like i'm the battery man okay and then samsung was like we're the battery people like, all right. But still, batteries die after five years. People who had their Teslas were realizing a $20,000 battery after five years isn't really palatable to them. But the one thing that the infrastructure deal did in the Biden administration was they went back to the Sandia labs and said, OK, I know you've been making bombs this whole time, but now I got to figure out a breakthrough in battery. So you, Sandia, you're going to figure out how to create A battery that lasts 25 years costs 90% less. It can be used anywhere that we need to store information, including the size of a house or a car. Like, that's your job now. And they're now at, like, they're almost, well, before the Trump administration came, they were almost there. They almost had it. So, like, if you think about it, like, yeah, private industry might start it. But there's a way for government to say, you're not moving fast enough. You're monetizing it too much. Here we come. It's going to be 90% cheaper. It's going to last 25 years. And that's going to be the standard you're going to live by. And Sandia is going to bring it to you. That's amazing stuff the government can do for you if you put the right government in place.
Marc:Yeah, I think, yeah, if you put the right government in place, right, but, you know, that's the operative scenario. I don't think, even after, you know, Trump, you know, I don't think that the U.S. Has the political landscape, you know, or will to do that kind of stuff.
Renee:Dude, it doesn't have to be us. It can be you. I don't care who does it.
Marc:So that's what I was going to say, is that actually the EU, which I think they're a weaker voice than they should be, but the EU actually has, you know, some progressive thinking in this space. So, you know, particularly around privacy and sovereignty of data and that sort of thing. So potentially, potentially, you know, they could push for this kind of change. But I don't know. We'll see. We'll see. I really don't think it's going to. We'll see.
Renee:But there's a way to get there without us funding it.
Marc:Yeah. We're living in the world of the tech bros. And we're just, you know.
Renee:And they want us to fund everything. And if that means they have to hollow out all of our pockets to do it, they're willing to do it. That's right. That's why pay for your subscription and nothing else.
Marc:Nothing else. Yeah. Yeah. Like agents, agents should act on behalf of the people, not the platform.
Renee:That's that makes me warm and fuzzy. I hope it works out.
Marc:All right. It's not it's not it's not agency if it if it behaves.
Renee:I hear you.
Marc:You know, like corporate here.
Renee:This is how it works in America. The the agency we're talking about is not personal. It's corporate. Always, always, always. Right. Always.
Marc:I get you. I get you. But but that's the thing. Like the agents can can act for multiple parties. And you know what? An agent that that is representing, you know, whatever, you know, Bob's bait shop. That's okay that it's acting on behalf of Bob's Bait Shop. But my agent shouldn't act on behalf of...
Renee:Bob Bait Shop.
Marc:Right. Bob's Bait Shop or OpenAI or Google because Bob's Bait Shop paid Google and paid OpenAI to load Bob's Bait Shop into my agent.
Renee:I mean, how else are you going to pay for Google or OpenAI? Like, how else are they going to make money? It's a valuation they need to make to be who they want to be. Like they're literally going to have to monetize that whole platform and that's how they're going to do it.
Marc:I'm making a prognostication here right now, right now. Digital marketing and advertising is dead. Digital marketing and advertising is gone. In 25 years, 25 years, it will not exist or it will exist in such a different form than that. It will be unrecognizable. Think about it.
Renee:No, I can't stop thinking about it now. Damn you, man. That's all I'm going to think about for like the next week. This is it. This is all I'm going to think about.
Marc:That's right. We talked about this before. Agents talking to agents. Why do you need advertising that is meant for eyeballs when it's agents that you're trying to target?
Renee:You know how much I love a good ad, too. That's a shame. Damn it, man. A lot of artistic people who can't get jobs in Hollywood end up making stuff for content, and I'm going to miss it all now. It's not going to make any difference.
Marc:We've gone way afield here.
Renee:Yes, we have. Let's just end this one, shall we?
Marc:All right. Look, everybody, thanks for joining us on this tour through two centuries of infrastructure bubbles.
Renee:If you have thoughts on whether AI is the next one, send them our way at nostalgicnerdspodcast at gmail.com.
Marc:And if you enjoyed this episode, subscribe, rate, slap, whatever. Tell your friends all the usual stuff.
Renee:And until next time.
Marc:All right. Stay nerdy.
Renee:Stay nerdy, folks.
