The text below is a transcript of the audio from Episode 35 of Onward, "The 5 patterns of technology breakthrough".
Disclaimer: This transcript has been automatically generated and may not be 100% accurate. While we have worked to ensure the accuracy of the transcript, it is possible that errors or omissions may occur. This transcript is provided for informational purposes only and should not be relied upon as a substitute for the original audio content. Any discrepancies or errors in the transcript should be brought to our attention so that we can make corrections as necessary.
---
Cardiff Garcia:
Hello and welcome to Onward the Fundrise podcast. I'm Cardiff Garcia of the Economic Innovation Group, and I'm joined as always by Fundrise CEO and legendary surfer of tech waves. Ben Miller. Ben, hello.
Ben Miller:
Hey, Cardiff. How are you?
Cardiff Garcia:
Just teasing up the episode there. Why don't you tell everybody what we're about to talk about, because this is one of the most exciting topics that we've covered on Onward, I think, to this point.
Ben Miller:
Yeah, so our team did a lot of research and analysis for this episode. We went back and looked at the top hundred or couple hundred largest tech companies going back 30, 40 years and looked at revenues and market caps. When we did it, we found some surprising patterns and then also as I dug deeper into them and the patterns were so visible that I even wanted to name them. So I took a shot at
Cardiff Garcia:
Custom named them.
Ben Miller:
Yeah, some of them are custom names, some of them are well known in the industry.
Cardiff Garcia:
Are borrowed, let's say.
Ben Miller:
Yeah, because some of the patterns are very familiar to insiders. Everybody in tech who has been around knows these patterns and some of them I think are not as well known, the financial metrics around these. I think I hadn't seen anybody ever do this kind of research looking at hundreds of tech companies over decades and decades. I think a few of these things will be really new even to insider insiders.
Cardiff Garcia:
Yeah. I want to point a couple of things out to our listeners. One is you have not shared these findings with me before the chat on purpose. I'm going to be as surprised, shocked, delighted as our listeners are, and hopefully I can also then act as an avatar for our listeners and ask follow-up questions and all that kind of stuff. And also we want to emphasize before we get started that this podcast is not investment advice. This is for entertainment and informational purposes only. And oh my, is it going to be entertaining. I can't wait. Ben, these patterns that you described and then we're about to introduce, they don't just apply to these hundreds of tech companies that your team studied over the last several decades. They apply to the waves of technological innovation that those tech companies represented. Is that correct?
Ben Miller:
Right. But before we get into it, lemme just say one thing about patterns is that why do patterns matter? And so VC investors will often describe themselves as pattern matchers because they often say is that they see more data across more companies, and so they're really good at recognizing patterns and entrepreneurs, it's so funny, if you're starting a tech company, you're in a tech company, you work in tech, these patterns are so consistent, yet it feels like every time they happen, people don't know that and they seem new. I sum it up as nearly everything's happened before in the past. And so we might as well learn from it. So there's been five major waves of technological innovation
Cardiff Garcia:
In the last few decades, to be clear. Not like forever.
Ben Miller:
Yes. In what I would call the modern technology era, and that would be with the launch of that first computer. And those five tidal waves were mainframe, personal computer, internet, mobile, cloud computing, and now AI.
Cardiff Garcia:
So mainframe is the first one. It's PC, so it's PC, internet, mobile, cloud, AI.
Ben Miller:
Got it. But yeah, the first was mainframe and there's a whole history there. And the personal computer was the first disrupting wave. The first computers were mainframes.
Cardiff Garcia:
And so you've discovered five patterns that seem to apply consistently to all five of these waves, correct?
Ben Miller:
Yeah. First, people should know that the wave pattern is a pattern, and that catching the wave, recognizing the wave, helping create the wave is the most important way you make money in tech. I didn't want to skip that part, which is that there are waves in technology, there are big waves, which is these big breakthroughs. These waves go through these sort of lifecycles, which is like a breakthrough adoption, maturity, and then old age or senescence. So just before I get into all these patterns, the first pattern is there are waves. And if you don't know, that's a pretty important one, but okay, let me give you the other one. So the other ones, some of these I had to name myself. So one are waves, two, I'm calling the bridging pattern, three, democratization, four, centralized decentralize, and five, the three x multiple.
Cardiff Garcia:
Some of those names sound more exciting than others. Even the idea that VCs describe themselves as pattern matchers just doesn't sound quite as sexy as saying I invest in technology. Right. Pattern matching sounds like something you do to choose your drapes, but let me just go back through each of these five. So waves, the first consistent pattern is that it is an actual wave. Second is bridging, third is democratization, fourth is centralization decentralization, and then fifth is the three x multiple. Right?
Ben Miller:
Right.
Cardiff Garcia:
Okay. Do we want to define each of these and then dive in or should we just start going with the first one?
Ben Miller:
I'll do a quick summary of each, then we can dive into 'em. They're each pretty deep. And the one that I feel like is the one I named, I'm calling it the bridging pattern. It's the most important pattern in tech, and it's actually the primary job of the tech industry. You would think it would have a name, but it's basically to take the innovation from the hardware to the human. So just to try to give an example of that, the mobile phone and in the chips inside or the hardware, an application might be the app store. It's where all apps have this shared infrastructure, and then an app might be Airbnb. And so those are three parts of the stack and each part of the stack's getting closer and closer to the customer. And that's the bridging pattern because getting this innovation, this sort of unlock to the customer is how the technology industry monetizes. And then the third pattern, democratization pattern is then the technology ends up giving more people more power, more computing power, more ways to solve problems in more places. If you think about it like the iPhone has a hundred thousand times more computing power than the computer that landed the space shuttle on the moon.
Cardiff Garcia:
And another way of saying that is that all this computing power has also just become a lot more affordable.
Ben Miller:
So that's democratization. And one that's near and dear to my heart.
Cardiff Garcia:
Indeed, the mission of Fundrise is exactly that, to democratize investing in private assets.
Ben Miller:
And so then the fourth pattern, academics, we call it centralized decentralized. The software industry calls it bundle unbundle, and the hardware actually used to call it construct deconstruct, but basically there's consistent patterns of how technology will end up causing power to centralize or power to decentralize.
Cardiff Garcia:
Yeah. Ben, can you say a little bit more about that? Can you untangle what you said there about bundling and unbundling and how that connects to this concept, this pattern of centralization, decentralization?
Ben Miller:
Yes, I'd love to, and we can go deeper into it when we get to the next part.
Cardiff Garcia:
Save it for later then. Yes. And the last one, the last pattern is the three x multiple, which sounds tasty. I got to say, this is the one everyone's going to like.
Ben Miller:
This is the financial pattern, and this is what we found in our research were financial patterns. And so the three were a consistent market value. And what we found was each wave was about three times bigger than the last wave. So the internet was three times bigger than the PC from a market cap or market valuation of all the companies, mobile was three times bigger than the internet. So each wave ends up being three times bigger. And that was interesting. We were trying to estimate the size of AI and how big AI's market cap could be, and we actually ended up at a number of 15 trillion based on the consistent patterns.
Cardiff Garcia:
That's what it is right now? Or that a projection based on the three x concept?
Ben Miller:
Yes, projection. AI is not worth 15 trillion today.
Cardiff Garcia:
I was going to say, okay.
Ben Miller:
At maturity, so what's the value of a wave at maturity each time it's grown about three x, and they were consistent patterns around the market split market value split and market share and winners and losers. So I'll get into that when we get to it. But they were really consistent patterns in the three x multiple. I almost called it the constructive wave pattern. For those who are real nerds.
Cardiff Garcia:
I like three x. That sounds futuristic because it is the financial pattern of these five. I quite like that name, but that's great. Okay, so we've got these five patterns, which again are waves, the bridging pattern, democratization, centralization/decentralization, and then the three x multiple. Shall we start going deep on each of these one by one then? And we'll talk about some of the examples that we have to illustrate how each of these particular patterns ends up manifesting?
Ben Miller:
Yes they'll make more sense as we get into the details and people who have more background. So I thought we could start with the waves, the five waves, or I guess it's really six if you want to go back to mainframes, but the computer was invented and the sort of initial wave that actually was replacing punch cards with computers. Then the subsequent waves are PC, internet, mobile, cloud, and now AI arguably, we don't know for sure yet, but it seems very likely at the moment. And so just to try to put some color on that, I think people may understand better with some examples. So PC, the wave hits and you end up with hardware winners, platform winners and application winners because over time, as the wave matures, you end up with consolidation. So examples in hardware and PC are Intel, IBM, compact computers, Dell computers, those are all the hardware winners.
And then the platform that winners end up being Windows. Windows operating system is the software platform that sits on top of the hardware. And then all the applications that people build sit on top of Windows. An application might be Adobe or it could be Microsoft's office like Excel and PowerPoint. So you see, if you were going to build an application, you'd say, okay, I want to build something. I don't want to have to design all the operating systems that need to talk to the hardware. I want to just sit on top of somebody who's done that. And so that's what the platform is. And give another example, which I think we talked about the mobile hardware is iPhone, Samsung and Qualcomm. They're chip makers. And then the platforms, there's only two platforms, Apple App Store and Android. And then all applications sit on top of that, Instagram, Uber, Airbnb.
That's actually how, and I could go through every single wave, but that's how every wave is organized and the innovation starts in the hardware and works its way up to the platform application. And actually if you look at every wave, interestingly, there's a lag between when the hardware innovation’s created and when the platform gets created and when the application finally becomes the thing customers use. So typically the hardware is invented seven years before the platform. So like the Intel X86 is created and you don't see Windows until seven years later. So anyway, there's this very consistent wave. And why that matters so much is that it's very hard to, once a wave is established, to basically create a startup or a competitor. So once the players in the wave have won that wave, you can't really disrupt them and they just have a monopoly type profits. And what we've been seeing in big tech is the late stages of that, and then people haven't been happy with it, but the next wave creates this massive opportunity for new winners and that's why waves are so important.
Cardiff Garcia:
Yeah, one thing that strikes me about each of these particular waves also is that it's not like each of these waves only ended up having a huge effect on information services or information technology. These are what a lot of economists might refer to as general purpose technologies. Some are more important than others, but each of them also has an effect on how every other part of the economy operates. And it's natural I think for that to just take some time for it to have that kind of effect, it has to spread throughout the economy, throughout society, and it has to gain wider acceptance to the point where people realize that it's worth buying, it's worth consuming, it's worth investing in if you're a company. And that doesn't happen overnight.
Ben Miller:
It takes on average over a decade from when the initial innovation is created to actually customers who are non-technical are seeing benefits. So the lag was a lot longer than I thought, but that's why the bridging pattern is so critical because essentially most people, which is how companies make money by selling to many companies, many people are non-technical. And so you have to get the innovation to a point where someone who's non-technical can use it. And that takes a really long time, basically a decade.
Cardiff Garcia:
And by the way, not every technology that looks like it has the potential to transform society necessarily plays out the way its most optimistic boosters necessarily expect. And here I come back to crypto, crypto is something that people thought, especially chain technology might end up really radically changing the way we do things, the financial sector, even the healthcare sector, and it just hasn't played out that way. Crypto for the most part, remains a speculative investment asset, but it has not actually radically transformed anything. And so you can understand why it takes some time to adopt all these technologies because each time a company says, Hey, maybe I'll invest in that, there's a chance it won't work out. And so it helps to have a lot of first movers who say, yeah, we'll try it. And then they do a lot of the beta testing or whatever. And then when it works, then you see a larger number of companies say, yeah, okay, we'll take it on and we will employ it at our company as well
Ben Miller:
Because the gains can be so large, if AI is worth 15 trillion, then it makes sense why all these companies are currently investing tens of billions to try to win it. There is still a possibility that AI ends up not being a wave, or at least not anytime soon. That's not what I personally think. And then the other thing that's actually interesting when you go back and look at the patterns is that the early winners don't always win. And so if you went to each one of these waves and looked at early winners, so I go the internet, the early winners were Yahoo and Lycos and AOL. And now Google ended up winning all of those. And the same thing happened with social media right, Friendster, MySpace for early winners, and Facebook ended up being the dominant platform. So it's likely that a lot of the players that currently seem like the absolute winners, AOL at one point was worth a hundred billion dollars. That was a thing. 1999, Yahoo was the second most valuable internet company, and neither of those really exist.
Cardiff Garcia:
Yeah, seriously. One last thing I'll say about this first pattern is that very often what ends up happening is that you do get for a time a financial bubble and everybody worries about it or they point out how crazy the frenzy is and they say, wow, look at these valuations are crazy and so forth. And that could be true, but a financial bubble also very often enables a lot more capital investment in a new technology than otherwise would've taken place precisely because of all that enthusiasm. And sometimes as it did in the early 2000s, it does lead to a crash or possibly even an economic recession if it's widespread enough. And then you have a longer period of consolidation around these new technologies, you have people start taking a more judicious approach to employing them. You sometimes get new regulations in the aftermath of the crash and so forth.
But the point here is that progress can be uneven sometimes. Maybe we'll get lucky and the next big technological innovation, if it is AI let's say, will take off and it'll be great for everybody and for the economy and society, even in the absence of a big financial market collapse or a recession or whatever, maybe not. I have no idea. But the point is that sometimes the progress is very uneven, and it doesn't actually mean that big early frenzy of investment was necessarily a bad thing overall, even though it looks quite bad. And even though a lot of people really do lose their money.
Ben Miller:
In the early phase, there are winners and losers, and if you're going to put your money into it, that's the reality. The winners have extraordinary gains and the losers usually go to zero even when the companies seem like the sure things like, AOL, Yahoo, or if you go back earlier, Netscape was the first sort of popular browser and it was worth whatever, a few billion dollars. And Microsoft Explorer ended up taking 99% of the market by like 1999.
Cardiff Garcia:
Yeah. Remember when I said that this podcast was not investment advice? I'll break that rule just momentarily and say, pick the winners. That's better, that's better. That's investment advice.
Cardiff Garcia:
You can put that on the record.
Ben Miller:
And then the other point you're making is that there can be massive societal benefits even from the losers. And so a good example is that as a telecom industry, put something like half a trillion dollars into the ground building telecom infrastructure, and then basically the whole industry went bankrupt in the early 2000s.
Cardiff Garcia:
Yeah, WorldCom collapsed, but left us with all this great infrastructure, communications infrastructure. That's right.
Ben Miller:
And so it's a great example of you could see hundreds of billions going to AI. It'd be great for society and then not for the investors. And that's actually a pretty typical pattern. And that's okay as long as you can try to either index the winners or maybe be judicious in the beginning, but for venture capitalists, this is the game they play. This is how they make the big bucks. So for them, they're obsessed with these patterns. They're trying to figure out is this a wave? Who's going to win this wave? And then how do I play it? And if I'm an entrepreneur, which I'm also an entrepreneur by day, that is critical. You have to know what's happening in the wave. And I feel like a lot of times people don't see the broader context. Let me just go to my favorite one, which is the bridging pattern, because I feel like that pattern actually explains what's happening during these waves. Why does it take 10 years? That seems like an extraordinarily long time, but actually it can take even longer for the innovation to work its way from the bottom of the stack, from the hardware up to the customer, like my mom.
Cardiff Garcia:
Your poor mom catching strays on this show for being tech unsavvy. How rude.
Ben Miller:
She definitely gives the Fundrise IR team their share of questions.
Cardiff Garcia:
That's great.
Ben Miller:
She always writes in Ben's mom here, and then she's always logged out and can't get in.
Cardiff Garcia:
Can't complete the sentence. So just catching strays, we apologize, Ben's mom.
Ben Miller:
But this is a really interesting one because there's really clear patterns in this bridging dynamic and that bridging dynamic or that bridging pattern is important if you're going to try to invest into it or entrepreneur or even work in the tech industry, seeing what's happening, seeing actually what the work is is remarkable. I feel like oftentimes if you're a software engineer, you actually don't fully recognize or you're whatever, you're a product manager, what is happening because you are so far in the weeds. The broad pattern is right, you have to go from highly technical to highly non-technical, and the economic value is created by the user interaction, the user interface that unlocks that simplicity. And so I'll give you a few examples of that over time. The most famous one is the original computers were black and white screen or whatever, green screen with just a DOS command line, you had to only type text and you had to know what kind of commands the type. I remember trying to figure out how to get my games to run, and I had no idea what the commands were. That was the original interface. You had to be at least technical enough to understand the command line. Here I am exposing my ignorance.
Cardiff Garcia:
Well, I was going to say, you're showing your, which is also my age. There's a lot of young people who are like, what the hell is this guy talking about?
Cardiff Garcia:
Yeah, what is the DOS command line? Yeah. Yeah. My Commodore 64.
Ben Miller:
So what Xerox Parc figured out was that people didn't want to interact that way, they wanted a graphical user interface, pretty pictures. And so they created this gooey, and then Steve Jobs saw it and copied it, and then Microsoft copied Steve Jobs. And so the way we interact with the computer today, which has got the file folder and it's all organized, it's actually a metaphor for an office because it's got files and a desktop and documents, but in reality, those things exist. They're just a visual metaphor, but allow you to navigate it. That was an enormous breakthrough, and that created the computer revolution as we knew it starting in the mid ‘80s. Same thing with the internet. The original internet, the browsers were text and images opening up in different windows, and Marc Andreesen from Andreesen Horowitz, famous venture capitalist, created the first browser called Mosaic by combining the text and images onto a single window, and then they wouldn't let 'em license the code.
So they went and created Netscape. So again, what made the internet take off was creating a user interface that non-technical people could use. And you saw the same thing happen with the iPhone. The original mobile phone was the Blackberry, only text. And what Steve Jobs did was he realized, so its touchscreen plus a graphic user interface was the iPhone breakthrough, and actually the computer was a graphic interface plus a mouse, right? So it was that combination of how the user wants to interact with something very technical. You have to understand, you have to map essentially what the technology is to how people think. You have to map it to their five senses and they have to map it to the job to be done. So that's why this bridging or mapping from hardware, from technology to people, that's actually what is happening inside the tech industry. That's what the platform industry does to hardware and what the applications are doing to the platform.
Cardiff Garcia:
Yep. Yeah, that makes sense. And the thing I'm trying to work out as you were explaining that, Ben, is how does this now apply to the more recent waves of technological innovation? How might there be a bridging between some of the early technology that we've seen in terms of AI, at the very least, these kind of chatbots and whatnot, and then the applicability for users, because some of us, by the way, are already using these things, early adopters and so forth. The technology itself seems so incredible, so powerful, but I don't actually think that we're really past the earliest innings of this.
Ben Miller:
We're definitely at the early innings, and it's why it's so applicable. If you understand the pattern or you understand the map of how things play out, you can understand where we are, what's coming next, and then how to invest or how to entrepreneur. As I said, if you look at the computer, the initial screen was DOS text, internet was text, Blackberry text, AI’s first use case is a text chat bot. It's just text. And the interesting thing is the LLMs, the large language models that drive the chatbot ChatGPT have been around for years. And the innovation, the unlock that actually surprised OpenAI was the chat interface on top of the LLM. So that was the interface breakthrough, but we're still in the text period. And as we said, it goes from text to images. And so I'll give you example of mobile. Mobile starts out text like Twitter, and as it advances, it goes to images and text, which is Instagram and Pinterest, and then it advances some more and it gets to TikTok.
And so the reason why it advances that way is that in the beginning of the new platform and the new hardware is it doesn't have enough compute to do something really intense. And so text is low compute, images require more processing because there's more data. And videos obviously require the most amount of data because there's just way more data obviously in a video, an image than a text. For example, internet started with GeoCities. GeoCities was the third biggest internet company in 1998. Then images plus text is like eBay. And then video internet is YouTube, and each one starts after the other and each one gets to popular explosion when speed, cost, quality all come together. And so with AI, we're not actually at that point, but it's definitely coming because of Moore's Law and Huang’s Law, which we can talk about later. So what's happening behind the scenes and why you haven't seen applications and why you're sort of still wondering, is this AI thing going to amount to anything is people have to bridge this by doing this exercise of trying to map what the technology can do to how a person wants to use it.
And so what happens is that in the beginning, software developers, software engineers are doing that with hardware and they're creating this platform. And that's what OpenAI did. That's the initial customer of all technology. Our software technologists, our other technologists, that's where we are mostly today with OpenAI. We have OpenAI access to their API, and we're trying to build an app and other people are building apps, but that pattern is just repetitive. And so seeing this sort of snapshot and also knowing there's this lag, I didn't mention the lag on average from platform to application is 36 to 48 months. So iPhones ‘07, Uber, 2011, four years.
Cardiff Garcia:
So large language models, at least when everybody first started getting to use them all at once, roughly coming on what, 18 months to two years now?
Ben Miller:
November ’22, I think would you say is when it all launched? And so 36 to 48 months would be 2025, 2026.
Cardiff Garcia:
Okay. All right. We'll see what happens. Clock's ticking. It gets more and more exciting. Shall we move on to democratization?
Ben Miller:
Yeah, I’ll do democratization. I mean some of these people in the tech industry, they're so familiar with it. Democratization and bundle unbundle. I'll mention sort of one interesting thing about this, again, a little bit technical, but I found it uncanny. So programming languages also follow this bridging pattern. So in the beginning when you had mention of hardware, invention of computer, everything was zeros and ones. So the original software language was in zeros and ones, called binary code machine code. And then somebody invented an abstraction that's a little bit more human friendly called assembly code. It's like a hexadecimal. But anyways, I barely have my arms around it. My CTO was giving me a tutorial over the weekend. And then the third generation that came after that in the 70s was what now is how everybody's software programs Java, JavaScript, C++. And the reason why it's so relevant to AI, I'll tell you in a minute. But each new generation of languages are more human friendly, more abstract. And the fifth generation, which is speculative, might be AI.
Cardiff Garcia:
One of the promises of AI is that people who never learned how to code can still take advantage of these languages, right? Because for example, if I work a lot with statistics and economics and whatnot, but I never learned how to code, I never learned R or Stata or whatever. Ok that means there's a lot of things I can't do in terms of charting trends and analyzing trends. But if I can just go to an AI and say, “Hey, I want a chart of the following”, and then just say in plain language in English, in other words, what I want, then the AI can just write the code for me and spit out the chart. I'm actually quite looking forward to that day. I think it's coming quickly too, by the way.
Ben Miller:
Right. And that is a perfect example of democratization. And so you're democratizing the ability to code by
Cardiff Garcia:
Even to dummies like me? Absolutely.
Ben Miller:
Both of us. Where you get to an abstraction that's high enough that you become less and less technical and it makes it more and more accessible to more people. And the idea that either one of us could say, okay, let's code up an app that does something really nerdy that we would be excited about. That's something that was impossible for me without learning one of the third generation… I skipped fourth generation, by the way, just to mention them. R is one that you just mentioned. SQL is another one. And SQL is the most commonly used programming language in the world. The data analysts can use SQL, but if you want to do software programming, you had to actually be at the third generation. So each generation requires less technical background. And so that's a really good illustration of this sort of democratization. And so another one is just looking at just broadly Moore's Law over the last 30 or 40 years.
Gordon Moore was one of the founders of Intel, and he noticed this trend in the manufacturing of chips, of semiconductor chips that every two years they doubled their computing power. You could actually have twice as much compute for half the price every two years. And so in 1968-69, there were about a thousand transistors on a chip, and by 2023, there were 50 billion. That's how you have a hundred thousand times the power in your iPhone to the space shuttle computer is that the Moore's Law increased the power exponentially, drove down the cost, increased the speed, and then also decreased the size. And that's a form of democratization of power because essentially if anybody can get it and it's really cheap and it's really powerful, then more people can do stuff with it. You were talking about wanting to do some cool economic analysis. So in 1980, you would have had to go to a Cray supercomputer. Cray computers were like the state-of-the-art for computing. That computer back then cost $40 million, which today, if you took an iPhone and said, okay, if I could buy an iPhone in 1980 with the same power it would've cost 200 billion.
So how much does an iPhone cost? Like three, four, $500? So if something that would've cost 200 billion, it would've been impossible. But if you look back at the amount of computing power, how much it costs in 1980 versus now talking about a reduction in cost of $200 billion, just absolutely insane. And one other fun fact I found is that that Cray computer weighed 5.5 tons, so it took 30 people to assemble it. And you're talking about having something I think 50 million times more powerful in your hand. That's behind all of this change. And that's essentially a democratization is more computing power becomes available, things that were only possible for somebody who is either very technical or big corporations can now be done by individuals who are less technical. And that pattern means that you open up new markets and open up new business applications that weren't possible before.
Cardiff Garcia:
Yeah, absolutely. Worth noting by the way that Moore himself said that at some point, Moore's Law was likely to stop holding. that at some point it might break down, and yet it lasted for an astonishingly long period of time. I have no idea if it continues to apply now, but it's remarkable that kind of exponential growth kept going for as long as it did. And for all I know might still be going on. I just don't know.
Ben Miller:
Yeah. What's funny about that, so people have been lamenting the death of Moore's Law for a long time. They said it's going to come to an end, going to come to end, because at the beginning of the last decade or about 2012, chip was about 28 nanometers wide. And now it got, or now about last year got to about 5 nanometers wide. And so people were seeing that at some point it can only be so wide before it can't be any thinner, otherwise, its quantum physics of it disrupt the ability to work. And the irony is, and I always see this once when mainstream people declare something over is when it actually gets rebirthed. So have you heard of Huang’s Law?
Cardiff Garcia:
No. What is that?
Ben Miller:
Yeah, Jensen Huang is the CEO of Nvidia.
Cardiff Garcia:
Nvidia, yep.
Ben Miller:
In 2018, he went out and he made the speech and he's talked about supercharging Moore's Law, and then people just started calling it Huang’s Law after him. So this is why Nvidia is worth so much money. So over the last decade, Moore's Law, if it doubles every year, it's equivalent to 32 times more compute power over 10 years. So Nvidia, which took a different approach, they don't just look at the packing basically like transistors on a ship, they also look at the networking and the algorithms and the architecture, they look at everything. They take a holistic approach is how they describe it. They have been able to increase compute by a thousand times in the last decade.
Cardiff Garcia:
So in other words, they blew Moore's Law out of the water.
Ben Miller:
400 times more. And that is something that's been going on since about 2012. Most people didn't know about it. And that's also why people in AI are so excited because they see it's not two years from now, everything's going to be twice as computationally powerful. You're talking about like 10, 20x. And so the rate of improvement is actually vastly accelerated. And I can put some charts up or something, but it's why being optimistic about AI actually makes a lot of sense. The underlying hardware changes are extraordinary.
Cardiff Garcia:
Excellent. Shall we move on to pattern number four, centralization decentralization? Go even deeper on that one?
Ben Miller:
Yeah. This one again, is like most tech people think of it as a monetization strategy. And so it was actually from a famous quote by Jim Barksdale, who was the CEO of Netscape. He said something like, ‘Gentlemen, there are only two ways I know how to make money: bundling and unbundling.’ There's an economic way to think about this, but there's also a technological way. And obviously the mainframe was a centralized technology, and PCs were decentralized and the internet actually was even more decentralized than the PC. And actually what happens is a lot of times when things get highly centralized, they usually get ripe for decentralized, become distributed. And when sometimes gets things overly decentralized, they get ripe for centralization. So cloud started recentralizing, but let me give you some examples of famous ones. So it used to be that if you wanted to do any graphic rendering, if you're going to make a video game or maybe if you're an artist, you needed to buy a workstation.
And the most famous was Silicon Graphics Workstations. And this little startup said, ‘Hey, what about if we just created a graphics card you could put in your PC instead of having to buy a whole workstation?’ And those graphics cards were called GPUs, which was NVIDIA's invention. And so they unbundled the ability to do graphics rendering from a whole workstation. It's like a deconstruction. IBM, which is the famous centralized player used to have everything we think of in tech used to be just IBM, they were the hardware, they were the software, they were the application, they were everything. And so networking technology allowed computers to talk to each other, was all integrated into IBM. And then it got deconstructed by Cisco and Juniper, and they created this massive new market, which is like a deconstruction of existing players. So if you think about the present, and you look at who seems very centralized and likely to get deconstructed, it's Google and maybe Apple. Google probably ends up having a successful AI application, but their dominance of internet seems ripe for deconstruction. And you can look back and see, like Microsoft famously missed mobile. And so if you miss one of these waves, it creates this enormous risk that you actually get disrupted and go out of business.
Cardiff Garcia:
Although interestingly, in the case of Microsoft, they very much caught one of the next waves, which was cloud, and they've done incredibly well because of it.
Ben Miller:
Yeah, yeah. Same with Apple. So Apple missed the internet, and then Steve Jobs showed up and invented mobile or reinvented mobile. One of the things, it's in a later pattern is that one of the big tech incumbents actually can miss one wave but they can't miss two. If they miss two, they're gone.
Cardiff Garcia:
Yeah. One interesting example on unbundling bundling is actually from my former profession of journalism. As people know, newspapers have not been doing great financially for the last few decades, frankly. There are still some big national newspapers like the New York Times and the Wall Street Journal that have a lot of different things contained within them. But most newspapers, especially at the local level, the ones that still exist, they are a lot smaller than they used to be. And part of what happened is that your newspaper used to have everything. It had obviously the news, but it also would have sports and it would’ve had crossword puzzles. And crucially, it also had advertisements in there as well. The internet comes along, what happens, you have unbundling, you have now online sites that can be very niche, so you might get your sports on the internet instead of in the newspaper, niche news sites as well.
Advertising crucially went to places like, I don't know, Google or Craigslist or whatever. And so the business model was totally gone. But more recently people have noticed that it went too far the unbundling because people started thinking, if I have a website, I don't need a homepage anymore for a news site. Why? Because everybody goes to find their news in these social media places. They go via Facebook or they go via Twitter or what have you. And people realize that actually some readers of the news like things to be consolidated in one place, at least organized a little bit. So now homepages are making a comeback. That's a very small example, but it's one instance where you can see that in some ways unbundling, if it goes too far, contains within it the seeds of its own reversal, and then you get more bundling. And then when bundling goes too far, it creates within itself the potential for some unbundling later. So it really can move in this kind of cyclical way.
Ben Miller:
Yeah, you can feel it today because people feel overwhelmed by information, because everything's unbundled. Where in the old days, you only got your news from CBS, NBC, ABC, it was entirely bundled. Let me do pattern five. That's the one that's really new information. I didn't see it anywhere out there. And that's the financial patterns in all of this research we did. So I mentioned that each wave was three times bigger than the last. So we took what we thought was the maturity of each point of the wave. As we said, a wave, start with it. There's a breakthrough, adoption curve, maturity, and then it declines, senescence. It just, no one uses mainframes anymore. And the PC industry 1995 was worth about $186 billion. And then the internet peaks a decade plus later, and peaks at about $650 billion, about three times bigger. And then mobile is 2 trillion, cloud is about 5 trillion. And AI we think is then 15 trillion. And that's actually on average a 17% CAGR or 17% annual growth rate. That’s pretty consistent, and the reason why it's been pretty consistent is that underneath of it, the underlying motor is Moore's Law.
I think that’s the reason why it's so consistent, and it's actually funny because there's this famous story in the venture world, this guy Mike Moritz, who was the CEO of Sequoia or whatever, lead managing partner, he took it over from the founder Don Valentine. And when he took it over in the early 2000s, he looked at Don Valentine's success. Don Valentine had done Cisco, Oracle and Apple, which were, up till then, the greatest returns in history. And he said, how am I going to match this? This is insane. And then he made this intuitive leap that he says, as Moore’s Law continues, then the markets that technology can address actually get bigger because Moore’s Law is just exponentially cheaper and better. And so to see the math match his intuitive leap that it actually, cloud was three times bigger than mobile, mobile is three times bigger than internet is like uncanny.
And then a couple other interesting financial patterns that we saw is that the market share split between hardware, platform, and applications was actually fairly consistent per wave. So just an example. So in mobile, the hardware gets about 25% of the economic value. The platform gets about 25%. And the applications like Uber or Airbnb, those are applications. Those you got about 50% of the market cap. So you see this sort of 25, 25, 50 split. That's an interesting pattern. This is very much a financial pattern, but it's a winner take all business. This is why you said the winners and losers are so important. And so it's unbelievable how much the top two companies take of economic value. So in the platform level, so the platform is an Apple or Google, they take 80-90% of all the economic value of the platform, right? It's super, super, super consolidated. And that's been consistent going back to the PC, right? Microsoft and Apple were 80% of all the market value. And you go to hardware, same thing, super consolidated, it's usually about 50% of the market value is in the top two players. So you see this very consistent hyper consolidation or winner take all dynamic across every platform, which is why the VCs are so obsessed with trying to pick the winners.
Cardiff Garcia:
I'm intrigued by the idea that there eventually is this stable ratio that emerges of 25%, 25%, and then that last bit 50% for the applications. But of course, I'm imagining that as a new technological wave sort of evolves, that ratio itself is also fluid. Because at the beginning I imagine that quite a bit of the value is in the hardware because that is what exists. And that is the thing drawing in all the enthusiasm. And that later over time, you end up with all these new applications that end up taking more of the value because people start seeing the value to them, and then it grows.
Ben Miller:
That’s for sure. I was just looking at the mature wave and how the economics are broken down. So on a typical mature wave, there's only five or six hardware companies that have 95% of economic value. In the platform, there's only 3 or 4, and in the applications there's only 15 to 20. So you're talking about, at most 30 companies, representing 80-90% of trillions. So in AI, if it's 15 trillion, you're talking about a very small number of companies being very valuable.
Cardiff Garcia:
Worth an outrageous amount of money. Yeah.
Ben Miller:
Yes. That's why the venture industry exists.
And I think that a lot of the popular desire is to see big tech companies go down or at least to be reduced in power. But what's interesting is that the waves typically add new major big tech players. And so you get more competition, not by reducing power, but by adding new players. I call it the plus one minus one pattern, because if you go look, you had mainframe, you added Microsoft and Apple. Internet, you added Google and Facebook. Mobile, you added Apple and Instagram, which basically was Facebook. Cloud, you added AWS. Amazon wasn't really a technology company. And now AI, you're adding OpenAI. So I actually think that one of the lessons is that the key is in order to get more competition among big tech, is to encourage the addition of new players in these major breakthrough waves and make sure that the regulatory environment is conducive to that kind of competitive dynamic.
And then look to some losers. And the funny thing about these losers is that you actually probably never heard of 'em. People who were the major players in these waves up until recently all know this stuff, because they were around. The inventor or the dominant player in word processing was this company called Wang. It was started by this guy An Wang. It had 33,000 employees and 3 billion in revenue in 1980. So this is a huge company. And when the PC showed up, he famously had a quote saying he's not going to disappear in a year. And he disappeared in two. DEC, Digital Equipment Corporation, which is a company that had 130,000 employees.
Cardiff Garcia:
Huge.
Ben Miller:
Yeah, gone. These disruptive waves are very consequential. And people in the industry, a lot of them worked at these places like John Chambers, who CEO of Cisco was vice president of Wang Laboratories. And a lot of people who worked at Sun Microsystems, which ended up not being able to transition, they were also a server company, ended up going on to run a lot of these other big companies. So these waves, everybody who's in leadership at tech has lived the highs and lows of these waves, the near-death experiences of Apple and Microsoft, because if Microsoft hadn't caught cloud, they would be gone.
Cardiff Garcia:
Yeah, it's fascinating to think of some of these counterfactuals. The other thing that's curious about all these companies too is that we are right now mainly talking about companies in the information space and the seemingly unrelenting ongoing digitization of the world, right? But these companies occasionally do get involved in other kinds of technology as well. Apple made a move into EVs, looks like that may not happen from the outside. For example, Elon Musk, Tesla are involved in artificial intelligence. And so there is some overlap with other kinds of technology. And I don't know, maybe as a follow-up project, I'd be interested to know if some of these patterns, some of these trends also apply to things like clean energy, to new transportation, to things like biotech, all these mRNA vaccines and whatnot. I have no idea what the answer to that is. And I understand why we're focusing on these particular kinds of technologies because it seems like one has logically followed from the last. It would be interesting to know if this also applies to other technologies that maybe apply more to, and stay within, the more physical realm of atoms as opposed to bits as, I think a lot of tech people refer to it.
Ben Miller:
Okay, you don't know this, but I am planning on doing one on military technology.
Cardiff Garcia:
I know it now.
Ben Miller:
I have a professor I've lined up. I've done a lot of thought and reading on this over the last couple of decades. That one, I know for sure it's applicable. I'm loosely confident based on what I know that it’s true for clean energy. I think one way to bookend this or capstone is the big consequences. So Marc Andreessen famously said, software's eating the world. And so I thought I'd put some numbers against that. Some of these, everybody knows. So lemme do the one that most people know, and then I'll do the one not as many people know. So e-commerce was 1% of all commerce in year 2000 and is now 16% of all commerce. So it's slowly been eating up commerce to the deep dismay of the retail industry and people who own retail buildings like malls. So here's an interesting stat that I feel I hadn't seen before. How many employees work in retail trade? So according to the BLS, 15.6 million. So 16% of 15.6 million is 2.5 million people. So e-commerce should employ 2.5 million people because 16% of all commerce. How many people does it actually employ? 500,000.
Cardiff Garcia:
Oh, man. Yeah.
Ben Miller:
Yes. So it’s a good example of software driving a change in work patterns, more efficient, productivity, however you want to describe it.
Cardiff Garcia:
Productivity growth. Yeah, absolutely. That’s technology.
Ben Miller:
I’ll give you another one I think is interesting, less well known. How much do S&P 500s, so companies spend on software a year? So in 2016, they spent 3% of all their revenue on software, and in 2024 or 2023, they spent about 6%.
Cardiff Garcia:
That's a big move actually to double the share of your revenues that you end up spending on software. Yeah, absolutely.
Ben Miller:
So if that slope continues, it should be about 10-11% by the end of the decade. 10-11% of all corporate spend spent on software, which doesn't sound that much until I tell you how much is spent on labor.
Cardiff Garcia:
Okay, yeah.
Ben Miller:
So the S&P 500 spends 10-11% on SG&A, on people.
Cardiff Garcia:
It'd be a perfect match if that earlier trend actually continues. By the end of the decade, we'll be spending the same amount on software as on labor.
Ben Miller:
On labor, and it's possible it’s accelerating with AI and with Huang’s Law.
Cardiff Garcia:
Yeah. Worth noting, by the way, that might sound scarier to people than it should because of course this would also imply that the economy itself will benefit from these productivity gains. And so it doesn't mean that jobs will be permanent lost, that there'll be permanent net job loss. It means that everybody will be way richer and that wages will be higher, and that in fact, there still will be people getting hired. That's not a guarantee. But so far, if you look at history, that is how it has played out with some very big intermediate disruption. So I'm not guaranteeing anything. I want to be absolutely clear. I'm not guaranteeing anything. It's hard to predict these things, but I'm saying we don't need to necessarily be frightened by the fact that companies are spending more on software. It also could just mean that they're going to make the people who do work for them more productive, which means faster real wage growth, yay. You can afford more things, have a better living standard, et cetera. Or yes, it could mean that a lot of people do lose their jobs and that there's a period of disruption in between, and that's terrible. But we don't know is the point like that statistic by itself is not dispositive of anything else. I just want to make that point.
Ben Miller:
Yeah, I have one more stat that is consistent with what you just said. It's a positive, consistent narrative. Over the last 30 years, it used to take 8 employees to produce a million dollars in revenue, and now it takes 2. So it takes 2 employees to make a million dollars in revenue for the S&P 500 on average. So that means that revenue per employee is 4 times more revenue per employee. What was the SG&A or percentage of revenue spent on labor over the last 13 years?
Ben Miller:
Is it declining? Is it increasing? Or was it flat? You have revenue gains per employee, but do you have less employees as a percentage of revenue?
Cardiff Garcia:
I don't know.
Ben Miller:
So no, it's been flat at 10-11% for the last decade and a half.
Cardiff Garcia:
So relatively stable trend. By the way, that's across some very divergent business cycles as well.
Ben Miller:
And across a lot of different kinds of businesses. So it’s exactly what you said, which is even though companies are getting more productive, more revenue per employee, they're maintaining on average the same amount of employment as a percentage of revenue, which actually means more employment and more wages, because revenue's going up.
Cardiff Garcia:
Yeah. By the way, a simple way to think about this is that in the last, I think eight years, you said that share of spending on software has doubled and yet unemployment, right this very second remains below 4%.
Cardiff Garcia:
Labor market's healthy.
Ben Miller:
In other words, software spend doubled and employment spend didn't change.
Cardiff Garcia:
Got better, actually.
Ben Miller:
Wages went up because you had better productivity. So anyways, the consequences of the technological gains have been wealth and prosperity. And so I think with everything we just said, the pattern is telling us that we're likely to go into a very big economic boom, a technological boom. And as you said, there might be short-term disruption, but the consistent pattern of mainframe to PC, to internet, to mobile, to cloud have been very beneficial for society. And though United States has been leading all of these technologies now for decades. So another optimistic note coming from me.
Cardiff Garcia:
I love it. I love it. And probably a good place to wrap up. Hey, Ben, on the three x multiple, you weren't thinking of naming that Miller's Law instead have your own law man.
Ben Miller:
No way.
Cardiff Garcia:
Not yet? You don't want to make that bold a proclamation just yet to say it's a law?
Ben Miller:
Yeah. The challenge is that Moore's Law, we just broke Moore's Law. Because AI is going at 10 times faster. And so it's question of are we actually going to see much higher productivity growth, much higher market values than what we see in the past? Because the challenge with extrapolation is you're extrapolating the past into the future. And if the future is not like the past, it's not a good trend. And so I actually think we are likely to see, you said 10x the past because you have 10x the computational power, the rate of change is 10x faster. That's 150 trillion.
Cardiff Garcia:
What a future that would be, entrepreneur by day, crime fighting podcaster by night, I think Ben. Always a pleasure, man.
Ben Miller:
Yeah, thanks. It was awesome, Cardiff.
Cardiff Garcia:You've been listening to Onward, the Fundrise podcast featuring Ben Miller, CEO of Fundrise. My name is Cardiff Garcia of the Economic Innovation Group. We invite you to please send your comments and questions to onward@fundrise.com. And if you like what you heard, rate and review us on Apple Podcasts, and be sure to follow us wherever you listen to podcasts. Finally, for more information on Fundrise sponsored investment products, including all relevant legal disclaimers, please check out our show notes. This podcast was produced by the Podcast Consultant. Thanks so much for listening, and we'll see you next episode.
