Matthew Prince has been watching the Jetsons lately. Now, the Jetson family is very dependent on the resources provided by Rosie the XB500. That may have seemed far-fetched when the Jetsons originally aired when I was a kid, but as AI becomes more powerful and accessible every day, AI bots may well be playing a similar role in all of our lives. Matthew joins us this morning to talk about the radical changes that AI is bringing to the internet, commerce, and society as a whole. Welcome back to Mountain Money.
Thanks for being here. Okay, for many of us, AI has gone from like this abstract concept to a tool we use every day in a matter of months. As users, of course, we don't see what's going on under the hood. We just tap things into our computer and magic happens, but can you give us a sense of how AI is starting to affect internet volume?
Yeah, I mean, it's been really pretty wild. Back in November of 2025, so not even a year ago, I was asked at what point is non-human automated traffic going to pass human traffic online? And we at Cloudflare pulled all the data, and we looked at it, and the best estimate we had was it would be in the second half of 2027. Just a few months later, I was at a conference down in Austin, Texas, South by Southwest, and was asked the exact same question: when will automated traffic pass human traffic? And so we pulled the data again, and it turned out it had moved up, and that was going to be the first half of 2027. And so I was blown away when my team came to me in May and said, we're already seeing now more automated traffic online than human traffic. And I think any of you who have used ChatGPT or Claude or Grok or any of the AI tools that are out there, you ask it a question and you watch as it just goes and scours the internet. And if I was buying, let's say, a digital camera or something, I might personally go look at like five sites. My agent is going to look at like 5,000. And so that's just driving an enormous volume. And with a caveat that I've been wrong on every prediction so far, if we extrapolate the current trends out, in five years, automated traffic will be a thousand times as much as human traffic. Not because human traffic is going to go down—we think it's going to hold pretty steady—but because you're going to just see that much more automated traffic out there. In other words, humans will basically be a rounding error on the internet.
Well, we don't see the resources that it takes. I mean, we're going to get to the web creators and the vendors and what all those people have to do, but something has to make that happen with that volume. Who builds that network and who's going to pay for what has to be an expansion?
Yeah, I mean, at some level, that's what we do at Cloudflare. A huge percentage of the internet sits behind our network. Almost everyone who's listening to this has probably used our network hundreds of times just in the last 24 hours. And when we're doing our job right, you don't even see it. We run servers in over 300 cities around the world. We're in hundreds, if not thousands, of data centers, and we're providing that infrastructure that makes the internet go. But to give you some sense, during COVID, the internet traffic doubled over the course of two weeks, and the internet in Europe almost crashed. I was on this crazy phone call with Netflix and Disney and a bunch of other big streaming providers, YouTube, and it was sort of like a standoff kind of situation where everyone didn't want to turn down their video resolution because they were like, "Oh, people were paying for our 4K stream and we don't want to turn it down because they're not turning it down." But the problem was that there was so much traffic that was going through, it was actually clogging the pipes up, especially in Europe. You're seeing there's a finite amount of capacity. What's interesting is that was just at 2x. We really think there's going to be a thousand-x, and so the amount that we're going to have to invest in order to make sure that's done efficiently... And there's some good news around it. Like video traffic actually tends to be pretty heavy; agents don't tend to pull videos. So there are things that are good news there. But there's going to have to be real significant investment in order to keep up with what's there. And I think your question of who pays for it is going to be one of the most interesting questions. The business model of the internet over the last 30 years has been advertising, really led by Google. Bots don't click on ads, so it's not going to be advertising going forward. Someone's going to have to pay for that, and that's one of the things that we're spending a whole bunch of time at Cloudflare thinking about.
So, the pipes are going to be blocked and the websites themselves are also going to get that traffic. So, now they need different servers. You can't be one of those little servers where you expect a thousand hits a month.
Yeah. I think there's a lot of things going to happen. So one is we've just got to take a lot of the kind of aging infrastructure—like if you imagine that you're a town and you haven't invested in your sewer system for a long time, or you haven't upgraded the various things that are here. And again, we're sitting in Park City, so there's a lot of places where we should be investing, we're not, but that's a conversation for later, maybe. But the same thing is true across the internet. There's a lot of old technology which powers the internet that we're going to have to modernize, and in the process of doing that, we actually can make it much more efficient and I think we'll be able to stay ahead of this. But there is going to be real investment; someone's got to pay for it. The business model has been advertising. That's not going to be the business model going forward. In fact, I think the most interesting question maybe in the world right now is: what's the future business model of the internet going to be? What does that look like, and how does it change incentives going forward?
Yeah. Are we going to pay per gigabyte per month?
First of all, let's talk about advertising and the business model that's existed. I will get angry phone calls, even though this is local radio, from Google executives because every time I say this, I get it. But I think you can make a pretty strong case that everything that's wrong in the world today is Google's fault. Let me explain. Google, I actually think, has been a pretty amazing company, done amazing things. But they are an advertising-based company. They make money when they capture your attention and can put ads in front of it. And Google begat Facebook, which did that even better, which begat TikTok, which has done it even better. And we're kind of going down this sort of attention economy hole. If you think about like if you're in the job of media today—I was sitting with Arianna Huffington, who started The Huffington Post, and she was so excited to tell me how they were doing all this work to be able to create content as inexpensively as possible and then A/B test headlines to figure out which ones provoke the largest amount of rage, right? Literally, they were trying to stimulate as much cortisol production in people that were reading these things as possible because that got them to click on them and then they can have an ad. And the best way to do that is to really divide people apart, to say you're different and here's how we're going to do it. So, a lot of media today is actually about how we're going to divide everyone. That's not making us better as a society. That's not actually contributing to things like what you do here at KPCW, what we do over at The Park Record, which is local media, which has gotten just decimated. When my wife and I bought The Park Record, we did it because we thought local news really matters, and we were concerned that the paper was going to go out of business because there just isn't as much money in advertising as there used to be. And so you couldn't thrive and survive and exist. And yet, in order to have a community, you've got to have local news. What I think is interesting is that if we think about what the future looks like, when I talk to Sam at OpenAI or I talk to Dario at Anthropic, what they want to pay for, what they are so excited about, is paying for actual real knowledge creation. And so like at The Park Record, we actually think this year we'll make more money off AI licensing deals than we will off digital advertising. Why? Because if you don't have The Park Record and you're coming on vacation to Park City and you want to know what's the new hot restaurant, or where should I stay in town, or who's playing at the Egyptian Theatre, you don't have that. And if you're going to build a sort of artificial general intelligence, you've got to actually have that local information. And so I think what KPCW does, what The Park Record does, actually is going to be much more valuable going forward in this new world of feeding these AI systems. And we're going to have to have a structure that enables The Park Record, KPCW, to be able to avoid other people just taking it, because I think you're seeing that now. They just take it.
We talk about how markets need supply and demand. That's not exactly right. You need demand for sure, but if you have infinite supply, you don't have a market. We're sitting here in the KPCW studios; we're breathing air. There's no market for air. If I go scuba diving, though, there's a market for air because there's no air underwater, and so there's limited supply. What's been hard for a lot of media for a long time has been they just made everything available to everyone, and the AI companies just took it and were just taking that information. What we're trying to do is work with all of these creators that are out there in order to say: you should have the ability to control who has access to your information, and after that, we can then establish a market for you to do this. Technology doesn't have to just completely decimate creative economies. Look at the music industry. 23 years ago, the entire music industry was worth $8 billion, which is a lot of money, but it was the entire music industry—the Beatles and Rolling Stones, still $8 billion for the entire thing. And the reason why was because things like Kazaa and Napster and Grokster were basically commodifying music. People were just stealing it and no one was paying for it. And then Steve Jobs steps on stage almost exactly 23 years ago to the day and announces iTunes: 99 cents a song. That isn't the business model that initially won, but what was really powerful about it is it said artists are worth supporting, music is worth paying for, and we're going to stick a stake in the ground. Now that turned into, over time, the business model that has won, and most people now have a Spotify subscription or Apple Music subscription, or Tidal. If you go to YouTube or TikTok, they're also paying back to artists for the music that they are using on those various platforms. And today, just Spotify alone last year sent $12 billion back to music creators. And we can debate whether they're sending it to the right people or not the right people, but there's never been a time in human history that more money is going into music creation than ever before. And so the question is, can we take that same model and think about how we do that for more general knowledge creation? The way I think about it is, for the first time in human history, we have a mathematical model of all of human knowledge. That's what these AI systems are. They are actually mathematical models of what humans know, and they're all put into this system. And I think of it like this giant block of Swiss cheese, right? There's a lot of cheese; there's a lot of holes in the cheese. And what the AI companies want to pay for is not yet another story about what happened at 1600 Pennsylvania Avenue. They want to pay for people who actually fill the holes in the cheese. I flew up to Stockholm to meet with Daniel Ek, who started Spotify, and I was talking to him about this and how it might work. He said, "You know the really interesting thing? If you go to Spotify and you search for like Taylor Swift's 'Shake It Off,' we return a result and we're pretty sure we've given our user what they want. But if you go instead and you search for 'I want a song to a disco beat about how much fun it is to dance with my cat,' not a lot of good results." But they take those misses because they know they are misses, and they publish them back to music creators. There's a guy who's in Copenhagen who's literally making—and this is going to make everyone think they're in the wrong business—he's making €40 million a year just writing songs for unfulfilled Spotify queries. And he's not alone; they have hundreds of people that are actually making over a million euros a year writing songs for those things. I think that's really beautiful because if you think about it, you're searching for a song, what you're searching for is an emotion, you're searching for a feeling. And instead of it being just slop, these are people who are writing things that are actually fulfilling. They're filling in the holes in the cheese. And as we think about what that future business model of the internet is, I think we should learn from the mistakes of the advertising-driven business model of the last 30 years and try and find something that actually advances human knowledge for the next 30 years. And again, that's what I'm spending a ton of my time thinking about.
So let's go back to the sort of traditional internet model of shopping, right? I would do a Google search. Your example: I want to buy a digital camera. So I go 'digital camera' and I get like six things, and then I methodically go and click on each one, and how many am I going to look at? Maybe five, right? And one of them I'm going to be suspicious of. That's going to change.
Already changing. Yeah. How's it going to work?
I think it's going to be like the Jetsons. Like if you watch the Jetsons, when George wants, I don't know, eggs, he just asks Rosie, and then Rosie goes and finds the eggs and brings them back. And what's interesting is that's going to have really big implications on a lot of things. One of the things I really worry about is small businesses, because if you think about like why do you do business with a small business, it's usually for one of two reasons: it's either convenience—you have a limited amount of time and so it's the bodega that's down the street or it's the thing that is on sort of your route to drop your kids off at school, and so you stop and use it because of that—or because you have some sort of emotional attachment to it. The guy at the deli counter always remembers your name, and so you go back to the same place. It doesn't necessarily mean that they have the best prices, doesn't mean they have the best quality, but they're going to do those things. Your agent doesn't care about either of those things. Your agent has almost infinite time because of the rate that they can just do things, and they're going to really not prioritize for emotional attachments. They're going to prioritize for what you've actually instructed it to do. And so that's going to drive just very different behavior. And so I've been spending a lot of time with the folks like Tobi, the CEO of Shopify, and the folks at Salesforce, Visa, Mastercard, American Express, to really say we've got to make sure that we build the tools to, as this massive change happens, make sure that small businesses can participate. And the thing that I keep trying to test is, in order to have healthy markets over time, you have to have the ability for new entrants to exist. And what I worry about is that AI systems are trained on the past. They're trained on the established incumbents, and I think those established incumbents are going to have bigger and bigger advantages going forward. And so we've got to figure out some way where if you're a new entrant where nobody knows anything about you, and you can't appeal to convenience and you can't appeal to emotion, how do you break into the market? And that's a really hard, really interesting problem that I don't know the answer for. I'm way better at solving how we get KPCW paid for the content you create. But I think we've got to think about all of these changes and what it's going to bring to society and how we make sure that small businesses can thrive.
So if we go back to Roger's camera analogy, if we have Nikon, we have Canon, will that make the cameras better and cheaper, or will that solidify them as the leaders and then make things more expensive so they can make it, or will Toshiba and Olympus suddenly get better and compete with them?
I think that you're already seeing that consumers are largely winning in this space in the short term. Products are getting a lot better, cars are getting safer, digital cameras are getting cheaper. I think that's just going to accelerate, and AI is going to drive a lot of that. What I worry about is the longer term, because if you don't have new entrants in a market, what over time happens is that you get consolidation. Canon buys Nikon and squishes them all together into something, and you don't have new things that are there. I was meeting with the guy who up until recently was the CEO of PayPal, and before that he ran small business for Intuit. And he said, "I worry in the future that there are five companies: one that builds things, one that holds the money, one that holds the land, one that transports things, and an AI company." And that's obviously extreme, but I do think that left to its own devices, AI could be a massively consolidating factor. And so the future that I'm trying to play for is one where there's not five AI companies, there's 500,000—we make it as easy as possible for those to exist. One where there's not just a small number of really powerful media companies, but one where really anyone can be a content creator and can still get their voice out. And then also one where businesses large and small can thrive and succeed, and new entrants can win as opposed to something where just the consolidated largest players are able to win because they have the best portfolio of what's out there and they can access your agent and convince it to do something. Like the idea of a brand: a brand today is all about emotion. It's a shortcut for you to understand what is the experience in terms of quality or value. Like if I see a Walmart logo on the side of a building, or a McDonald's logo, or a Marriott logo, I know when I walk into that building what that experience is going to be like. Thinking about what a brand is in a world of agentic commerce I think is again one of these really fascinating questions, because it's not going to be some colorful swirls of paint. It's going to be something that sort of helps an agent understand: is this someone that has high quality? Is there value that's here? Do they actually fulfill the promises that they make? And I don't think anyone is really right now thinking through how that's going to exist in the future. And again, we've got to have that if we're going to be able to have new entrants be able to break into these markets and succeed.
Isn't there another possibility though, which is that if the agent can search 5,000 websites and I can project as a smaller business something that attracts their attention, then I might have a better shot at getting this?
So that's the question: how do we make it possible to attract attention? We have to provide the tools, in the same way that when the internet first came out, we made sure that there were tools. It was disruptive to some businesses; it helped other businesses. But there's going to be a bunch of change for this. And what I think is really important is we keep our eye on what does that long term look like. It's not five AI companies, it's 500,000. It's not a few media companies, it's anyone being a creator. It's not these massive conglomerate companies that sort of control everything. We want to make it so that new entrants can get into the space and actually win over time if they have better products and can actually fulfill customer demand better. Those are all the things that I think we have to play for. And as we ask what's the right business model and what are the right incentives, keeping in mind the goal is really important, because left to its own devices, I think you do see massive consolidation around the AI companies, you see massive consolidation around the media companies, and we could have really very few actual businesses, and small businesses might get really hurt.
We talked about local media. What do you think the impact of this is going to be on big media, the existing large legacy companies?
Yeah, I mean, we're already seeing this. So take two great media organizations: The New York Times—again, amazing media organization—and Reddit, right? You might have different feelings about Reddit, but Reddit's a pretty amazing media organization as well. And if you look at their corpus, all of their content, they have the same number of tokens, which is just a fancy way of saying words. So they've got kind of the same amount of stuff, the same amount of information. The New York Times has been publishing a lot longer, so its catalog is a lot deeper over time. Reddit is higher volume, but ends up being about the same. Which one got the better media deal? Which one got paid more for their content? The answer is Reddit, by at least seven times what The New York Times did. And this is again—I already have pissed off the Google folks, so let's piss off The New York Times now. The question of why is because at some level, if you can't license The New York Times, you can just license The Wall Street Journal and then ask AI to rewrite it as if you're a New York liberal and you get The New York Times. Now, that's deeply unfair to the Times; they do incredible work. And I could tell the story exactly in the direction which would be unfair to The Wall Street Journal. But what I think most of media has been has been basically everyone telling the same story to their narrower and narrower tribe. And I think that that business is going to go away. Whereas if The New York Times of the future would actually do not just hotel reviews, but imagine hotel room reviews where you said, "I'm going to stay at the Marriott Marquis and I like to sleep in late in the morning. Am I better staying in room 1313 or 1314?" Well, it turns out 1314 faces north and has better shades, and therefore you should stay in that one versus 1313. Like, that's exactly the sort of thing that is going to be really powerful media: more local, more original, more actual knowledge creation and dissemination, as opposed to kind of what has been much more of the media landscape, which is tell the same story and put your own editorial spin on it.
So, The New York Times originally was print, now they're digital, and Reddit is 100% digital. So The New York Times has to kind of get more into the digital world. I mean, will they stop printing?
The New York Times has been an incredible story around their ability to embrace digital. They've been really one of the success stories that's there. But at the same time, the world is changing now. And what people don't totally... Not to age your listeners, but most people who are listening to this watched as we went from consuming on the internet from a browser, to then social where most of the consumption was actually through like your Facebook feed, to then mobile. Those were all platform shifts. What I think people don't totally appreciate is AI is yet another platform shift. The interface through which you consume information, the interface through which you do commerce, will increasingly look like an AI system. If you look at even Google today, Google doesn't return 10 blue links anymore; it gives you an AI summary. That's going to become more and more of what it is. So what's been interesting is in each of those platform shifts, advertising has survived as the dominant media. I don't think that's going to be the case going forward. I think most of us in the US are probably going to spend, let's call it $1,200 a year on our AI agents. And then those agents are going to do work on our behalf. And if one of those agents, if I say, "What car should I buy or what digital camera should I buy?" and it says, "You should buy a Ford because Ford paid them—there's an advertisement that paid them," I'm going to very quickly switch my business because it's an agent and it has—this is a stronger word, but it has a duty of kind of loyalty to me. And advertising is directly in conflict with that. And so we're having this huge platform shift where people are increasingly going to view content through one of these spaces, and the advertising business model that has supported these things in the past is going to come under increasing pressure because it just isn't aligned with how these things work. And so we've got to figure out what is that future business model in order to support the buildout that Roger was talking about. We've got to actually continue to invest in making sure that the pipes can flow.
Yeah. But another thing is that the legacy media organizations, The Wall Street Journal, The New York Times, we can talk about whether they put a political spin on it or not, but there were curating checks and balances on the truth of what was in the stories. And one of the things that's scary is what happens to that function, right?
I hope it gets pushed to KPCW and The Park Record. I hope it gets pushed lower. And by the way, if you're Sam at OpenAI, what does he want? He wants unique information that no one else has that is highly reputable. And if he doesn't get that, if his systems return bad answers... What's been interesting is none of these guys seem like... One day OpenAI is the best, and the next day it's Anthropic, and then the next day it's Google, and then the next day it's the Chinese models. They're all kind of leapfrogging each other. And so at some level, who wins in this space is actually going to be who gives the right answers. And that's who you're going to be willing to pay for. And so I actually think that there's an enormous incentive for the AI companies to want to get the right answers because that's how they get paid. And so I think that they're going to actually look for more people who are more responsible. I'm actually pretty optimistic. In the future, if what we're paying for is filling in the holes in the cheese, which is just a way of saying we're paying for the actual creation of real human knowledge, think about how much better that is than paying for the stimulation of cortisol, which is what so much of media, or dopamine, which is what so much of media is designed around today. I am super optimistic that if we get this right, we actually do a lot of things to get back to where journalism is about finding what the truth is as opposed to selling what you think the truth is.
Rosie was a pretty benevolent person—I mean, a robot. She wasn't sort of polluted by an advertiser who's paid her to point her in a different direction. When we hear the term 'bot,' it sounds kind of evil. Talk to us a little bit about terminology and how it affects the way we perceive what's happening.
Well, I mean, these are all synonyms: bot, crawler, agent. And it just depends on if you want to describe it as a bad thing. You can just tell people's sort of beliefs towards AI based on whether they're like, "Oh, it's a bot," and then they're kind of the anti-AI folks; if they're like, "Oh, it's an agent," then they're like the pro-AI folks. They're all the same thing behind the scenes. And again, I think we can recognize what some of these challenges are while also appreciating we're living in an age of absolute miracles. You can see that just this last week there was an announcement about how there's new drug development which is being driven by AI, which is curing a ton of amazing diseases. A doctor had prescribed me a series of things; I ran it through AI just as a sanity check on does this make sense, and it came back with really sensible answers that actually then prompted me to ask questions of my physician. I think that these are all things that, again, it's just incredible that you can hold it. My car was having a problem and I just was like, "Hey, here's the error that I see," and instantly it was like, "Do the following six things and it will fix it," and it fixed it. That's a miracle that that happens. So I think that consumers are going to benefit. Nobody wanted search engines. The crazy thing about search engines is it's a lot of work. You think the search is done when it returns the results. The search isn't done; that's just a treasure map and you've got to go find the things. It's better as a consumer to ask things. People don't want search engines; they want answer engines, and then you can get back to doing the things that you love and fixing your car.
yourself as opposed to having to go to the mechanic. Like that's actually better. So I think these are miracles, these technologies that are there, but there are real risks. And as we think about this—and again, I think in the position that we're at at Cloudflare where so much of the internet depends on us—it's up to us to be good stewards and really think through what are the long-term implications and how can we make sure that what we come out of is better for society than what we've had for the last 30 years.
You know, we've talked about some of the challenges being faced by the AI companies. Let's talk a little bit about how the consumer-facing companies of the world are reacting to all this. How are they sort of dealing with this transition from humans to bots hitting their sites—the bots that aren't going to read the ads?
Yeah. I mean, to give you a sense of how nobody knows the right answer, look at four great retailers: Shopify, Walmart, Target, and Amazon. At some level, they're all in the same business. They're all retail or retail facilitators. They obviously have slightly different business models. They kind of come at the problem from different directions, but they all have radically different strategies on this. In Shopify's case, they literally are unifying across all their merchants a catalog to be able to publish it back and inviting agents to come in—come one, come all. On the opposite end is Amazon, who is literally suing any of the agent companies, saying, "You're not allowed to shop at amazon.com."
And then in between, you've got Walmart that kind of is waffling between... at first they were like, "All the agents are welcome," and then they saw that that didn't work so well, and so now they're kind of retreating from that. And Target is like, "Agents are welcome as long as they are specifically blessed and they're run from, you know, target.com." That seems like the worst strategy of the four to me. But all of those... I don't know which is going to be the winning strategy, and it might be that the winning strategy is something in the short term, but that it backfires over the long term. And so I think that right now, if you're in business and you're saying, "How do I respond to this?" and you don't know the answer, know that you're in good company, because I think a lot of people don't know what the answer is.
And again, I think that what we've all got to sort of articulate is in the future, what do we want? And what I want is a world where there's not five AI companies or 500,000. It's a world where there's lots of different media outlets and anyone can be a content creator, and one where new small businesses can start and thrive and win over time, and we have lots and lots of successful businesses. I think if we start articulating that, then we can ask, "Are we moving toward that or away from that?" and that can help be the barometer of how this is going.
Now, can't the bots, instead of shopping at Amazon, go past Amazon and go to the distributors or go to the manufacturer themselves, thus giving you even a better deal?
I mean, that is I think what Amazon's concern is—that at some scale, they can just get disintermediated from the relationship, because if you just go to a bot and type something in, then you don't have a relationship with Amazon. They're just sort of a merchant that's out there. But if they can find a better price somewhere else for the same product, then they're going to go around that. And so I think figuring out what is your distinctive advantage in this world that's coming, where you may not have a direct relationship with a lot of your customers because that relationship is going to be through Rosie the helpful robot—that's going to be the sort of thing that I think is one of the real challenges that businesses are going to struggle with.
