Cloudflare CEO: The Man Deciding The Internet’s Next Business Model

I’ve got the CEO of Cloudflare, Matthew Prince, on the show today. Matthew, welcome in.

Thanks for having me.

How’s life?

It’s good. I’m in Park City, Utah. Fall in Park City is beautiful, so it’s great.

Yeah. Nice. Well, I’m very excited to have this conversation. I think there are a couple of big topics looming on my mind. One is, basically, you’ve been running this company for 16, 17 years now, so I want to hear what it’s like to be in year 16 or 17 running what feels like a very AI-first company.

1. How The Internet Got Here

Before that—and we should definitely talk about all the work you guys are doing in crypto—I think maybe the best place to tee up this conversation is around the evolution of the internet. I think you’ve had a better view into how the internet has evolved than nearly anyone standing on Earth right now.

Maybe just give me your take on the past 2 decades. Give me the run-through of the evolution of the internet and take us up to this really weird point in time that we find ourselves in today.

Yeah, it definitely is a weird point in time. I feel incredibly lucky that I was born at a time where I got to experience the internet and kind of grew up on it. Back in 1993, when I was in college, I helped create what was one of the first online-only news publications. Wired gave us an award, and we were supposedly one of the first 1,000 websites that was publicly accessible.

To see where we are today, when there are literally hundreds of millions of sites and the internet has just exploded—and even this conversation that we’re having is being carried over the internet—is incredible. Something like this, which even just a few years ago would have cost tens, if not hundreds, of thousands of dollars to produce, we’re able to do very efficiently today, which is just an incredible miracle.

I think what’s been interesting, though, is that if you look back over the last 20 years of what the internet’s been, it’s really been driven by a business model that is around advertising. Ads have been a huge and incredibly successful way to make content available online. If an ad is served to me, I’ve got a lot of disposable income, and I buy stupid stuff from time to time, so there’s a really high CPM for that—basically, a price that you pay for those ads.

But if there’s a kid in Sri Lanka, even though they may not get the same kind of payments for that, they’re still going to get something. Ads have really been the only micropayment system that has worked at real scale, and it’s carried the internet to what it is.

I think the hero of the last 27 or 28 years of the internet has been Google, which has really built that ad ecosystem. We all think of search, but they also built Google Analytics, DoubleClick, and all of the infrastructure that really powers that ad ecosystem. Without the economic engine that they created, we really wouldn’t have the web in the form that it is, and we really wouldn’t have the internet in the form that it is. That’s been an amazing transformation.

Last November, I was at Web Summit in Lisbon, and someone asked me, “When do you think that automated traffic—that is, agents and the various things that run around on the internet—will pass human traffic?” I went back to my team, we pulled all the data, and based on everything that we saw, we thought it was going to be the second half of 2027.

Fast-forward to March of 2026. I was at South by Southwest and got the exact same question. We had pulled all the data again, and we were like, “Oh, wow, it’s moved up.” Now it looked like it was going to be in the first half of 2027. Then my team came to me in May and said, “You won’t believe this, but we’ve actually already passed automated traffic over human traffic.” It’s wild that it has continued to grow like that.

With the big caveat that I’ve been wrong in every prediction so far, if we see the current trends continue at the rates they’re going, we now think that agentic traffic will be 1,000 times as much as human traffic in 5 years. That’s just incredibly remarkable. In the not-so-distant future, humans aren’t going to decrease how they use the internet, but bots are going to increase so much that humans are really going to be just a rounding error online.

I think that’s wild to think about, and the implications of that are pretty substantial. If the internet was really built on ads previously, the problem is that bots don’t click on ads. The entire business model of the internet is about to change radically.

What that changes, too, I think has not only amazing and profound implications for what the internet looks like, but it’s going to have really deep implications for media generally, how society is organized, and a lot of the polarization that we’re experiencing today. I think the most interesting question in the world is: What’s the future internet business model going to look like? That’s where I’m spending a lot of my time.

The thing that I find so interesting about this is that there are obviously great questions to be had about what businesses will do well, what businesses will not do well, and how you position your company. But from a societal perspective, we had this open internet—you were talking about the ’90s—then AOL comes around and starts to centralize the internet. Social networks have basically entirely centralized the internet.

Social networks combined with Google have centralized a lot of the internet, and that has all these societal implications. Maybe it’s driven a lot of the polarization, like you mentioned. Think of how many small businesses and bigger businesses are just set up on Instagram ads and Facebook ads—the whole ad network of the internet.

How do you think this impacts society as a whole? We’ll get into the business component in a second.

There are parts of the internet that have gotten really centralized, and there are parts of the internet that are still remarkably decentralized. I’m actually really proud of the role that Cloudflare has played.

At some point, there was an investor early on who said, “Who’s your biggest competitor?” I said, “Facebook,” and they looked at me like I was crazy. But I think that’s actually kind of true, because when Cloudflare was first getting started, Facebook was gobbling up a huge amount of the internet. It was easier for you to be behind Facebook than to have your own website.

Part of the reason for that was that having your own website was hard. You had to deal with security problems, performance, and all this other stuff. People were actually giving up on creating those things.

Wait, to make sure I understand: In the sense that instead of going and creating your own website, your own app, or your own blog, it’s like, “Just launch it on Facebook.”

A distribution Facebook page. I mean, in certain parts of the world—in Nigeria, the internet is largely Facebook. There just isn’t as much of that ecosystem set out there.

I think Cloudflare is to Facebook as Shopify is to Amazon. We’re taking some of the parts that are really difficult for people to deal with—security and other things—and helping make that easy. That has actually made it easier to create the internet.

I think part of what’s really amazing right now is that if you look from about 2012 through about 2024 and early 2025, the amount that the web grew was minuscule. In fact, for a period of that time, it actually shrank. There were more websites being taken offline than there were being created.

What changed in really late 2025 was largely because of AI. People who never thought of themselves as developers were suddenly able to create websites and applications and put them online. We’re actually seeing growth of the internet that looks closer to what internet growth looked like in the early 2000s. I think it’s been a pretty amazing time for all this to happen.

Back to your question, though, in terms of societal impacts, if you think about it—and I get in trouble every time I say this, so why not just say it again—I think at some level everything that’s wrong with our world today is Google’s fault.

That’s an overstatement. Google, I think, is actually an amazing company. They’ve done amazing things. But they did create this business model that was all around, “How do we capture people’s attention and then monetize that attention?” That begot other systems like TikTok and other things that are really just trying to capture people’s attention and sell it.

The best way to capture people’s attention is either to generate dopamine or cortisol. Make people irrationally happy or irrationally angry. A lot of what media is designed for today is to push one of those 2 buttons in your head so that you can then sell an ad against that.

What I’m actually optimistic about is, as we think about what the future business model of the internet is, instead of it being about how we get humans to have a fight-or-flight response, how could we instead reward actual knowledge creation? I’m pretty optimistic about what that looks like and, again, what the positive impacts on society could be if we get that right.

2. Why AI Labs Fear Google

Yeah. I would love to hear more about your relationship with some of the big labs. I’ve heard you talk very publicly about Google. I was reading earlier this morning about your push to have them, I think it was, unbundle search from AI and their AI crawlers, if I understood that correctly.

I haven’t heard you talk much about your relationship with and your view on some of the other labs—OpenAI, Anthropic, and it’s debatable whether you throw Perplexity into there. What is your stance on these? I don’t know if you can bundle all of them together or if you view them as separate entities here.

Yeah, they’re really quite distinct and all have different personalities. But the one thing that I think all of the major labs share is an incredible fear of Google.

Think about how OpenAI started. OpenAI started after a dinner conversation between Sam, Elon, and Larry Page. Larry basically said, “Listen, if AI out-evolves humans, then humans deserve to be extinct.” Sam and Elon came to humans’ defense and said, “You have kids. We have kids. Come on.” Supposedly—and the guy wasn’t at the dinner, but this is how it’s been reported—Larry called them speciesist, basically saying, “How dare you defend humans if the machines are better?”

I think there was almost a religious aspect to the founding of OpenAI: “We’ve got to do something, because otherwise Google is going to run away with this game.” Google had all of the resources to run away with the game. They had the best researchers through DeepMind and some of their other areas. They had the best silicon. They had access not only to chips from people like NVIDIA, but they had built their own TPUs and their own silicon, which were hyper-efficient. And they had the most data.

Even today, Google sees twice as much of the internet as OpenAI does, and that gives them an enormous advantage. I think there will be many books written on how, even with all those advantages, Google has kind of screwed it up so far. But I would certainly not count them out.

It’s important to recognize that every other AI lab exists in part because it started as almost a religious opposition to that attitude inside Google. That’s created some interesting challenges and opportunities.

We have an amazing relationship with them. Eighty percent of the major AI companies use Cloudflare. We have great relationships with and they’re customers of both OpenAI and Anthropic, as well as X and Grok and the whole group, including a bunch of the Chinese labs. We talk to them all the time.

One of the important points, as we think through what the next business model of the internet is, is that, sure, we get it—we’ve got to pay for the data that we’re taking. But you’ve got to make sure that everyone else pays for it as well, and that includes Google.

It can’t be that the number-one competitive threat that they all worry about, and that they have almost a religious opposition to, gets all of the data on the internet for free while everybody else has to pay. A lot of the work that I’ve been doing at Cloudflare is to ask how we can make sure that 2 things happen.

First, we need to make sure that Google plays by the same rules as everyone else. They shouldn’t be able to leverage their monopoly of yesterday in order to get a monopoly tomorrow. Secondly, how can we help folks like OpenAI, Anthropic, and others get back to the same level that Google does?

Google sees twice as much of the internet as OpenAI, 4 times more of the internet than Anthropic, and many times more than someone like X or Meta. If we want to play for a world where there are as many successful AI companies as possible—and I don’t want there to be 1 or 5; I want there to be 500,000 amazing AI companies from all around the world—we have to make sure that everybody has a level playing field to play from.

Again, I think that’s a big piece of what we’re trying to think about as we think about a healthy ecosystem here.

Yeah. This might be a false dichotomy, but I saw you sign—I think that you signed—the letter saying, “Don’t regulate the open-weight models.”

But I also think that, hearing you speak, you want the AI models to pay for training content. I’m trying to think if that’s a false dichotomy. How do those 2 positions coexist?

I think you can have open-weight models, and there are businesses built behind those, just like there are open-source companies that build great businesses. But you still need to compensate creators.

Take it from this: Let’s just imagine that, in the future, agents are creating 1,000 times as much traffic on the internet as exists today. Who’s going to pay for that?

There are a lot of things we can do just to make the internet more efficient. We’re working on taking old pieces of software that aren’t as efficient, like WordPress and other things, and rewriting them to make them significantly more efficient. We do a lot with the free version of our service and everything just to make sure that people can have it.

But at the end of the day, there’s real bandwidth, real servers, and a real cost to serving that. If you don’t put some cost in place, then you have this massive tragedy-of-the-commons problem. If I want to say, “Where should I go to lunch?” and the cost of scouring the internet is effectively free, then bots are going to literally scour every page on the internet and come back and say, “You should go to Wendy’s.”

That’s totally inefficient. There should be something there. We have to figure out some way for the people who are getting the benefit from scanning the internet to be the ones actually compensating for the infrastructure, as well as for the work that goes into creating the original content that powers these bots.

There’s no way we’re going to be able to keep up with 1,000 times the traffic unless we create some business model that helps create and pay for that infrastructure.

3. What Cloudflare Actually Does

Interesting. Let’s talk about the business model. Before that, I’m very familiar with Cloudflare because I think it was one of the first vendors we used. Our entire business runs on Cloudflare, which is great, and I thank you for that.

4. Ads (Token2049, Avalanche Summit)

For people who may never have built software products, Cloudflare may be a little less familiar to them. Can you give us almost a state of the union on the scope and scale of how big Cloudflare is and how many internet businesses run on top of Cloudflare?

At its core, Cloudflare is a network. We run one of the world’s largest networks, spanning the globe across more than 350 cities, with locations within milliseconds of the vast majority of the world’s population. In all those places, we’ll have racks and racks and racks of servers.

We perform a number of functions. The things that people are probably most familiar with are that we help protect anything you’re putting online from various types of attacks, whether that’s cyberattacks, DDoS attacks, or other attacks. We protect that.

We also make sure that things are as fast as possible. If you’ve got someone visiting your site from Pakistan, someone else visiting it from Virginia, and someone else from Japan, regardless of where they are, we’re going to deliver it in the fastest possible way.

Over time, we've added additional functionality that helps protect internet users as they go out and use the network. They basically pass through us. More recently, starting in 2017, we started working on how we could make it so that anyone out there could develop software and run it on our infrastructure in order to be as fast and efficient as possible.

Today, that's one of the fastest-growing areas. If you use almost any of the vibe-coding platforms, whether it's Lovable, Replit, Base44, Codex from OpenAI, or some of the tools from Anthropic, when those agents build software for you, the most preferred place for that software to run is typically on Cloudflare's network.

That's been incredibly rewarding. Today, a lot more than 20% of the internet sits behind Cloudflare, and almost half of the Fortune 500 is built on Cloudflare. In your world of crypto, it's one of the places where almost all of the major crypto companies are building on Cloudflare because they get to start with a new paradigm and a blank sheet of paper. When developers are able to do that, what we find time and time again is that they're choosing to build on the Cloudflare developer platform. It's been amazing to see what people in both the crypto space and, now even more so, the AI space are building on top of us.

5. When Cloudflare Went AI-First

Yeah, 100%. When everything got really crazy, ChatGPT came out in Q4 of 2022, I think. There are these moments in my head—I think it was May of 2025 when Claude Code came out, and Tobi's letter to all of Shopify is another moment in time for me. When you saw some of these moments, were you thinking, “This is the evolution of cloud. This is what we've been building for,” or, “We need to completely pivot the business, go all in on this, and reinvent ourselves”?

I wish I could say we were so clever. We started working on the developer platform back in 2017, way before the real rise of AI. AI was still kind of a backwater and a laughingstock. It wasn't really the place.

Machine learning was the hot thing.

Yeah, there was some of that. I remember, in a VC pitch once a long time ago, we described Cloudflare as an AI company, and we got laughed out of that room so badly that we almost have an allergic reaction to saying it. At some level, that kind of was what we were doing. If you can get enough of the internet flowing through you and then run AI or machine-learning algorithms against that, you can do a lot of things to predict security threats, enhance performance, and do a bunch of other things.

As we saw AI rising, we were happy selling picks and shovels to the people who were adopting it, but we were a little nervous about leaning in too much. It wasn't totally clear at first what the ROI was going to be for a business. Secondly, we were really paranoid because we didn't want to take a bunch of our internal data and ship it off to some company that we didn't know very well.

I think we were actually slow adopters of it, but sometime in the middle of 2025, it became blindingly obvious that there was something here that was impactful beyond what we'd already seen with our engineering teams. It was impactful across the entire company. The real turning point for us was probably November of 2025, when we really leaned in and said, “This is the point where you actually have to restructure how a business is organized and run to make sure that we can take advantage of AI and use it for the things that it's really good at.”

Yeah, we'll get to that in a bit. I read your Wall Street Journal op-ed, which I definitely want to talk about. Before doing that, was there a moment in time when you thought, “This is how we have to rearchitect the company and the products”?

I felt like late October or early November of 2025 was the point at which it was—

Opus 4.6 had just come out.

Less around rearchitecting the products, more around rearchitecting the organization. I think we'd, through somewhat dumb luck, built the right architecture. To give you an example, the hyperscalers—AWS, Google Cloud, and Microsoft Azure—were really built for what was the mobile era.

If you look at the evolution of where software is run, back when I was in college, if you wanted to stand up a web server, we literally had to call Gateway, the server manufacturer, to ship us a physical box. That was going to be our web server, and it took months for it to arrive and for us to get it all wired up and the software installed. Then VMware came along with virtual machines that took one box and made it into many, so you could actually get more efficient.

Then folks like Docker came along with containers, and really the whole hyperscale world and all of the mobile world are built on this idea of containers. The problem with containers is that if you imagine every single knowledge worker on Earth has one agent running for them—which seems pretty conservative; I think most knowledge workers are going to have more than one agent—and you imagine those agents are all running in containers, you have to import a whole operating system and toolchain, all this stuff, just to power them.

Not even looking at GPUs, just looking at CPUs, you need 40 times the number of CPUs produced in the world just to run the agents for those knowledge workers. That's not going to work, so we've got to do something else.

What we did at Cloudflare was ask, “What's beyond containers? What's actually more efficient than containers?” The answer is a technology called isolates, which most of your viewers will be familiar with. It's what powers the tabs in your browser, basically, and the way you can have lightweight systems where you can very quickly open a new web page, which is really an application.

We took that same idea and applied it, and what that allows is much faster, much more efficient deployment of code. That's exactly what agents need. We happened to build the right technology for this moment in time, and it's amazing to see how many of the agents users are using are either running in isolates on Cloudflare's network or, as they're generating code, are generating that code and running it in isolates. It's much more efficient, much less expensive, and a lot faster.

6. AI Is Flattening Companies

Yeah, 100%. Let's talk about how you run the business. There was a moment that I think a lot of other founders and CEOs looked at. I feel like there are these moments in time for AI. I mentioned the Tobi Shopify one, and your Wall Street Journal op-ed was another one that I saw passed around a lot in founder group chats.

If I remember correctly, you guys had this monster quarter in Q1. Correct me if I get any of this wrong. Monster quarter in Q1, business is ripping. I think it was early May—late April, early May.

You guys announced this huge, record Q1 and then proceeded to RIF, I think it was around 1,000—1,100 people—which was roughly 20% of the company. If I remember the Wall Street Journal op-ed correctly, there was a line in there that got passed around a lot: “You are builders, sellers, and measurers.”

You said the builders make the things. Maybe these are product people, engineers, and designers. There are the sellers: maybe the marketers, salespeople, account managers, or whoever it may be. Then there are the measurers.

If I remember the point of your op-ed, it was that AI can now do all the measuring. The AI does the audits, finance, legal, compliance, and middle management. First off, correct me if I got that wrong, but second, how did this happen? You were not very AI-pilled, and 6 months later, you’re laying off 20% of the company. How did we get there?

I think, first of all, we just started adopting it, and we were very realistic about what AI is good at and what it is not yet good at. I’m not in the camp that thinks everyone’s going to lose their job and there’s no hope. We’ve seen how much more productive our developer team is when assisted by AI tools. If a developer is now 10 times as productive, there’s so much for us to do that I’m going to hire as many developers as I possibly can.

On the other side, in sales, a great salesperson’s favorite thing is to spend time with customers. We found that, pre-AI, most salespeople were spending maybe 40% of their time with customers, but about 60% of their time was spent building presentations, filling in Salesforce records, and doing all of that kind of back-office work, which nobody really loves. We use tools so that now most of our sellers are spending 90% of their time with customers and much less time actually building those tools. That’s driven significant increases in productivity, which has been amazing.

I think what we realized was that there are a bunch of things about a business that you need to measure, and you need to measure them accurately. The problem with humans is that they have biases, and especially the problem with humans on a team is that they tend to have shared biases across the team. That actually influences how accurately and how well humans can measure things. AI has biases, too, but what’s interesting—and I think actually under-talked about—is how those biases are uncorrelated with the rest of the organization. It’s a true, outside, fresh set of eyes to look at various problems.

If, in the past, the Harvard Business School recommended number of direct reports for any manager was supposed to be around 6, we’ve found that you can actually have a better manager relationship with your direct reports. You can be fairer to them, give them better feedback, and see when they need help more accurately if you use these tools. We found that you can take somebody who in the past would have had 6 direct reports and actually make that 12.

What’s powerful about that is that, as you do it, the organization flattens. In fact, if you want to measure how hierarchical an organization is, the easiest thing to measure is the average number of direct reports in that organization. The larger that number is, the flatter the organization is. The smaller that number is, the more hierarchical it is. We were around 6, and we’re making our way toward 12 today.

As we’ve done that, it’s actually flattened the organization enormously, which meant we just didn’t need as many middle managers because the managers were more efficient at doing that. Someone who was a middle manager at Cloudflare would be a great senior manager at some startup.

What was amazing was—I mean, it sucked. It was incredibly emotional. There were really dear friends of mine that we let go. They’re very talented people, and what I’m proud of is that we worked to make sure as many of those people as possible got placed in really great jobs. They’re going to go on and do great things. Some of them are starting companies, and some of them have gone to work at other great companies.

We really prioritized that, in addition to giving what was incredibly generous severance, continuing to vest stock, and doing a bunch of things that were pretty unusual. What has been surprising to me is that, when I talk to my peers at other companies, especially other large public companies, they’re all like, “Yeah, we’ve got to do that, too.” But they don’t want to stick their necks out because they saw our stock go down when we did it.

It went right back up, but I got all that. I got a bunch of death threats. There are a whole bunch of things that don’t make these things fun to do as a leader. But honestly, that’s what leadership is. When you realize you’re going to have to make that change, you have to do it.

What I know is that, when we did that at the beginning of this year, it was much easier for somebody to get a job than it is today. As more of these companies start to make these changes, it’s going to get more difficult. As a leader, once you realize you need to make these changes, every day you wait is actually a disservice to the employees you’re eventually going to lay off. That’s just chicken leadership at the end of the day.

The right thing to do is, once you make the decision, take the hit. It sucked, and I got lots of arrows, but I’m proud of the fact that we’ve been able to place as many of those people—great people—in great positions elsewhere. I know they’re going to create amazing new things going forward.

Today, Cloudflare is running more efficiently, flatter, and just better than we ever have before. That’s because we’ve got great people building products, great people selling products, and fewer people doing the measuring because the tools we have are significantly better at doing those things.

I have a very good insight into what crypto founders are doing, but not as good of an insight into what public company CEOs are doing. What do you think happens over the next several months? Do you think this is very early days and a lot of your peers are going to follow suit?

I think everyone is. To date, we’re one of the few companies that did it from a position of strength. Most people are doing it because they felt like their business was threatened in some way or another. But everybody is looking around and seeing the same thing.

There are some reasons why. Because we started with a security background, we could figure out ways to integrate these technologies safely into the work that we did in a way that I think some other companies are struggling with. But people are going to figure it out, and the harnesses that are getting built are really robust.

We open-sourced the one we call Cloudflare OS because it’s been so valuable to us. I think, as that happens, there’s going to be a significant shift across companies. Again, it doesn’t feel good, but the kindest thing you can do if you’re the leader of any organization, if you know you’re going to have to restructure how your business works, is to do it as soon as possible.

7. What Happens To Human Work?

Walk me through what you think happens maybe 3 to 5 years from now. One thing I’ve noticed at Blockworks is that, usually with headcount, everyone wants headcount in a company. Everyone wants to add more people, and everyone wants the budget.

We just had the first person tell me—I said, “You really need to hire for your team.” They said, “I’ve got 8 agents right now. Don’t give me another human. I don’t want another human to manage.” He said, “I hate human management, but I’ve never had more fun managing these agents.” Give me your sense of where all this goes.

It’s wild how many people we see internally who were managers and are actually saying, “No, no, I want to go back to being an individual contributor.” I think what it means to be a manager and what it means to be an individual contributor is going to get really funny and blurred over time.

It’s wild to see all these really senior people going to places like Anthropic and saying, “No, I want to go back to being an individual contributor.” I think that’s going to spread across this. I’m not a doomer who thinks we’re all out of jobs.

I think, again, if you're a talented engineer, there's always going to be work to get done. You're going to use tools in order to do that more efficiently, but I think the world is bright for that. Over the course of 2026, we were hiring 1,111 interns. It's just incredible.

You are about to hire 10,000.

We start at the beginning of the year. If you go over the course of the entire year, it'll be over 1,000 interns.

Why are you hiring 1,000 interns?

Because they're amazing. They're all AI-native. Everybody else has stopped hiring interns, so we like to zag when everyone else is zigging, and they came in and were just amazing.

Usually, we think our job is to train the interns on what it's like to have a real job and do all that stuff. We told them this year that, yes, that's still half, but the other half is that they have to train us. We've taken interns and embedded them in senior teams to show, “Hey, have we tried this new tool, or have we done this thing?” And it's worked incredibly well, with knowledge transfer in both directions.

I spent a ton of time just having dinners with interns in all the different cities where we had interns. It's wild how talented the group was. I'm actually pretty bullish that we're going to have great people who are building things. I'm pretty bullish about selling, too. There’s going to be agent-to-agent commerce and things like that, but if you're doing a $100 million software deal, that's still going to be humans.

Yeah. Enterprise sales has never been more important.

Yeah, yeah, yeah. And so I think those two areas are really big. If you're getting a degree right now in accounting, I'd be a little bit more nervous. I think there's a ton of work to do. Places like us are hiring like crazy, especially builders and sellers.

I think that's actually the most fun part of a business that's out there today. So I'm optimistic. I'm not in the “it's going to take everyone's jobs” camp, like Dario and some of the people in the labs are. I'm much more in the camp that it's going to refocus how we run businesses. It's going to put more and more importance on building things and selling things, and it's actually going to make us better at measuring our businesses as well.

8. Why Public Markets Work Better

Yeah, I agree. How has it been to manage investor relations and the public markets? I always thought that crypto would grow up one day and become more like the traditional capital markets, and volatility would get suppressed. What has happened to me is that the equities market has become more like crypto.

Crypto has always traded on a lot of narratives, and the public markets are no different right now. I was cracking up at Benioff's launch of Agentforce—I think he called it—or, excuse me, Agentforce. It's really just a connector or an integration. It was beautiful marketing. When I saw that, I thought, “The marketing there is for the public markets as much as it is for the customers.”

I'm very curious how you've managed, and how you think about managing, public-market investors.

I’ve loved being public. Again, to use a bad analogy, the rates of spousal homicide are significantly lower in jurisdictions that have no-fault divorce. Whereas, if you make divorce hard, one of the things that happens is spouses kill each other, which is obviously not good.

The problem with private markets—private equity, venture capital, and all those things—is that there’s no no-fault divorce. It’s really tough to get rid of your venture-capital investors. It’s really tough to get rid of your private-equity investors. By the way, it’s really tough for them to get rid of the founders and CEOs of the companies they run. It’s painful.

As a result, you find yourself in a weird situation where nobody tells the truth. Everyone’s hedging and couching. I loved our VCs, and I loved the folks there, but public markets are so much better. If I say something stupid on this podcast, a bunch of our public-market investors will call me and be like, “That was stupid. I sold your stock, and here’s why it was stupid.” Half the time I’m like, “Yeah, you’re right. That was stupid. I’m sorry,” and hope you’ll keep following us.

The other half of the time I’ll be like, “No, it wasn’t stupid. Here’s the strategy behind it, and let me explain it.” We can have really honest conversations. I’ve really liked that. It’s a much more truthful relationship.

The other thing that everyone warned us about was that public markets are incredibly short-term-oriented. They only care about your next quarter and all those things. That’s just not been the experience we’ve had. We have these amazing investors. There’s a Scottish investor called Baillie Gifford.

Yeah.

The average holding period is 16 years. The average holding period for a VC fund is less than 10 years—I mean, VC funds are only 10 years, 12 at the absolute max, right? You’ve got these public-market investors with incredibly long time horizons who are really focused on what the strategy is.

If you listen to a Cloudflare earnings call, our CFO doesn’t get very many questions. Everyone is focused on, “What’s the strategy? What’s the future?” I think you can run businesses that are very long-term-oriented in the public markets and have a lot of success, as long as you continue to deliver innovation and growth. Again, I think that’s something we’ve done a good job at, and hopefully we’ll be able to continue to do that.

What do you think is the most misunderstood aspect of Cloudflare?

I think our investors probably understand us sometimes better than our customers do. A lot of our customers still think of Cloudflare as the security CDN that’s out there. That’s such a small part of the overall business and what we’re doing.

The other thing that was interesting is that, because we started out servicing the long tail, it took investors a while to realize that most of our revenue comes from really big enterprise customers. Those are probably the 2 areas where customers still see us as doing just 1 thing and don’t understand how all the pieces fit together, and investors sometimes think we’re too SMB-focused. We’re like, “Yeah, it’s less than 10% of revenue today.”

9. Who Pays For The Agent Internet?

Yeah. Maybe let’s get into Cloudflare, how the business has evolved, and where it’s going. I was doing a little prep reading for this episode, reading your blog posts and the product announcements. It became very clear to me that you were building the full stack for the machine-internet economy, or however you want to call it.

You’ve got Pay per Crawl—the tollbooth, right? Pay-per-crawl and pay-per-use, a monetization gateway. You’ve got the wallet. On the crypto side, you’ve got the browser, the identity layer, and the stablecoin with NET Dollar. What is the end vision for what you’re doing here?

It’s funny. I was watching The Dark Knight last night, actually, and there’s a scene where the Joker is with Two-Face in the hospital, and he’s like, “Everyone thinks they have a plan,” right? I kind of empathize with that. I’m not sure. I think people assume that there’s some grand unifying plan here.

Really, we’re responding, first and foremost, to what we need ourselves to build—sort of tooling. We’re building the things that our own team needs in order to continue to build, and that drives a lot of what our customers ask us for. Then I think right now we’re trying to figure out what that future business model of the internet is. At some level, it has to be—

I mean, again, if you worry about—

The tragedy of the commons problem, where your agent has an incentive to just crawl the entire internet, no matter how much that costs, because the costs are being borne by someone else. We’ve got to flip that around. The agent has to pay, but it has to pay tiny fractions of fractions and fractions of fractions. The most interesting thing, I think, that we’re working on internally—and, by the way, trying to find anyone to partner with—

We have no pride of ownership here. We don’t need to own how this works. We’d love to partner with someone if anyone has solved this problem, but we have yet to find anyone who has. We’ve got to do something that can process micropayments at incredible scale.

To give you some sense, the public number we’ve released is that Cloudflare handles about half a billion web transactions per second. We think between 1% and 10% of those will be monetizable through some sort of microtransaction. So that means you’ve got roughly 10 million financial transactions per second that you’ve got to do, very, very small. It has to be super efficient, because you can’t have a lot of transaction costs associated with that, but you have to be able to do it at scale.

Visa, which is the world’s largest payments network, handles about 80,000 transactions per second at peak during the holidays. We need something that’s literally 3 orders of magnitude larger than Visa and probably has to scale to 100 million transactions per second in very short order. On the busiest day in the history of all the financial markets, there were about 2 million equities transacted per second at the peak. So we’re still an order of magnitude larger than all of that.

That’s a really tricky problem. But we have to solve it because, if we don’t, I don’t know how we’re going to pay for the 1,000-times traffic that’s coming from agents, and how we’re going to do it in a way that’s fair to the people who are putting content online and still gives them an incentive to put that content online. That’s some of the infrastructure that we’re working on, either building ourselves or hopefully finding something that we can partner with.

It’s turned out to be—I thought there would be a lot of people who’d been solving that problem. It feels like a lot of people in the crypto space and everything else are shooting to be the next Visa.

I think the real opportunity is to be the white space that Visa can’t even handle. You can’t do micropayments—

—on a network like Visa. It’s just too expensive.

Yeah. I’m going to try to find this blog post I wrote. I wrote this in 2018. I’ll show you this. We launched the business—we launched Blockworks in late 2017. This is June 2018. I said, “The Future of Micropayments Is on a Blockchain.” I said, “Before you know it, your browser will have a crypto wallet extension that you will use for micropayments across every form of the web.”

I think this is a good reminder that things tend to take much longer than you think they will, but—

They do. Again, I think there’s going to be a role for blockchains and crypto, especially for getting things into and out of the network.

Yeah.

The question is, is consensus fast enough?

Well, that was actually my next question. I wanted to ask you: there are no blockchains set up to support what you need. You guys launched the 2 main crypto things that I’ve seen you do. You have the wallet and you have a stablecoin, right?

Well, we have an idea of a stablecoin.

We have NET Dollar. Yeah, an idea of a stablecoin.

We have a name.

A name. You have a name. Yeah.

What we were really trying to do when we announced that was kind of flushing the pheasants out of the grass. We were trying to get people to talk to us. Over that time, we’ve met with everyone in this space.

Come talk to us. Yeah.

Come talk to us. We were like, “Hey, we’re doing this. Help.” I’ve spent a ton of time, and there are a lot of smart people out there. They might be building incredible businesses.

The problem is that they just can’t—the volumes that we’re talking about are really tough to figure out. The largest vendors that we’ve seen in the space and potential partners are getting up to about 2 million transactions per second. It’s usually some sort of hybrid, which is kind of blockchain-based but also has other things—batching, pooling, and other ways to get up to the speeds that you need.

But that’s still an order of magnitude slower than we need on day 1, and we think very quickly it’s 2 orders of magnitude slower than we need. So we have to have some way to scale that. I think the problem is that we’re kind of a customer of 1, at least at first. Over time, I think you’re going to have other people, but that’s what we’ve struggled with.

I think a lot of people are saying, “We can do Visa-like payments much more efficiently,” and again, that’s amazing. A lot of people should be going after that. But saying, “No, we’re going to make it so that every time an agent accesses a resource online, there’s a tiny fraction of a penny that they’re paying,” which is both helping support the infrastructure cost and rewarding the creator of that content—that’s a much more niche, specialty area. But I think it’s a really interesting one.

If someone’s listening and says, “Hey, I know how to solve that, or I have solved that,” we’re still very much open to that while internally also trying to figure out whether we can build this ourselves.

That’s what I was thinking about when I was thinking about the problem you’re in. It’s a build-versus-buy—or maybe not build-versus-buy, but do you do it internally or do you work with a vendor? Basically, there is not an L1 blockchain that can support 100 million transactions a second. So why not go build it yourself? Even NET Dollar, right? It’s like—

Yeah. We’re running both of those things in parallel. Building stuff takes time and comes with risks, but we have a team that’s working on that and hiring like crazy for that team. At the same time, we’d be totally happy if someone came out of the woodwork tomorrow and said, “Hey, I have a solution.” We would either partner with that firm or acquire them.

You know what you could do? There are a lot of L1s that actually have phenomenal technology and maybe could get semi-close to this. They’re an order of magnitude off, but their tokens are just in the dump. You could do the first ever—I mean, first big token M&A deal.

M&A deal. Again, I would not be surprised if there’s some team that’s thinking about that. But the key is you’ve got to have the real technology that can keep up with this.

We’re not religious about this. It doesn’t necessarily have to be a blockchain. It would be great if it is, because those are great things. I do think that getting funds out of the system and getting funds into the system obviously makes sense using a lot of the rails that already exist. The most efficient ones are going to be in crypto, and that’s going to make a ton of sense.

But whatever it is that processes the high-velocity transactions, I think that’s a very difficult computer science problem. We also have to play to our strengths. Some of the companies that have done this in really smart ways have built it with beefy, heavy servers connected with extremely fast networking.

That’s a way of doing it, but it doesn’t play to the strengths of Cloudflare, where we have relatively commodity servers scattered all around the world. We can do things like make sure that their time is synced precisely, so that you can avoid double-spend problems and other things. We need to play to the strengths of what we’re able to do.

We have a team thinking about this and working on it, but we have no pride of ownership. If there’s someone else who has a solution out there, we’d love to talk to them.

10. Do Agents Need Blockchains?

There. You know, this is a crypto podcast, after all. I think a lot of people here are trying to figure out: are agents the next big—

—thing in crypto.

I think—I forget who said this. Maybe it was Eric Voorhees, who started another company called Venice that’s doing quite well right now. I think he was the one who said, “Maybe humans weren’t actually the end customers for blockchains after all. Maybe we’ve just been waiting for these agents.”

So I’m curious: you are the person building this right now.

You are having a build-versus-buy conversation internally. Why do you need to hit a chain at all? Why don't you just use internal ledgers and sign receipts?

Yeah. I mean, we may, but the thing that's nice about chains is that they can then be across multiple different organizations. The thing that's bad is that consensus is slow, so you've got to balance those two bits. I would guess that it ends up being something that's a little bit hybrid.

One of the things that I've always appreciated about Cloudflare is that we're not very religious. We tend to try and figure out what the right solution to the problem is, and we don't just think that we have a hammer, so everything's a nail. We have a hammer, but we have a screwdriver, we've got a drill, and a whole bunch of other things. We're going to try and combine that to create the right solution.

Again, I think if you're participating in this, there will be a way to get funds in and out of this across a bunch of the different blockchain protocols. But we'll also partner with PayPal, Visa, American Express, and traditional banks. If you want to just move money out, we'll let you do that, too. We want to be the universal translator that makes everything work.

Our interest is in making sure that our customers—the people who are putting content and applications online—are able to have successful businesses and be able to thrive. And so I think that—

You know, sometimes I think some parts of the crypto ecosystem can get pretty religious and be like, “You have to do it exactly this way.”

Yeah, we're kind of like, listen, this is an amazing tool, and we're super happy to have it. We've got these other tools, too. The right answer is probably an amalgamation of more than one of them.

Do you worry at all that you have a very, what I'd call, neutral business today—security and DNS and .com, browsers—and it's all very neutral plumbing? Money tends to be regulated and custodial and sometimes adversarial in new ways that maybe your business has never faced before.

Never faced before? I mean, content—

You've never faced regulation. Yeah, regulation.

I never thought I was going to have to be an expert in Spanish copyright law.

I was reading about—who was it? Italy or something. Something in Italy.

Yeah. I'm in Spain and Italy, and I'm like an expert on Indonesian politics at this point. I mean, I think we've always kind of been in some of that. I agree that there are all kinds of additional problems that come when you touch money.

On the other hand, I think we don't need to be the ones that actually hold the money. We just need to be the ones that help transact between the websites and the agents and make that all work. When we talk to regulators, they're really concerned that if we don't get the business model right, this incredible resource that we've had implodes.

Because we don't come at it from “we want to be a bank,” but rather, “we want to make sure that content creators and business owners are getting paid for the work that they do,” we've so far gotten really constructive interaction with regulators around the world. But I'm not naive. We will have to deal with all sorts of new problems that come along with this. Even if the transactions are very small, if you do a lot of them, you can create fraud or all kinds of other problems. We'll have to deal with that.

Yeah. I remember your—

I don't remember the name of the site. I remember you kicked someone off the site like a decade ago.

The Daily Stormer was one. 8chan was another. Kiwi Farms was another. There were others. It's funny: if we get a court order to block something in some region, we comply with that. We have to.

But I thought you just didn't comply with that in Italy.

We didn't comply. We did not sign up in Italy. Italy has this thing called Piracy Shield.

It's like the piracy shield or something.

Piracy Shield, yeah. Basically, you have to sign up, and then anyone who runs a media company can put any website on a list. The requirement of Piracy Shield is that it has to be taken down globally.

That’s a problem for a lot of different reasons. Literally, they tried to take offline a site that was created by the US Department of State. So we had a little bit of Italy being able to regulate outside of Italy. If Italy wants to be crazy outside of Italy, that's fine, but regulating beyond their borders seems unreasonable.

The second thing is that there's no judicial oversight. There's no due process, and the rule of law matters. That means there has to be some actual oversight by some enshrined body. It can't just be a group of media owners that gets to decide what's on the internet and what's not on the internet.

If Italy sent us a court order saying, “Block this thing in Italy,” we'd comply with that. If Italy sets up a scheme where an unelected cabal with no judicial oversight can block something everywhere in the world, obviously we're going to push back on that, because that's insane.

Yeah. Yeah. Let's end on a happier note instead of ending with Italy. Matthew, is there anything that you see right now—across AI, running companies, blockchains, really anything—that you have a vantage point into and think hasn't become obvious to the general public yet?

Actually, one fun fact about Italy is that if you're convicted of a white-collar crime, which is what they're threatening me with, you can serve your punishment time. You don't have to actually go to jail, but you do have to stay in Italy. I was just thinking, well, I can just go hang out in Tuscany for a while. Well, that might not be safe.

Yeah. Tough, tough.

11. What Breaks In The Agent Internet?

Oh, you know, how dare you make me go eat all that?

How dare. Matthew, across any topic here—AI, running companies, blockchains, really anything—is there anything that you think you see right now, through your vantage point, that hasn't become obvious to the general public yet?

I think what's wild to me is that during COVID, in 2 weeks, internet traffic doubled, and parts of the internet almost collapsed. Nobody talks about this, but at the time I was on a crazy phone call with Netflix, Amazon, YouTube, Akamai, and a bunch of the big streaming services.

It was like this Mexican standoff. The late Senator Ted Stevens said the internet's just a series of tubes and got mocked mercilessly. The internet's just a series of tubes. Those tubes have a finite capacity. If you send too much traffic through them, they start failing.

We were watching this both because there were a lot more people putting their kids in front of the TV to entertain them during COVID, but also because the peak times of internet use in Asia started to line up with the peak times of internet use in Europe. People were basically getting up and streaming earlier in Europe, and then when people were coming home from work in China, they were also streaming.

A lot of that just concentrated across the major exchange points in Europe, and it was causing things to fall over. The Mexican standoff between the various streaming companies was that everyone had to decrease their streaming rates. They couldn't stream at 4K anymore, otherwise the internet would go down. No one wanted to do it, but eventually everyone agreed to do it, and that's what kept the internet going.

12. Did Ads Break Society?

That was at 2x. We think it's going to be 1000x. The good news is that agents don't tend to stream video as much, at least today, but who knows what the future looks like? There's a lot of stuff that's just going to break unless we find ways to make it significantly more efficient or find ways to pay for it and support that. That's the thing I'm spending the most time thinking about, and I don't think the rest of the world has woken up to that yet.

The other thing that I'm spending a bunch of time thinking about—and I kind of feel like I have an idea of a solution for the first bit, but not the second—is that I really worry that in the future it's going to be incredibly hard to start a new business. What is it that gets you to buy from a new small business that's starting? Typically, the answer is either convenience or emotion.

Either it opened and it's right on the way to work, so you can stop there quickly, or it's emotion: They put a Facebook or Instagram ad in front of you, and you're like, “Oh, yeah, I have to buy that skincare product,” or whatever it is.

What's going to be interesting in the future is that I think agents don't care about those things. Agents are coolly calculating and have an almost infinite amount of time, so convenience and emotion don't work for a new entrant. The challenge is that if we don't have new entrants, what happens over time is that you get consolidation among the major players.

Agents are going to trust the things that they know and that they can find the most information on, which is going to put even more pressure around consolidation. I worry that for small businesses, there can be incredible pressure. It's going to be very difficult for new entrants to enter the market.

I said earlier that I want a world where there aren't just 5 AI companies; there are 500,000. I also want a world where there aren't just a small number of very powerful conglomerates doing everything for us, but instead we have a way for new entrants to come in. I don't know that we know exactly how that's going to work, but it's something that I'm spending a lot of time thinking about.

The other side of that problem is that I'm an extreme outlier in terms of my financial resources and other things. I would guess that in the not-so-distant future, I'm spending about $1 million a year on the various agents that work on my behalf. I bet the average U.S. consumer will spend about $1,200 a year.

My theory would be that the money comes from replacing things like your Netflix subscription and a lot of the stuff that you use for entertainment. That shifts over to spending on the various AI agents. I think that a lot of things around advertising-based business models are going to change.

If you saw the Anthropic ads where they talked about how creepy ads with agents, or ads with an AI advisor, are, I think that the fight for those dollars is going to be so fierce that it's going to be really difficult to have ad-based models. Ads inherently mean that you're corrupting the system from giving you the right answer.

If I suddenly realize that my agent is recommending I buy a Ford rather than a Nissan because Ford paid them for advertising, I'm going to switch my million dollars to someone else. I think most consumers are going to switch really quickly.

To give you a sense, Google makes about $750 per U.S. user per year. Meta makes about $250. So $1,200 is a big number to fight for, and I think people are going to be fighting for that like crazy.

What I worry about is the kid in Sri Lanka, because right now the kid in Sri Lanka still has access to most of the web because it's ad-based. But if, in the future, in order to get in and compete, you've got to be spending $1,200 a year—or maybe even $1 million—it's going to be really tricky.

13. Will AI Become The Internet?

I really worry that this is going to exacerbate the problems of the haves and have-nots. It's going to lock in the advantages that a lot of the developed world has. I haven't heard any solution to that problem, and I think it's something that we're going to have to think about and worry about as a society.

Won't ChatGPT and Anthropic just become the internet? I guess that's the bear case.

Yeah. I was talking to—I don't know how old this kid must have been, maybe 14 years old. His internet is TikTok right now. He doesn't Google; he just TikToks everything.

I'm like, “10 years from now, the next 14-year-old will be on ChatGPT.” Not only that, my dad, who's 85, just lives in his chat systems all the time.

What people don't totally appreciate is that there have been platform shifts. You started out browsing the web on your browser, on your desktop or laptop. That then shifted to social, where that became the interface—that's the kid on TikTok. Then mobile was another shift.

For all of those, the dominant business model remained the same, which was advertising.

Yeah.

We're now shifting to the interface that you're going to look at being AI chat. Exactly what it looks like is still up in the air, but I go back and watch old episodes of the cartoon The Jetsons and I'm just like, “How did George—if George wanted to buy something, what did he do?”

The answer was that he asked Rosie, the helpful robot, and Rosie went out and did it for him. I think that's going to be a lot more of what happens. Most of the information you consume, you're going to get through some sort of AI interface. Most of the commerce that you do, you're going to do through some sort of AI interface.

That becomes the view that you have for the rest of everything. Again, that's going to put enormous pressure on what have historically been advertising-based business models. We're going to have to find some way to continue to reward the creators who are creating that.

I can see a path, by the way, to doing that financially. I flew up to Stockholm to meet with Daniel Ek, the creator of Spotify. One of the things he said—it was a fascinating conversation—is that there are 2 reasons why people become musicians: to get rich or to get famous.

He's like, “If you had to pick which one motivates people more, it's actually the getting-famous part.” By the way, that's totally irrational. Being anonymous and rich is the best thing in the world, so aim for that. But fame drives a lot of this.

What I think is really going to be tricky is that I can find a way to get content creators to continue to be rich in an AI world. But if what you're consuming is sort of a regurgitation, sort of the CliffsNotes version that you're getting through AI, how do I ever make sure that that person gets famous?

We should create some sort of—I don't know if it's the Nobel Prize or the Academy Awards—but, for the first time in human history, we have a mathematical model of all human knowledge. That's what the LLMs are.

We can measure who, in the last year, contributed the most to Texas history research or to the biology of mammals, or whatever it is. We can actually say, “Okay, it was that person,” and we should actually be giving people those rewards.

Fame is still something that really motivates people, and being recognized by your peers and being recognized for really contributing to human knowledge is super valuable. I think we're going to have to do that in order to get people to continue to do the work that feeds the information into these systems that we're all going to be using.

Yeah, Matthew, amazing conversation, man. I'm not only a customer, but a very happy customer of Cloudflare. I really like the way that you see the world and enjoy hearing your thoughts on everything. I appreciate the time.

Thank you. Thank you for having me.