Who Owns the Front Door to AI?

Speaker 1

We're all using these incredible new AI tools, right? But have you ever actually stopped to ask who's picking up the tab for all this? Well, today we are diving deep into the real business model of AI. We're going to map out how we can build a sustainable AI native internet, looking at everything from court cases to eye-watering funding rounds. Because when it comes to AI, it's not about control. It's about making the math work. Yeah, these tools feel free, or at least super cheap to use. But behind that simple, clean interface, there's a massive, incredibly expensive engine running. So the big question for this explainer is, if we aren't paying the full price, who is? And more importantly, is that system built to last? Okay, let's get into it. So the first major problem we need to talk about as well, it's a total paradox. You know how technology usually gets cheaper and better over time? Think about your phone or your TV. With the most advanced AI, the exact opposite is happening. Building on top of the best models is actually getting more expensive, not less. We're calling this the AI cost squeeze. So what does it really mean for an AI to think more? Well, when you ask it a complex question, it's not just one single action. As this quote gets at, one simple request from you can set off a whole chain reaction of computations and calls to other tools. And every single one of those steps adds to the cost. The bottom line is, the smarter the AI gets, the more expensive it is for the companies building with it. And you can imagine this kind of unpredictability is an absolute nightmare for businesses, especially for startups. I mean, think about running a company where your main cost is a complete mystery from one minute to the next. One user might cost you a tiny fraction of a cent, but another could rack up hundreds of dollars in a single afternoon. It makes building a stable business on AI almost impossible. And just how wild can these costs get? Okay, here's a real-world example. Anthropic, one of the top AI labs, had an early pricing plan for 200 bucks a month. Sounds reasonable. But they admitted that a single power user on that plan could actually cost them tens of thousands of dollars a month to support. That is not a small rounding error. That is a business-destroying kind of problem. So you have to be wondering, how on earth does a company survive those losses? Well, the answer is with this. $13 billion. That was the size of a recent funding round for Anthropic. And it's a direct result of that cost squeeze. Only the most gigantic, well-funded companies can afford to absorb that kind of loss and still keep prices low for you and me. And this isn't just normal funding. It's a massive concentration of capital into just a few hands. I mean, look at this growth. Anthropics valuation pretty much tripled in only six months. This is a huge shift where just a handful of giant companies are soaking up almost all the investment needed to build this foundational technology. So what's the end game in all this? It's something called vertical integration. When your number one cost is computing power, the only real way to control it is to own it yourself. And that's exactly why we're seeing companies like OpenAI get into the business of designing their own custom AI chips. They're trying to own the entire process from the silicon on up just to wrestle those costs into submission. Okay, so the economics are completely stressed out from rising costs. But believe it or not, that's only half the story. At the exact same time, the fundamental bargain that powered the internet for the last 25 years is completely falling apart. The old deal is dead. For decades, the internet worked on a simple, unspoken deal. Publishers let search engines like Google scan their content for free. And in return, the search engines sent them valuable visitors or traffic. But these new AI answer engines just shatter that deal. They scrape the information, give you the summary right there, and take all the value from the publisher without sending anything back. The clicks are just gone. And hey, this isn't just some random theory. This is Matthew Prince, the CEO of Cloudflare, a company that literally sits at the heart of the Internet's plumbing. He says it plainly. The move from search to answers has completely cut off that old quid pro quo. The value exchange that built the web as we know it is basically over. So if the old deal is dead, what does a new one even look like? Well, interestingly, there's a technical standard for this that's been sitting around for years. It's called HTTP 402 Payment Required. It could be used to build a pay-per-crawl system, where a website could just automatically tell an AI bot, hey, you want to use my content? No problem. Here's the price. It would turn a legal right into a simple line of code. But while the tech folks are debating new standards, the courts are already busy putting a price tag on data. And that price tag is huge. This number, $1.5 billion, is what Anthropic reportedly paid to settle a massive copyright lawsuit from book authors. Now, this wasn't just another legal payout. This is the largest settlement of its kind in U.S. copyright history, and it sets a huge precedent. It's basically sending a message loud and clear. The era of scraping high-quality data for free and just calling it fair use is finished. That data now has a formal, legally recognized, and very, very high price. Okay, so let's take a breath. We've got skyrocketing costs on one side and a broken, legally messy system of value on the other. It sounds like a complete mess, but out of all this chaos, we're starting to see the outlines of a new, more sustainable model, a new deal for the web. And this new solution really stands on three main pillars. First up, identity. AI agents and bots need clear identities so publishers know who's accessing their content. Second is attribution. We need to be able to track the chain of value to see where the good stuff comes from. And third, and maybe most important, settlement. A real system to make sure that when value is created, everybody who is part of it gets paid fairly. Now, the goal here is not a zero sum game. It's about creating a virtuous loop where everyone wins. Users, that's us, get to keep using amazing, nearly free AI tools. Publishers and creators finally get paid fairly for all their hard work. And this is key. The AI companies also get paid, but only when they can prove they created real, measurable value. Suddenly, everyone's incentives are pointed in the same direction. The glue that holds us all together is trust and solid proof. This new system has to be run with receipts. Every single part of the process has to be measurable, with trustworthy data showing who did what and when. You know, it's probably not a coincidence that OpenAI just went and bought an analytics company called Statsig. The big players are already building the plumbing for this new, measurable web as we speak. So we have this vision for a much more stable and fair AI internet, but how do we actually get there from here? What are the key things you should be watching for to see if this new model is actually starting to happen? Here are three big things to keep an eye on. First, those custom chips. We need to see if that vertical integration actually brings down costs for developers or if the big platforms just end up pocketing the savings. Second, that paper crawl idea. Will web standards like HTTP 402 actually get adopted and turn writes into code? And third, watch the regulators. Will antitrust moves force data sharing in a way that truly lets smaller companies compete? This really brings us back to the ultimate goal, doesn't it? A system where users, creators, and AI companies can all win together. Now, it is a complex and really challenging road ahead, no doubt about it. But for the first time, a real map for a truly sustainable AI native internet is starting to come into focus. The only real question left is, can we actually build it?