The internet's next great battle may not be over content or bandwidth, but over who is allowed to read the web — and who has to pay for the privilege. That is the argument Cloudflare chief executive Matthew Prince has been making with growing urgency, most recently in a wide-ranging conversation on The Verge's Decoder podcast, part of a two-part series on the future of business.

Two and a half years ago, Prince described the moment as a wild pivot point for the internet. In hindsight, he told host Nilay Patel, things are even wilder — because AI has moved from curiosity to core infrastructure, and Cloudflare now sits squarely between the world's websites and the machines that want to consume them.

The clearest evidence lives in Cloudflare's own telemetry. In June, the company reported that bots accounted for more than half of all traffic crossing the internet — the first time automated traffic has outnumbered humans. That share keeps climbing as AI companies scrape ever more of the web and autonomous agents begin acting on users' behalf, booking, buying, researching and browsing without a person ever touching a keyboard.

The gatekeeper's bargain

Cloudflare's position is unusual, and unusually powerful. Because it proxies traffic for a substantial slice of the web, it can decide — on behalf of site owners — whether an AI crawler receives a page, is blocked outright, or is charged for access. Prince has framed the choice as a three-way menu: block all AI tools, allow them, or permit only those willing to pay.

That third option is the one quietly rewriting the web's business model. Publishers and platforms that once gave away content in exchange for search referrals now watch AI systems summarize that same content and keep the traffic. Cloudflare's pitch is essentially a tollbooth: an enforcement layer that turns crawling from a free-for-all into a negotiated transaction.

The company has already moved from rhetoric to deadlines. NBC News reported that Cloudflare set a cut-off for blocking AI crawlers that bundle search with AI training — a direct challenge to the practice of using crawler traffic for search indexing while simultaneously harvesting the same pages to train models.

Not every relationship is adversarial, however. Seeking Alpha reported that Cloudflare and OpenAI teamed up on a research pilot for AI search indexing — a signal that the fight is less about keeping AI off the web than about setting terms for letting it in.

Giving agents an identity — and a wallet

As FinTech and security outlets have emphasized, the story is as much about payments as it is about crawling. Finextra reported that Cloudflare is giving AI agents both an identity and a wallet — cryptographic proof of who an agent is and what it represents, paired with a mechanism to pay for content and services it consumes.

That combination matters. An agent that can authenticate itself and spend money is no longer a scraper to be blocked; it is a customer to be served. It also opens a gray zone: who is liable when an autonomous agent misbehaves, overspends, or signs a contract on someone's behalf? Banking and security publications, including BankInfoSecurity, have framed the shift as a commercial reset in which the web economy is being rebuilt around machine-to-machine transactions.

Dead internet theory, upgraded

The cultural resonance has been hard to miss. The Next Web and Yahoo Finance both reported that AI agent traffic is growing by nearly 8,000%, describing the once-satirical "dead internet theory" — the idea that the web is mostly bots talking to bots — as a measurable fact rather than a conspiracy.

That reframing cuts both ways. If most traffic is synthetic, then advertising metrics, audience measurement, search rankings and social engagement are all built on increasingly unstable ground. It also changes what is valuable: original, verifiable, human-created information becomes scarce precisely as synthetic content becomes free.

"Bots now account for more than half of the traffic Cloudflare sees — and the share keeps going up as AI companies scrape more of the web and agents start doing things for people."

The paradox inside Cloudflare

Cloudflare's own workforce tells a parallel story. The Next Web reported that the company cut 1,100 jobs and then grew its engineering team by 45%, with Prince arguing the pattern will repeat across the economy: automation eliminates one class of role while creating outsized demand for the people who build and secure the systems doing the automating.

"The pattern will repeat everywhere," Prince has said of the shift, describing a labor market that shrinks in some places and expands sharply in others.

For a company positioning itself as the web's referee, that is a convenient thesis — and a self-interested one. Cloudflare profits whether traffic is human or machine; the only question is whether it gets paid for the machine half.

How the story gets framed

• The Verge frames it as a personality-driven question: can one CEO save the open web from AI, or will he simply become its landlord?
• FinTech and security outlets (Finextra, BankInfoSecurity) treat it as a commerce and infrastructure story about agent identity, authentication and payment rails.
• NBC News focuses on enforcement — deadlines, blocking, and the policy mechanics of crawler access.
• The Next Web and Yahoo Finance zoom out to macro effects: dead internet theory, a rewired advertising economy, and a labor market in flux.

What happens next

The open questions are substantial. Will site owners opt in to paid crawling, or simply block? Will regulators treat agent authentication as a technical standard or a competition issue? And will users accept a web where access is metered by whoever controls the proxy layer?

Prince's answer — that the web needs a market for machine access before it is consumed for free — is pragmatic and, to critics, self-serving. Either way, the era in which humans were the internet's default users is over. The next phase will be defined by how quickly the web builds a tollbooth that AI can actually pay at.