Cloudflare CEO: AI Agents Are a 'Bane' for Website Operators

Cloudflare CEO Matthew Prince told investors on the company's third-quarter earnings call that some of Cloudflare's customers view AI agents as "a bane to their business and want to keep them away," according to The Register. The comments came alongside a record quarter for the network infrastructure company, with revenue reaching $782.5 million, up 39 percent year on year, and shares jumping around 15 percent in after-hours trading.

What Happened

According to The Register's account of the earnings call, Prince used a hypothetical example to explain the agent problem: a person asking an AI agent to find a restaurant could trigger the agent to scrape a thousand different restaurant websites and menus, only for the user to pick one. "The problem is that 999 of those restaurants have now had a cost that is just been applied to them, but they are not getting any benefit," Prince said. He argued that while web search has always generated some low-value traffic, the sheer volume of agentic activity now changes the scale of the problem, with Cloudflare previously predicting bot traffic could outnumber human clicks by a ratio of 1000 to 1.

Prince also used the call to address security, telling investors that nation-state actors and criminals likely exploited the recently disclosed PHP2SHELL vulnerability for years before it became public. "These vulnerabilities, like the PHP vulnerabilities that were disclosed the other day, had been in the wild for years," he said, adding that Cloudflare is combing through logs looking for historical evidence of exploitation.

On infrastructure, Prince predicted that containers, the technology underpinning most hyperscale cloud and mobile application deployments, are heading toward obsolescence. He credited containers with being "a significant improvement over bare metal or even virtual machines" but argued that "the container-based architecture of the hyperscalers and neoclouds just can't keep up in a cost-effective way with the new agentic paradigms of today." He pitched Cloudflare's serverless Workers platform as the alternative to containers; The Register noted this was a Cloudflare-centric view given the company's commercial interest in the product.

Prince said demand for Cloudflare's agent-related products, which can block bots, reduce the cost impact of agentic traffic, or help sites monetize it, is growing quickly. He claimed that since July 1st, "approximately five percent" of the web has opted into Cloudflare's content protection products. He named Meta as a customer and said Meta's new "Muse" agents will generate more traffic for third parties to manage, indirectly benefiting Cloudflare. He also said the current wave of disclosed vulnerabilities will not let up soon, telling investors "I wish I could tell you we were through the worst, but we're not," while framing the ongoing security disclosures as a profit opportunity given Cloudflare's early access to quiet vulnerability disclosures.

Why It Matters

Prince's comments describe a cost-shifting problem rather than a technology failure: the compute and bandwidth burden of agentic browsing lands on site operators who built infrastructure for human-scale traffic, not machine-scale scraping. That is a direct training-data and compute infrastructure story, because every agent query that pulls down a menu or scrapes a page is effectively treating the open web as a live data source, repeatedly, at a volume Cloudflare says could reach 1000 bot requests for every human click. The MachineHours AI Acceleration Index puts the compute subindex at 56 against an overall score of 46, a gap consistent with Prince's argument that infrastructure demand is outrunning broader capability and adoption trends. Whether containers actually become obsolete is a separate question from whether Cloudflare profits either way, since the company sells both the blocking tools and the serverless alternative. The pattern fits a broader dynamic tracked in ongoing analysis of AI infrastructure investment, where vendors at every layer of the stack argue their product is the necessary response to agentic scale, regardless of whether enterprise agent adoption has caught up to the infrastructure spend.

Further Reading and Context

For background on how autonomous AI agents are reshaping enterprise workflows and the infrastructure built to support them, see MachineHours' guides on AI agents in the enterprise and the broader analysis of the agent economy. The shift from conversational tools to autonomous systems capable of independent action is also covered in MachineHours' piece on Claude 4's agentic capabilities.

What to Watch

Watch whether Cloudflare's claimed five percent opt-in rate for content protection products grows, since that figure is the clearest available measure of how many site operators are actively trying to manage or block agent traffic. Also worth tracking is whether other infrastructure providers echo Prince's prediction about container architecture losing ground to serverless models built for agentic workloads, since Cloudflare's framing is self-interested but not necessarily wrong. Finally, Prince's comments about historical exploitation of the PHP2SHELL vulnerability suggest more disclosures tied to long-dormant bugs may surface as Cloudflare and others review logs, a trend that could keep security spending elevated through the next several quarters.

Further reading and context

• The Rise of AI Agents — Which Enterprise Software Companies Win and Lose