Why Cloudflare Started Funding Work Outside Its Main Roadmap
In May 2015, Facebook announced Instant Articles, a format that let publishers package their pages and serve them directly from Facebook’s own infrastructure. Google responded that October with Accelerated Mobile Pages (AMP), which applied the same idea but ran on Google’s infrastructure. For anyone who cared about the open web, the trajectory was alarming: if both efforts succeeded, the majority of web content would end up hosted and controlled by two large platforms.
For Cloudflare, the threat was existential as well as philosophical. The company’s stated mission is to help build a better Internet — but that mission presumes the Internet continues to exist as a distributed system of independent networks. If publishers moved their infrastructure onto Facebook and Google, there would be little left for a company like Cloudflare to do.
The motivations behind both formats, however, were legitimate. The web was designed for devices on fixed, wired connections. As usage shifted to mobile devices over wireless networks, performance suffered — particularly in developing regions where connectivity is unreliable and users are far from content origins. Facebook and Google reasoned that by serving complete, cached copies of pages from their own edge infrastructure, they could materially improve usability in places where Internet adoption was still growing.
The rationale was sound; the endpoint was not. But the problem was still real, and it demanded a response.
Making Room for Long-Term Bets
Internally, the risk was clear to Cloudflare’s leadership, but any company juggling customer priorities will defer strategic work in favor of immediate needs. To avoid losing that battle entirely, Cloudflare in 2016 funded a small skunkworks team in Austin, Texas, led by Dane Knecht, the company’s founding product manager. The idea was to give strategic projects with long payoff timelines physical and organizational space away from headquarters.
That team evolved into what is now the Emerging Technologies & Incubation (ETI) group. It is the origin of products including Cloudflare for Teams, 1.1.1.1, and Workers. Austin has since grown from an outpost into what will be Cloudflare’s largest office by the end of the year, housing members from teams across the company — but its genesis was this effort to find a productive answer to Instant Articles and AMP.
What Collaboration With Google Produced
Cloudflare met with both Facebook and Google. Facebook’s worldview centered on its own app and left little room for partners. Google, by contrast, was founded out of the open web and, despite ongoing criticism of AMP, maintained an interest in preserving a distributed ecosystem. Cloudflare engaged directly with Google’s AMP team on those concerns.
Cloudflare built a number of products intended to extend the AMP ecosystem and make it more open. Taken on their own, those products were not commercial successes — but they had an important side effect. Building a third-party AMP cache required a more programmable network, which pushed the company to prototype several serverless computing approaches before settling on Workers. In fact, many of the AMP-related products Cloudflare shipped were the first products built on Workers.
That outcome is consistent with how the ETI team operates: it takes frequent, low-cost attempts at new ideas, expects some to fail, and treats even failures as opportunities for learning and internal collaboration. That structure, rather than any single product bet, is what the company credits for its rate of innovation.
From AMP to Signed Exchanges
Working with Google’s AMP team also opened the door to more collaborative work on Internet performance. Cloudflare’s mission is deliberately phrased as “help build a better Internet” — not build one outright — because the Internet is fundamentally a collection of networks and the work of many organizations. The company argues that meaningful performance leaps come from cooperation on open standards with other companies aiming to democratize access to optimization technology.
Signed Exchanges, the technology that emerged from that collaboration, preserve the key benefit of AMP — the ability to preload pages so they render nearly instantly — while restoring control to individual publishers. Pages are not dependent on a single provider’s infrastructure, and the format extends beyond search-result traffic. For users in regions with slow or expensive connectivity, that translates into a faster, more accessible web.
The work has implications for how content distribution and trust can be separated on the web — an area where developments continue to unfold. For now, Cloudflare’s ETI team has moved on to the next set of horizon-level questions, but the pattern is established: treat disruption as a prompt for internal experimentation, and measure success not just in shipped products, but in the technologies those experiments enable.
