Cloudflare has crossed a threshold that says as much about the future of software as it does about the company itself: more than 10 million developers now build on its platform, up from 7.4 million just one quarter earlier. The jump reflects not only Cloudflare's expansion beyond its original serverless compute product, Workers, but a deeper transformation in how applications are built, who - or what - is building them, and who they are built for.
From a Single Product to a Full Development Stack
Workers launched nearly a decade ago as a lean, globally distributed alternative to conventional cloud compute. Instead of running code in virtual machines or containers, it relied on isolates, a lighter-weight execution model that starts almost instantly and scales across Cloudflare's network without the overhead traditional infrastructure carries. That architectural choice, made early, has become the foundation for a much broader platform. Cloudflare now offers storage, databases, AI inference, media handling, real-time communication tools, observability, and developer workflow tools, all designed to work together rather than as separate services developers must stitch into place themselves.
Rita Kozlov, Cloudflare's VP of product management, frames the expansion as a response to a structural shift in software itself. Developers, she notes, are no longer building only for human users; they are increasingly building for a world where agents act alongside people, and sometimes in place of them.
A Traffic Shift With Real Consequences
The numbers behind that claim are striking. Daily requests from AI agents on Cloudflare's network grew by more than 1,700 percent over the past year, and for the first time, more than half of all Internet traffic the company observed was not generated by humans at all. That is a meaningful inflection point. Applications once built to serve browsers and mobile apps now have to anticipate constant interaction from automated systems: bots that summarize content, agents that complete transactions, and software that queries other software continuously, around the clock.
This shift changes the demands placed on infrastructure. Platforms can no longer be optimized solely for occasional human requests with predictable patterns; they must handle sustained, high-volume, machine-driven traffic without sacrificing speed or reliability. Developers responding to this reality are turning to tools like Workers AI, Workers KV, AI Gateway, Durable Objects, D1, and R2 to build applications that are AI-aware from the start, rather than retrofitted later.
Why Network Scale Has Become a Competitive Factor
Cloudflare's pitch rests heavily on physical and architectural reach. The company's network now spans more than 340 cities in over 130 countries, interconnects with nearly 14,000 other networks, and carries close to 600 terabits per second of external capacity. For developers, this scale is not an abstract bragging point - it determines how quickly an application responds to a user in a distant region, how resilient a service is against surges in demand, and how well it can support agents and APIs interacting with each other continuously rather than in isolated bursts.
Dane Knecht, Cloudflare's chief technology officer, argues that this scale removes a trade-off developers have long had to accept: choosing between speed, reach, and security. Building once on infrastructure already distributed globally, he suggests, is becoming the more practical path as applications increasingly need to serve not just people, but other software.
What This Signals for the Next Phase of the Internet
The broader implication is that the Internet's underlying demands are changing faster than many organizations have adjusted to. As AI tools write more code and automate more of the software lifecycle, and as agents take on persistent, autonomous roles, the infrastructure beneath applications needs to support a mix of human and non-human activity at a scale and consistency earlier systems were never designed for. Cloudflare's growth numbers are one data point in a much larger transition - one in which the distinction between who builds software and what uses it is becoming far less clear-cut.
