Cloudflare executives expect artificial intelligence and other automated systems to transform the composition of internet traffic so dramatically that human-generated activity could become almost negligible within five years.
Chief Financial Officer Thomas Seifert said during Cloudflare’s second-quarter earnings call on Thursday that machine-generated traffic had already surpassed human-generated traffic in May, earlier than the company’s previous forecast that the crossover would occur in 2027.
Seifert acknowledged that his earlier projections had been wrong, but offered an even more dramatic forecast based on the latest traffic data.
Register for Tekedia Mini-MBA edition 20 (June 8 – Sept 5, 2026).
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
“To give you a sense of how this trend is playing out, and with the big caveat that I have called it wrong at every point along the way, if the current trends continue, we think in five years, non-human traffic will be as much as 1,000 times as much as human traffic,” he said.
“In other words, humans will be a rounding error on the internet, not because human traffic goes down, but that’s just how fast we’re seeing non-human traffic grow.”
The prediction highlights the scale of the change underway as AI agents, automated software, bots and machine-to-machine applications generate increasing volumes of internet activity.
The shift is being driven in large part by AI. Unlike traditional web users, AI systems can make large numbers of automated requests, retrieve information, interact with websites and APIs, generate content and perform tasks continuously without direct human involvement for every action.
That creates a fundamentally different traffic pattern from the internet built around human browsing.
For Cloudflare, which operates a global network that sits between users and online services, the change presents both an opportunity and a challenge. More machine-generated traffic means greater demand for network infrastructure, application performance services and security, but it also creates new problems around capacity, abuse and cyberattacks.
Seifert said the company would need to become significantly more efficient if machine traffic continues expanding at its current rate.
The security implications could be particularly important. AI agents can generate traffic at much greater scale and speed than individual users, potentially increasing automated attacks, credential abuse, scraping and other malicious activity. As more agents gain the ability to interact directly with websites and online services, distinguishing legitimate automated activity from malicious traffic could become increasingly difficult.
Cloudflare’s own financial strategy is built around handling that growth without following the massive capital-spending model adopted by the largest cloud providers.
The company expects capital expenditure of roughly 14% to 15% of its forecast annual revenue of between $2.865 billion and $2.87 billion, implying spending of about $430 million. That is relatively modest compared with the enormous amounts being invested by hyperscalers to build AI infrastructure.
Cloudflare Chief Executive Matthew Prince used the difference to distinguish the company’s business model from traditional cloud computing.
“If you’re selling what is just commodity compute, if you’re basically letting an AI company use your balance sheet and your credit rating in order to buy servers that are the same as everybody else’s servers, then that’s just not attractive business for us,” Prince said.
Cloudflare’s strategy is to extract greater utilization from each server rather than simply expanding its physical infrastructure.
Prince said the company wants to “squeeze as much out of every Capex dollar as possible” by improving the way it uses memory, storage and computing resources within its existing equipment. He also argued that traditional hyperscalers have relatively low GPU utilization because their model largely involves providing customers with computing infrastructure and leaving customers responsible for maximizing its use.
“The hyperscalers, the traditional first-generation clouds, are in the business of buying a server and then trying to sell it back, lease it back, and get five turns of revenue off of it,” Prince said.
“We’re in a very different business where we’re selling actually work getting done.”
That distinction is central to Cloudflare’s strategy. Rather than competing directly with Amazon Web Services, Microsoft Azure and Google Cloud on the volume of computing hardware available for rent, Cloudflare is attempting to sell the outcome of computing activity through its serverless and edge-computing services.
Prince said Cloudflare does not want to rent servers because he views that market as becoming increasingly commoditized. Instead, the company intends to focus on extracting more computing output from its infrastructure through scheduling, software optimization and higher utilization.
The strategy appears to be resonating with investors.
Cloudflare’s shares rose about 16% in after-hours trading following its quarterly results, reaching a record high and extending its year-to-date gain to roughly 68%.
The market reaction also suggests investors were more focused on the company’s revenue growth and customer momentum than on its increased losses. The company reported a sharp increase in losses, with net losses more than tripling to $205.7 million, while revenue growth exceeded expectations and Cloudflare reported a record number of large enterprise customers.
The contrasting financial profiles of Cloudflare and the hyperscalers point to an increasingly important divide in the AI infrastructure market.
The largest cloud companies are committing hundreds of billions of dollars to data centers, GPUs, networking equipment and power capacity to meet the expected growth in AI workloads. Cloudflare, by contrast, is betting that software optimization and distributed infrastructure can allow it to benefit from the same AI-driven traffic growth without matching that level of capital intensity.
The scale of machine-generated internet traffic could ultimately test that strategy.
If AI agents continue multiplying and increasingly operate autonomously across the web, the internet could evolve from a network dominated by human requests into one dominated by machine-to-machine interactions. That would increase demand for infrastructure capable of processing enormous volumes of automated traffic while also requiring more sophisticated systems to determine which automated activity should be allowed.
Seifert’s forecast of machine traffic reaching 1,000 times human traffic within five years is therefore less a prediction about humans abandoning the internet than a projection of how quickly the number of automated interactions could grow.
Humans would continue to generate traffic, but the volume generated by AI agents, software services and automated systems could expand at a far faster rate.