AI agent traffic drives first profitable year for Fastly

AI agents aren’t just hype, they’re actually having a material impact on content delivery network economics—at least for Fastly. Its fourth-quarter earnings call this week provided evidence of how AI agent behavior differs from traditional web traffic and how much revenue that difference generates.

The edge cloud provider posted $172.6 million in revenue for Q4 2025, up 23% year over year. That marked Fastly’s strongest quarterly growth rate in more than three years and delivered the company’s first profitable fiscal year with $19.7 million in net income for 2025.

During Fastly’s earnings call, CEO Kip Compton attributed the results in part to changes in how AI agents consume web content.

“If you’ve used AI tools, I think you would appreciate that they often check a lot more websites, for instance, than you might,” Compton said. “And that’s more traffic, and all of that traffic is processed through the Fastly network for our Fastly customers.”

AI agent traffic creates distinct usage patterns

Fastly has been tracking AI agent traffic patterns closely enough to quantify the shift.

The company’s Q3 2025 Threat Insights Report, released in December, analyzed trillions of requests across Fastly’s global network. The data shows bots now account for 29% of all web traffic, with AI crawler traffic highly concentrated among a small number of platforms. Meta accounted for 60% of all AI crawler traffic in Q3, up from 52% earlier in the year. Google and OpenAI account for the remaining major share of crawler activity.

Fetcher bots, which retrieve content in real time when users make queries to AI assistants, show different concentration patterns. OpenAI’s ChatGPT and related bots generated 68% of fetcher bot requests. In some cases, fetcher bot request volumes exceeded 39,000 requests per minute to individual sites.

AI agents check multiple websites when processing queries, generating higher traffic volumes per interaction than browser-based users. That traffic runs through Fastly’s network for customers using the platform for content delivery.

Beyond traffic volume, Fastly is seeing AI workloads running directly on its edge compute platform. Compton cited examples including storage of large training datasets and customers using edge compute for inference and other AI-related processing tasks.

Customer strategy shifts from blocking to optimization

The approach large media companies are taking toward AI agent traffic has evolved over the past six months. Compton said the discussion with media customers has shifted from “‘how do you block it?’ to a much more nuanced and sophisticated conversation now, about ‘how do you optimize for it?'”

Media companies want their content to remain relevant to AI models and chatbots but need tools to manage how that access happens and enforce existing content licensing agreements. Fastly developed AI bot mitigation technology to address this requirement, allowing customers to permit beneficial bots while blocking harmful ones.

The company also became one of the first edge providers to support the Really Simple Licensing (RSL) protocol. Fastly is working with large media customers as design partners to refine how the protocol gets implemented at the edge. “It was an industry-developed protocol to essentially enforce content rights agreements related to AI models,” Compton said.

Infrastructure investment increases for AI workloads

Fastly is positioning its edge infrastructure as a layer for managing “the age of Agentic AI,” as Compton described it. The company’s distributed architecture handles both the increased traffic volumes from AI agents and the edge compute workloads customers are running for AI inference.

Fastly is opening additional points of presence in the Asia-Pacific region to support growth. The company’s software-defined infrastructure approach requires lower capital expenditures for expansion than legacy providers, according to management.

“As the Internet moves into the age of Agentic AI, it’s clear that the edge will play a pivotal role,” Compton said. “Our infrastructure is designed to power this edge intelligence layer, optimizing authorized AI agents and blocking abuse.”