- AI bot traffic is spiking cloud costs
- Agentic AI will amplify cloud infrastructure demands
- But managed AI traffic can create value
AI bots are driving sharp increases in network traffic and cloud computing costs, according to industry experts. Emerging agentic AI systems could further accelerate cloud spend by increasing autonomous workload activity.
We’ve written before about the debate around just how much AI will impact network traffic, and according to a recent report from Akamai, AI bot traffic has surged by 300% over the past year. It turns out that all that traffic comes with a hefty price tag.
Tom Howe, director of Field Engineering at data platform Hydrolix, recently noted one of its customers ended up with a six-figure overcharge from their ISP due to bot traffic. “The traffic bypassed the firewall, reached the origin web server and drove up costs by six figures,” he wrote in a blog.
It turns out that AI bots contribute to ISP and cloud costs in a few different ways. Ari Weil, VP of Product Marketing at Akamai, explained to Fierce that many AI bots contribute to high request volumes with terrible cache efficiency.
“Many AI bots bypass or invalidate caching assumptions, driving traffic back to origin or using premium edge compute path,” he said. “This is much more expensive.”
AI bots also tend to repeatedly fetch the same content, which contributes to egress, compute and storage costs, he added. Needless to say, there’s no incremental value for businesses associated with these increased costs.
“AI bots not only contribute to rising cloud cost, but they are causing the entire economics of the web to break down,” Weil said. “AI bots are generating all of these additional infrastructure costs, without generating any ad revenue, merchandise sales or new subscriptions for the enterprises.”
The agentic AI impact
Today, Weil said agentic AI only represents about 1% of the total AI bot traffic Akamai observes across its network each day. He added that’s expected to drastically increase over time.
Cisco ThousandEyes VP and GM Joe Vaccaro told Fierce that Agentic AI will have a slightly different impact on networks and current AI bots.
Where bots are just periodically indexing certain pages, “agentic systems generate unpredictable, machine to machine interactions as they carry out specific tasks, often touching multiple services in parallel,” he said. The end result is not only more traffic but more sustained traffic.
“The impact for networks and cloud infrastructure isn’t just volume or cost, but the need to support workflows that depend on a larger, more diverse, less predictable set of infrastructure and service dependencies,” he said.
Weil argued AI agents could actually be better for web economics because some agentic traffic will inevitably “represent real purchase intent, subscription activity or other forms of qualified demand, which create value” for enterprises.
“The challenge is not the existence of agentic traffic, but ensuring it is authenticated, governed and aligned with economic value, rather than operating as unmanaged and unpriced automation,” he concluded.