techcrunch.com web signal

AWS, Cloudflare Rebuild Core Systems for AI Agent Traffic

amazon agents ai infrastructure ai-infrastructure agents cloud

Key insights

  • AWS relaunched OpenSearch Serverless on May 28 with scale-to-zero specifically targeting AI agent workloads displacing human browsing traffic.
  • Single AI agent sessions can spawn hundreds of concurrent sub-agent queries in seconds, overwhelming infrastructure sized for human load patterns.
  • Cloudflare and AWS are treating agent-generated bursty traffic as the new baseline design requirement, not an edge case.

Why this matters

Scale-to-zero and sub-second provisioning are no longer differentiators; they are the minimum spec for infrastructure serving agentic AI workloads, and teams still planning around human traffic baselines are underbuilding. AWS's OpenSearch Serverless relaunch signals that major cloud vendors are explicitly pricing and packaging for machine-native traffic patterns, which will accelerate cost restructuring across database, CDN, and API gateway services over the next 12 months. Founders building on AWS or Cloudflare should expect pricing models, default configurations, and SLA structures to shift toward burst-optimized architectures, making today's cost projections unreliable for AI-native products.

Summary

AWS and Cloudflare are redesigning cloud infrastructure for a new reality: AI agents, not humans, generate the majority of internet traffic. A single agent session can spawn hundreds of sub-agents querying databases and APIs in seconds, then disappear entirely. That burst-to-zero behavior is incompatible with systems built for steady human browsing and minutes-long autoscaling windows. Essentially: (AWS, Cloudflare) are retooling core systems for machine-native workloads. - AWS's May 28 OpenSearch Serverless relaunch adds scale-to-zero and sub-second provisioning as baseline requirements. - Cloudflare is repositioning API gateway and CDN offerings around agent-originated traffic patterns. The cloud layer built for the human web is becoming a bottleneck, and the largest providers are moving first.

Potential risks and opportunities

Risks

  • Enterprises locked into long-term AWS reserved instance contracts face stranded capacity costs as agent workloads demand serverless burst pricing instead
  • Cloudflare customers using WAF and DDoS rules calibrated for human traffic patterns may see false-positive blocking of legitimate agent traffic within the next 90 days
  • Smaller cloud providers (DigitalOcean, Linode/Akamai) risk losing AI-native startups to AWS and Cloudflare if they cannot match sub-second provisioning timelines in 2026

Opportunities

  • Observability vendors (Datadog, Honeycomb, Grafana Labs) can sell agent-traffic dashboards and alerting tooling to enterprises whose monitoring still assumes human browsing baselines
  • API gateway and service mesh vendors (Kong, Apigee) gain leverage to reprice burst-tier plans as agent-spawned call volumes make existing flat-rate tiers uneconomical
  • Cloud cost optimization startups (Infracost, CAST AI, Spot.io) have a window to build agent-aware capacity planning tools before AWS and Cloudflare bundle that capability natively

What we don't know yet

  • Whether Cloudflare has published specific technical specs or pricing for agent-optimized traffic routing, as of May 29, 2026
  • What actual traffic split between human and AI agent requests AWS is currently observing across its customer base
  • Whether sub-second provisioning on OpenSearch Serverless applies globally or only in select AWS regions at launch

Originally reported by techcrunch.com

Read the original article →

Original headline: TechCrunch: Cloud Infrastructure Is Being Rebuilt for Machine Traffic — AWS and Cloudflare Redesign Core Systems as AI Agents Overtake Human Web Patterns