On May 28, Amazon Web Services (AWS) launched the new generation of Amazon OpenSearch Serverless, fully decoupling compute from storage: when agent-triggered tasks occur, compute resources can scale up in seconds, and scale down to zero when idle, stopping billing — even the previous version of Serverless required at least one instance to run continuously. Tia White, GM of AWS OpenSearch Service, told TechCrunch, “Agents create traffic spikes without warning and go silent just as abruptly. Enterprises need a search system that can keep pace without paying for idle compute.” The new version natively integrates with AI development platforms such as Vercel and Kiro, allowing developers to deploy production-ready search and vector database backends for agents without managing the underlying infrastructure. In the same article, Lai Yi Ohlsen, Senior Product Manager at Cloudflare, revealed that bot traffic has accounted for 31% of all HTTP traffic over the past six months, with AI crawlers, search engines, and assistants making up about a quarter of all bot requests. They predicted that “non-human traffic will surpass human traffic in the first half of 2027.”
This AWS release signals a systemic shift across the entire cloud industry: Databricks and Snowflake are positioning themselves as the memory and retrieval systems for enterprise AI; Microsoft has launched Azure updates targeting AI agent traffic bursts and shared memory across agents; and last month, Cloudflare also released an infrastructure product providing persistent environments and instant elastic scaling for agents. The TechCrunch report points to a positive feedback loop between infrastructure reconstruction and the scale of agent deployment: more enterprises are deploying agents for both internal and external scenarios, driving cloud infrastructure to be redesigned for machine traffic, and infrastructure improvements will further reduce the deployment cost and barrier for agents.