The internet is being rebuilt for machines

The traditional cloud infrastructure has been primarily built for human interaction, where users search, click, and stream in predictable manners. However, with the rise of AI agents, this model is being challenged. These agents operate differently; they can quickly initiate numerous tasks, querying multiple databases and calling APIs in mere seconds, only to vanish as swiftly as they appear. In response to this shift, Amazon Web Services (AWS) has introduced its latest innovation: OpenSearch Serverless. This new, fully managed search and vector database is engineered specifically for workloads driven by AI agents. AWS claims that this system can instantly scale its capabilities when agents trigger tasks and reduce to zero when not in use, marking a significant departure from traditional systems designed for human users. The tech world is beginning to recognize that infrastructures built for human traffic are ill-suited for a landscape increasingly filled with agent-driven activity. Although AI agents currently represent a smaller fraction of internet usage, machine-generated traffic is substantial and set to expand. According to Cloudflare, bots made up 31% of overall HTTP traffic in the past six months, with AI crawlers and assistants accounting for about a quarter of those requests. Li Yi Ohlsen, a senior product manager at Cloudflare, stated to TechCrunch that "non-human traffic will surpass human traffic in the first half of 2027." At a recent Google I/O developer conference, the tech giant announced that users would soon be able to delegate various tasks to AI systems, from researching products to booking travel. This evolution isn't limited to consumer-facing AI; businesses are increasingly implementing AI agents internally, generating new forms of machine-driven traffic. Consequently, cloud providers are grappling with the necessity to adapt their systems for this new reality, where agents autonomously access information and generate machine-to-machine communications. According to Tia White, general manager for Amazon OpenSearch Service, the timing for this new infrastructure is clear: "Agents are transitioning from experimentation to real-world applications, creating traffic patterns that previous systems weren't designed to handle." She added that the new system accommodates sudden spikes in activity and idles when not needed, allowing for cost savings. A key advancement of OpenSearch Serverless is its decoupling of compute from storage. This allows the system to rapidly scale up to meet demand from agents and scale down to zero, meaning users won’t incur costs when agents are idle. Previously, users had to maintain at least one active instance, leading to unnecessary costs. With this upgrade, AWS's offering is akin to a metered parking system, where users only pay for what they utilize. The new OpenSearch Serverless will also seamlessly integrate with AI development platforms like Vercel and Kiro, enabling developers to set up production-ready search and vector backends for agents without the burden of managing infrastructure. The cloud industry is witnessing a broader trend where companies like Databricks and Snowflake are positioning themselves as AI-centric memory and data retrieval solutions. Microsoft has made enhancements to Azure to better accommodate AI agent workloads, and Cloudflare recently launched infrastructure aimed at providing agents with persistent environments and immediate scalability. As the deployment of AI agents increases, there will be mounting pressure to reshape cloud infrastructures to better support machine-generated workloads, ultimately making these technologies more accessible and cost-effective.