Slate Auto to Reveal EV Pricing and Open Preorders on June 24

For years, cloud infrastructure has been built around predictable human behaviour — users searching, clicking, scrolling, and streaming content at relatively steady rates. AI agents, however, operate very differently. They can suddenly trigger large bursts of activity, launching multiple sub-agents that search databases, scan documents, and interact with APIs simultaneously before disappearing just as quickly.

To address this shift, Amazon Web Services (AWS) is redesigning part of its cloud infrastructure. On Thursday, the company introduced a new generation of OpenSearch Serverless, its fully managed search and vector database platform. AWS says the upgraded service is specifically engineered for agentic AI workloads, allowing systems to rapidly scale up when AI agents initiate tasks and scale back down to zero when no activity is occurring.

The launch reflects a broader industry trend, as technology companies are recognising that infrastructure designed for a human-driven internet may be insufficient for a future increasingly shaped by autonomous AI systems.

Although AI agents currently account for only a fraction of overall internet activity, machine-generated traffic is growing rapidly. According to Cloudflare, bots accounted for 31% of all HTTP traffic over the last six months. AI-powered crawlers, search tools, and assistants accounted for roughly one-quarter of all bot-generated requests during that period.

"No"-human traffic will exceed human traffic sometime in the first half of 2027," said Lai Yi Ohlsen, senior product manager at Cloudflare.

The growing role of AI agents was also highlighted at Google's I/O developer conference, where the company showcased systems capable of researching purchases, booking travel, browsing websites, and interacting with applications on users' behalf. Beyond consumer applications, businesses are increasingly deploying AI agents internally and externally, generating entirely new categories of machine-to-machine traffic.

As organisations adopt more autonomous systems, cloud providers are being forced to rethink how infrastructure handles workloads that constantly retrieve information, execute tasks, call tools, and communicate with other systems.

That challenge is one of the primary reasons for AWS's test of the OpenSearch Serverless update.

"Th" timing is straightforward. Agents are moving from experimentation into production, and they create traffic patterns that previous infrastructure simply wasn't designed for," said White, general manager for Amazon OpenSearch Service. "They spike without warning, they go idle without notice, and enterprises need search systems that can keep up without paying for unused compute resources."

A "major technical enhancement in the new version is the separation of compute resources from storage. This allows compute capacity to scale within seconds during periods of heavy agent activity and then shrink back to zero when demand disappears, eliminating costs associated with idle infrastructure.

"Pr" viously, even in our earlier Serverless version, at least one instance had to remain running because storage and compute were linked together," W" ite explained. "Yo" couldn't automatically scale compute exactly when you needed it, which meant customers always paid for reserved capacity whether it was being used or not."

A "S compares the new approach to paying only for a parking meter while a vehicle occupies the space, rather than renting a permanent parking spot.

At launch, OpenSearch Serverless will integrate directly with AI development platforms, including Vercel and Kiro, enabling developers to deploy search and vector databases for AI applications without managing the underlying infrastructure.

The shift toward AI-native infrastructure is occurring across the cloud industry. Companies such as Databricks and Snowflake are increasingly positioning their platforms as AI memory and retrieval layers for enterprise data. Microsoft has introduced Azure enhancements designed to support sudden AI agent activity and shared memory capabilities between agents. Cloudflare has also launched services focused on providing scalable environments for autonomous systems.

As AI agents become more widely deployed, demand for infrastructure optimised or machine-generated workloads is expected to increase significantly. Industry observers believe these changes could ultimately reduce deployment costs and make large-scale AI agent operations more practical for businesses.