The Internet is being rebuilt for machines

Cloud computing has long been built around people searching, clicking, swiping, and navigating seamlessly and intuitively. AI agents behave differently. They can spawn multiple events, spin up multiple sub-agents that query hundreds of databases, search logs, and call APIs in seconds and then disappear as quickly as they arrived.

Under these conditions, Amazon is also planning a large part of its cloud products. Thursday, AWS introduced the next generation of OpenSearch Serverlessa well-managed search with a vector database – essentially a method of storing and retrieving information at scale – designed specifically for performance. AWS says the new system can scale up immediately when agents start work and drop to zero when they are idle.

The implementation reflects a growing trend in the technology industry: the infrastructure that was built for the human-driven Internet no longer works in a world increasingly populated by agents.

Although AI assistants still represent a small part of the online experience, the amount of automated traffic is already significant, and is expected to grow. Cloudflare says bots account for 31% of HTTP traffic over the past six months. AI crawlers, search engines, and agents made up about a quarter of all bot requests during that time.

“The non-human population will surpass the human population sometime in the first half of 2027,” he said Li Yi Ohlsenchief marketing officer at Cloudflare, to TechCrunch.

At Google’s I/O conference last week, the company said that users will be able to activate it delegating work to others to AI systems, such as shopping searches, booking tours, browsing the web, and interacting with apps. But the buck doesn’t stop at consumer AI assistants. Businesses are increasingly deploying agents internally with their customers, and creating new types of automated vehicles behind the scenes.

As a result, cloud providers and infrastructure companies have been thinking about how to transform human-based systems into a world of automated data-gathering agents, call-to-action tools, and machine-generated traffic.

That’s where AWS’s new OpenSearch Serverless comes in.

“It’s time to focus. Providers are moving away from product testing, and they’re creating traffic patterns that weren’t created before,” Tia White, general manager of Amazon’s OpenSearch Service, told TechCrunch. “They run without warning, they don’t just go unnoticed, and businesses need ongoing investigations without paying for nothing or nothing.”

The biggest change in technology with this new generation is that it removes compute from storage, allowing compute to ramp up in seconds to match traffic bursts and down to zero, so customers pay $0 if servers don’t work.

“In the past, even in our old Serverless model, you had to have the same uptime and run time because the storage and compute were combined,” White said. “You couldn’t just change (calculations) at the rate you needed, so you always have idle computers storing your work, whether you’re using it or not.”

Think of it as always paying for a parking space, even if you don’t use it. With AWS hosted Serverless, it’s like paying for parking.

At launch, OpenSearch Serverless will integrate with AI development platforms such as Vercel and Kiro, so developers can deploy automated search and vector backends to agents without managing infrastructure.

Change is evident in the cloud industry. Databricks is Snowflake They are also positioning themselves as AI memory and business data recovery. Microsoft is out updates to Azure it is designed to handle AI agent bursts and share memory between agents. Cloudflare, like Amazon, last month launched architecture that aims to provide providers with a stable environment and instant scalability.

The more companies deploy AI assistants, the more pressure there will be to redesign the infrastructure around the number of automated tasks, which would make the assistants cheaper and easier to deploy at scale.