This startup is rebuilding the cloud with AI agents in focus

  • Cloud-native is out, AI-native is in, according to Agentuity
  • The startup is building a new cloud OS made in AI's image
  • It's not entirely clear whether Agentuity's work will show up on hyperscalers' radar

The push to make hardware and software systems “cloud-native” has been all the rage for a while now. But these days, a new viewpoint is emerging: the idea that everything – including the cloud – needs to become “AI-native”.

Enter Agentuity, a startup launched in January with $4 million in seed funding led by Boldstart Ventures, with participation from Bloomberg Beta, Southern Equity, and OneSixOne Ventures. Agentuity is aiming to remake the cloud in AI’s image.

“The cloud was built for humans. AI agents aren’t humans and they need something different, all the way to the infrastructure side and that’s what we’re trying to build,” CEO Jeff Haynie told Fierce.

So, what on earth is it building? An entirely new kind of computing, Haynie said, one that’s designed for machine-to-machine rather than human-centric interaction from the ground up.

Haynie said the company is building an agent operating system that sits on top of virtualized hardware. Right now, the latter is commodity cloud hardware (you know, because they were founded six months ago), but the OS it is building is cloud agnostic and the goal is to move the system onto in-house data center infrastructure at some point in the future.

Its OS runs on the open-source Xen hypervisor and most of the rest of the system — including security, storage and vector capabilities — is being built from scratch.

Haynie explained that’s because most of the tools today just don’t offer the capabilities needed for what Agentuity is trying to achieve.

“To do agent orchestration is wildly different than API orchestration. So, we’re having to build pieces of the stack that maybe if you were building something else you wouldn’t have to do that,” Haynie explained.

Right now, Agentuity is a 10-person shop. But it’s also eating its own dog food, using around two dozen AI agents to help write code for the new operating system. Haynie said around 35% of its code was written by these agents.

The company just went public with its work on June 2 and will showcase its wares at the AI Engineer’s World’s Fair June 3-5 in San Francisco.

Haynie told Fierce while Agentuity has been doing private previews with a number of users, things are still pretty early on the customer side of things.

To date, it has won a contract for a public-private project in Florida to use its platform to help crunch data for water quality treatment and safety. Haynie said it is hoping to work with a lot more customers over the next six months.

Hyperscaler implications

What does this all mean for the infrastructure hyperscalers have deployed and continue to build? Well, it depends who you ask.

Haynie said it could involve some refactoring of their systems. But J. Gold Associates Principal Jack Gold said “if it’s not built by the hyperscalers they’re probably going to ignore it.”

That’s especially true, Gold said, given the rise of model context protocol (MCP) and the Agent2Agent (A2A) protocol, both of which are designed to allow agents to more seamlessly interact with various data sources and each other.

We’ll be watching closely to see if Agentuity can carve out a comfortable niche for itself.