New startup Hyphastructure has launched a distributed Edge cloud network for physical AI workloads.

The company says the network will comprise locally distributed nodes to enable AI inference latency of under 10 milliseconds.

This will particularly target what it calls "physical AI," including autonomous robotics, vehicle-to-vehicle collision avoidance, and smart city infrastructure.

The Edge platform will use Intel's Gaudi 3 AI chips, along with software-optimized networking and bare-metal virtualization. Hyphastructure claims the Intel chips offer up to a 30 percent lower TCO compared to "traditional GPU-based systems."

“Hyphastructure delivers the infrastructure layer AI-powered industries have been waiting for,” said Michael Huerta, CEO of Hyphastructure. “By moving inference to the Edge, we’re removing latency barriers and opening the door to transformative applications from tele-surgery to robotics that were impossible before.”

Hyphastructure was founded in 2025. According to its website, the company describes its offering as a "virtual compute plant" based on how the virtual power plant infrastructure is deployed. Whether any nodes have yet been deployed has not been shared. CEO Michael Huerta is based in New York and is also a venture partner at Contrarian Ventures.