Edge cloud startup Zero Latency, formerly known as Hyphastructure, has launched a closed beta for its distributed AI inference grid.

Dubbed Zerogrid, the service routes AI inference workloads to Edge capacity, creating a low-latency inference grid.

The beta program is currently open to a select cohort of Fortune 1000 enterprises, Tier 1 telecommunications and fiber operators, and leading enterprise DevOps application platforms.

The offering is modeled on the concept of behind-the-meter distributed virtual power plants (VPPs). Zero Latency owns Edge computing clusters across the US, and coordinates them as a single "pool" of capacity, provisioning of which is aggregated and dispatched against workloads on a day-ahead and real-time basis, as well as longer-term arrangements.

According to the company, this aligns with Nvidia's AI grid concept - a reference design that has already been adopted by Akamai, as well as several telcos.

"Innovation through decentralization is not a thesis we arrived at recently," said Michael Huerta, co-founder of Zero Latency.

"It is the lens through which we have built, financed, and operated infrastructure for decades. We have applied the successes and hard lessons from deploying decentralized power infrastructure to unlock architectural and routing innovations for AI workloads. Zerogrid is the result: infrastructure designed for an inference world that the cloud was never built to serve."

According to a report from Fierce Network, Zero Latency currently has three Edge facilities live in California and Florida, and is targeting six more this year across the US.

Previously known as Hyphastructure, Zero Latency was founded in 2025 and is based in Charlottesville, Virginia. In October 2025, the company launched a distributed Edge cloud network for physical AI workloads, including autonomous robotics, vehicle-to-vehicle collision avoidance, and smart city infrastructure.