Xeal Plans to Deploy 100,000 Nvidia GPUs at EV Charging Sites for Edge Inference

Xeal's Edge Compute Network

EV charging firm Xeal announced plans to establish an edge inference compute network by deploying over 100,000 Nvidia GPUs. These GPUs will be housed in specialized "Latient Pods" at the company's more than 1,600 existing EV charging locations across the United States.

The strategy leverages Xeal's current electrical infrastructure, which includes 200MW of permitted and installed capacity, much of which is currently underutilized. This approach aims to bring new compute online rapidly by tapping into existing grid-connected energy resources.

Latient Pod Specifications

Each Latient Pod is designed to contain up to 48 Nvidia Hopper or Blackwell Ultra GPUs. These self-contained units feature independent cooling systems that do not require water hookups and operate quietly at less than 65 dBs. The pods are built to withstand outdoor conditions with NEMA 4 enclosures.

Xeal states that these pods can be installed and operational within hours, providing a quick deployment solution for compute capacity. The company also highlights potential benefits for site owners, including fixed rental or ancillary revenue and increased property value.

Utilizing Idle Infrastructure

Nikhil Bharadwaj, co-founder and CEO of Xeal, explained that Latient Pods offer a fast method to deploy new compute by using idle EV charging capacity. He noted that EV charging sites typically operate at less than 10% of their permitted capacity, providing available power for compute operations.

This model addresses challenges associated with bringing new compute online, such as grid interconnection, by utilizing already permitted and grid-connected sites. An Nvidia representative acknowledged this as an innovative way to deliver low-latency inference closer to users and optimize existing infrastructure.