EV charging firm Xeal plans to deploy 100,000 Nvidia GPUs at roadside sites

Xeal plans to use idle capacity at its 1,600+ US EV charging sites to run distributed inference on over 100,000 Nvidia GPUs housed in Latient Pods of up to 48 Hopper or Blackwell Ultra accelerators each. The first pod goes live by the end of the year, with sub-20ms Metro Edge latency claimed.

• Xeal has 1,600+ sites and 200MW of permitted power, typically used at under 10%
• Each Latient Pod holds up to 48 Nvidia Hopper or Blackwell Ultra GPUs
• Pods use self-contained cooling with no water hookup and run under 65 dB
• Metro Edge latency is claimed under 20ms; first pod online by year-end