Humanoid Collapses Robot Compute Onto One Nvidia Jetson Thor

UK humanoid-robot startup Humanoid is deepening its collaboration with Nvidia, consolidating two onboard computers into a single Jetson Thor module while keeping the same compute budget, according to a release carried by Robotics and Automation News.

The next-generation stack layers in Nvidia's Isaac OS, described as "an open, deterministic runtime framework," alongside the Holoscan Sensor Bridge, Isaac Lab and Isaac Sim, with Multi-Gigabit Ethernet carrying up to 100Gb between components. The Sensor Bridge works by placing camera data directly into GPU-accessible memory, cutting the CPU workload and the latency between sensors and the models reading them.

"Building general-purpose humanoid robots requires significant computing power and an architecture that can bring sensing, AI and control together in real time," founder and CEO Artem Sokolov said. "Nvidia's robotics platform gives us a strong foundation to do that."

The move puts another European humanoid builder on Nvidia's full-stack reference architecture, part of a steady run of robotics stories we have logged in the last 90 days that keep routing through the same compute vendor.