LEM Surgical’s Dynamis robot maps Nvidia ‘physical AI’ roadmap at CES 2026

Dynamis is a tangible example of “physical AI” moving from the lab floor to mainstream trade shows and clinical settings. Unlike many CES demos, this robot is already FDA-cleared and in routine use for spinal procedures at a US hospital, with LEM now explicitly planning to layer Nvidia’s Jetson Thor and Isaac for Healthcare to boost autonomy and real-time perception. That’s a powerful combination: domain-specific robotics expertise plus a general-purpose physical AI stack from Nvidia. ([ggba.swiss](https://ggba.swiss/en/lem-surgical-showcases-its-dynamis-robotic-system-and-ai-roadmap-at-ces-2026/))

For the AGI race, this matters less as a standalone surgical product and more as a sign of where embodied intelligence is heading. Surgical robotics is one of the harshest proving grounds imaginable—high stakes, constrained environments, and deeply skeptical regulators. If open-world models and simulation-heavy training can make a dual‑arm robot a reliable assistant in that context, similar architectures will likely spill over into industrial manipulation, warehousing, and even domestic robots. Each deployment generates rich multimodal data (video, force, EMR context) that can feed back into better world models and decision policies.

Nvidia’s presence in the roadmap is also revealing. The company is not just selling GPUs; it’s trying to own the full software and tooling stack for physical AI. That further blurs the line between "training compute provider" and "vertically integrated robotics platform," and gives frontier labs a ready-made channel for putting increasingly capable models into contact with the real world.