NVIDIA UNVEILS $3,499 ROBOTICS CHIP FOR “PHYSICAL AI” AS IT EYES THE NEXT COMPUTING REVOLUTION

Nvidia, fresh off its meteoric rise as the defining company of the artificial intelligence boom, is turning its focus from the digital world of language models to the physical world of robotics.

- 

At its annual robotics and AI summit this week, the company unveiled a new $3,499 robotics chip, specifically designed to power what CEO Jensen Huang called the era of “physical AI.”

The chip—part of a new platform named Jetson Thor—is intended to give robots the computational capacity to perceive, reason, and act in real time. Where GPUs have already become indispensable for training large language models, Nvidia now wants its hardware to become the backbone of machines that can navigate factories, hospitals, warehouses, and even homes.

WHAT “PHYSICAL AI” MEANS FOR NVIDIA

For Huang, physical Al represents the next frontier after the generative Al boom. “Al has mastered language and vision,” he told attendees. “The next step is for Al to understand and interact with the physical world. Robots are going to be the next trillion-dollar Al industry.”

In Nvidia's framing, physical Al refers to systems where intelligence is embodied—machines that don't just interpret data but also manipulate their surroundings. These range from industrial robots that can adapt to changing assembly lines, to autonomous drones that can inspect infrastructure, to humanoid robots designed for labor-intensive tasks.

The new robotics chip is designed to make these visions feasible, packing enough compute power to run complex multimodal Al models locally while staying energy efficient.

INSIDE THE JETSON THOR CHIP

The $3,499 Jetson Thor module integrates Nvidia's latest GPU architecture with high-performance CPU cores, specialized tensor cores for Al inference, and real-time processing engines tuned for robotics workloads. Key features include: