Nvidia Is Tackling a Hidden Problem in Humanoid Robots — And Solving It Could Make the Technology Widespread

Nvidia is trying to solve a critical problem at the center of the humanoid robot boom: how to make powerful AI machines move safely and predictably around people in real-world situations. The AI chipmaker is betting that safety will determine which robots actually make it out of the lab and into factories and homes. 

According to a recent Bloomberg report, Nvidia is developing technology to ensure that humanoid robots, or robots that resemble humans with similar heads, torsos and limbs, can operate safely around people. The company says these robots must be capable of instant, high-stakes decisions before earning a place alongside human workers. 

Nvidia is selling chips and software that will let humanoid robots work in offices and factories and interact closely with people. The company said earlier this week that Halos, its new software built from its self-driving car technology, will give robots a much clearer sense of their surroundings. 

Nvidia and other major Silicon Valley companies are pouring resources into advanced robotics, pitching it as the next major frontier for AI rather than just a niche hardware category. Executives told Bloomberg that as AI moves from screens into physical machines, the industry could eventually support billions of intelligent devices working in homes, factories, hospitals and cities around the world. 

Big Tech companies are building full-stack platforms that combine chips and software to make it easier to design, train and deploy AI-powered robots at scale. Barclays estimates that humanoid robots will generate $200 billion in revenue by 2035. 

The humanoid robot industry still faces challenges

Today’s humanoid robots still operate under strict safety rules that prohibit any possible contact with a human being. In most industrial settings, if a human steps too close or a sensor detects a possible collision, the machine must slow down sharply or come to a full halt to comply with safety standards, per Bloomberg.

According to the outlet, these responses reduce the risk of injury, but also make robots less efficient whenever people share the same workspace. These safeguards limit how naturally humans and robots can work together. Tasks like passing tools back and forth, guiding a heavy object together or adjusting a part while a robot holds it become awkward or impossible when the system is designed to halt at the first hint of contact.

“If you think about safety in the context of a traditional robot, it is basically you need to put it in a cage or you need to have sensors which can detect that there’s an obstacle and the robot comes to a stop,” Amit Goel, a senior director of product management at Nvidia, said in an interview with Bloomberg. “But that is not enough for a humanoid robot.”

Nvidia plans to power robots with technology that lets them make their own decisions based on what they see and sense in real time. In one example, an autonomous forklift could use warehouse cameras as external sensors to look around a corner before turning, then decide whether to keep moving at full speed or slow down to avoid a collision.

“Safety design has to be much more advanced,” Pras Velagapudi, chief technology officer of Agility Robotics, told Bloomberg.