Amid growing competition in the AI chip industry, Nvidia is reportedly shifting focus to robotics. The company plans to release the next version of its compact computers for humanoid robots, named Jetson Thor, in the first half of next year, according to a Financial Times (FT) report from Sunday, December 29. Nvidia, valued at $3. 3 trillion, credits its technology for contributing to the rise of artificial intelligence and aims to become the leading platform in what it sees as an impending robotics surge. Deepu Talla, Nvidia’s vice president of robotics, mentioned to the FT that the market is nearing a significant turning point, describing it as a forthcoming "ChatGPT moment" for physical AI and robotics. This strategic move occurs amid expectations of increased competition for Nvidia's chips from companies like AMD and cloud giants such as Google and Amazon. As a result, Nvidia is investing in the "physical AI" sector to aid the growth of new robotics companies.

For instance, it participated alongside Microsoft and OpenAI in a February funding round that valued humanoid robotics company Figure AI at $2. 6 billion. Although Nvidia doesn't separate its revenues from robotics products, the FT indicates this segment is less significant compared to its overall earnings. Data center revenue, for instance, accounted for about 88% of Nvidia’s $35. 1 billion in sales for the third quarter. In related robotics and AI developments, PYMNTS recently highlighted MIT research on an AI system designed to help warehouse robots manage irregularly-shaped packages and crowded spaces safely. This breakthrough arrives as retailers and logistics companies face increased pressure to automate due to growing eCommerce demands. While robots excel at repetitive tasks like moving pallets, MIT’s PRoC3S technology potentially addresses the issue of robots performing intricate warehouse tasks, requiring human-like dexterity and spatial awareness. Erik Nieves, CEO and co-founder of Plus One Robotics, explained to PYMNTS that PRoC3S might reduce robot error rates by refining initial LLM-based assumptions with more accurate understandings of warehouse environments. He compared it to the difference between classroom learning and a practical field trip, noting that a warehouse robot operating solely on LLM guidance has only been told how to complete a task, whereas the PRoC3S concept involves running a digital robot through a simulated version of that task.