Could humanoid robots get a lot more human? Nvidia may have made that possibility a bit realer today with a smarter robot brain that has less energy demands.
The tech giant’s latest robotics offering is Jetson Thor, a super computer built for real-time AI computation on humanoid robots and smart machines alike, Nvidia announced in a press release on Monday.
The new module is built to handle larger amounts of information at less energy than previous model Jetson Orin. Powered by the latest Blackwell GPUs, Jetson Thor has more than seven times the AI compute power and twice the memory at more than three times speed and efficiency than its predecessor, Nvidia claims.
All this new power is supposed to unlock higher speed sensor data and visual reasoning that can help humanoid robots get better at autonomously seeing, moving, and making decisions.
“Jetson Thor solves one of the most significant challenges in robotics: enabling robots to have real-time, intelligent interactions with people and the physical world,” the company wrote.
It’s a considerable performance leap that Nvidia hopes will appeal to engineers. The company says early adopters include Amazon, Meta, Caterpillar, and Agility Robotics, a startup that makes commercially available humanoid robots for warehouses and other manufacturing facilities. The model is being considered for adoption by John Deere and OpenAI.
It’s also being adopted by research labs at Stanford, Carnegie Mellon, and the University of Zurich, to power autonomous robots in medical research settings and more, Nvidia said in a blog post on Monday.
The developer kit Jetson AGX Thor, which includes the Jetson T5000 module plus a reference carrier board, power supply, and an active heatsink with a fan, is now on sale on the company’s website starting at $3,499.
Coming soon—and available now on pre-order—is Nvidia Drive AGX Thor, a developer kit using the same technology but for autonomous vehicles instead. Deliveries for that are slated to start in September, the company said.
Nvidia’s growing bet on robotics
Although AI chips are Nvidia’s bread and butter, the tech giant is betting big on robotics and autonomous vehicles.
“This is going to be the decade of AV [autonomous vehicles], robotics, autonomous machines,” CEO Jensen Huang told CNBC in an interview in June.
Huang elaborated on his trust in just how much the robotics industry can scale at the company’s annual shareholders meeting later that month.
Along with AI, Nvidia expects robotics to provide the largest growth for the company, and combined, the two represent “a multitrillion-dollar growth opportunity,” Huang told investors.
Earlier this year, the company also released a family of AI models that can be used to train humanoid robots, called Cosmos.
Huang’s bet isn’t an empty one. Humanoid robots are advancing.
Just last week, China, one of the key players in the global robotics race, hosted its first-ever robot Olympics, World Humanoid Robot Games. At the three-day spectacle, companies showcased robots that can complete a 1,500-meter race in just a little over six seconds and achieve practical job skills like sorting medicine or taking food orders.
But still, the technology is hugely limited and far from widespread adoption. Even at the great robotics showcase in China, many of the robots suffered technical difficulties. One robot in the track and field race even ran straight into and knocked over a bystander walking off-course.
Big week ahead for Nvidia
Nvidia made the announcement at a rather convenient time for the company. The tech giant is reporting fiscal second quarter earnings on Wednesday afternoon, and the market is buzzing already.
Nvidia dominates the AI market, so the company’s earnings always draw huge speculation, but the importance this week is boosted by volatile policy changes and questions around the economic value of wide-scale AI adoption.
The company has been on a policy rollercoaster ride in its efforts to sell AI chips in China amidst the escalating trade war between Beijing and Washington. China is a major market for Nvidia, and the uncertainty is keeping company investors at the edge of their seats.
Also keeping investors occupied is a concerning new AI report from MIT researchers. The report found that despite the bold bets on AI in the corporate world, fewer than one in 10 AI pilot programs have translated to real revenue gains.
Nvidia just hit $4 trillion market value last month, becoming the first public company to achieve the feat. Now, the stakes are high, as it’s up to the tech giant to prove that it’s valuation is not just built on AI hype.