August 26, 2025
By Bob O'Donnell
As exciting as Generative AI may be, there’s another AI-powered technology on the near-term horizon that could end up being even more impactful: the field of robotics—or as Nvidia’s Jensen Huang and others have started to call it: physical AI.
The idea with Physical AI is that some of the same type of algorithmic principles that have been used to create large language models (LLMs) for text-based interactions could be applied to learning and recreating physical movements in the real world.
Nvidia has offered robotics-focused hardware platforms and software for over a decade, but its latest release—the NVIDIA Jetson AGX Thor platform (which is now generally available)—marks a significant step, because it brings the company’s latest Blackwell generation GPU compute architecture into the worlds of both industrial and consumer/humanoid robotics. Traditionally, the company’s robotics offerings have lagged behind the latest generation GPUs. The previous robotics platform, Jetson Orin, for example, was released two years ago, and it was based on the Ampere architecture, which was first introduced over three years ago. With this latest release, however, the company has brought its most advanced AI compute engine—the Blackwell generation GPUs—to the field of robotics. In fact, in doing so, it represents the first time that every one of Nvidia’s various vertical offerings are now leveraging its latest chips.
Because of the big jump in generations, there’s also a big increase in performance with the Blackwell-powered Jetson AGX Thor versus the Jetson Orin. Nvidia claims that the Jetson T5000 production module offers an impressive 7.5x increase in AI computing power and 3.5x improvement in energy efficiency. More important than the raw number increases, however, are the types of applications that are likely to be available when using the new board. Notably, Nvidia believes the new platform can enable a new range of humanoid-style robots that could be leveraged across commercial and, eventually, even consumer applications.
Nvidia is offering several variations of its Jetson Thor platform including a development kit, priced at $3,499, that includes a Jetson T5000 board and a number of I/O ports. The T5000 board, priced at $2,999, is also available on its own. It includes a 14-core Arm Neoverse V3AE CPU, 128 GB of memory, features up to 2.070 TOPS for FP4 calculations, and runs at 130 Watts. Later this fall, Nvidia will also offer the T4000 at $1,999, which offers a 12-core Arm Neoverse V3AE CPU, 32 GB of memory, features 1,200 TOPs and runs at 70 watts. Each of these Thor platforms also include a new Blackwell Multi-Instance GPU (MIG), which allows the systems to essentially break the GPU into multiple virtual pieces and respond very quickly and consistently to the wide range of activities and sensors that will be incorporated into robotics systems.
In addition to the hardware, Nvidia has multiple different software platforms and tools specifically targeted at robotics applications. The Nvidia Jetson software platform has been extended to work with these new boards, as has the company’s Isaac Groot humanoid robot foundation models and Nvidia Metropolis for Vision AI, among others.
Of course, when it comes to robots—particularly for consumer applications—there are a huge number of issues to think about beyond just the enabling technology. The social, psychological, and economic impact of humanoid-style robots starting to enter our lives is the kind of thing that’s likely going to take years to fully digest. This is, after all, true science fiction come to life, and I expect to see a wide range of potential issues starting to be raised and discussed. To summarize it simply, Rosie the Robot or the Terminator, which will it be?
Even beyond the simplistic good vs. evil debates, there are serious questions about what kind of consumer applications humanoid robots can really fill. Yes, a digital maid—a la the Jetsons’ Rosie the Robot—would likely be attractive to a certain number of households, but others are likely to be horrifically opposed. If people are concerned with the “robotic” autonomous driving features in cars, how do you think they’re going to feel about real robots? Plus, given the cost of the Thor Blackwell computing engine, it’s not difficult to imagine that the first generation of these devices will cost tens of thousands of dollars—well out of reach for all but the highest income households.
For these and other reasons, I expect we’ll see a progression of robotic-powered consumer devices that slowly get us accustomed to the notion of robots in our homes. This will also help with the sticker shock that these new types of applications will undoubtedly include. The problem is, it’s not very clear what these types of non-humanoid applications will be and how attractive they’ll be. Rumors of Apple having a smart display that automatically turns and looks at you while you’re talking to it via a robotic arm, for example, certainly don’t seem like a mainstream product to me.
In the industrial world, the story is very different. There are already huge numbers of robotic devices used in manufacturing, rapidly growing deployments of autonomous guided vehicles (AGVs) in warehouses, and lots of other robotic applications being developed. Here too, humanoid robots that can be used to enter dangerous or difficult environments are immediately appealing and will likely be widely adopted—even at the types of prices they are likely to command.
Of course, we are still very early in the era of Physical AI, and there will likely be unexpected developments that could completely change how the opportunity evolves, particularly in the consumer market. But with the launch of these new platforms, which can essentially be thought of as advanced robotic brains, it’s clear that the technological tools necessary to create advanced robotics applications are quickly coming into place. And, no surprise, Nvidia has managed to position itself very strongly with a combination of advanced hardware platforms and a wide range of robotics platforms and applications that may help build the kind of moat the company created with CUDA in the world of GenAI.
Lots of questions remain, but the time for starting the discussions on robotics applications—particularly humanoid ones—is now.
Here’s a link to the original column: https://www.linkedin.com/pulse/nvidia-brings-blackwell-robotics-bob-o-donnell-slw0c
Bob O’Donnell is the president and chief analyst of TECHnalysis Research, LLC a market research firm that provides strategic consulting and market research services to the technology industry and professional financial community. You can follow him on LinkedIn at Bob O’Donnell or on Twitter @bobodtech.
| 
