Being available for pre-keynote interviews with relevant media to discuss your expertise is generally a good idea. But what if said media hands over the mic and asks you to extemporize?
That’s what Machine Design asked Deepu Talla, vice president of Robotics and Edge AI at NVIDIA, and Ujjwal Kumar, group president of Teradyne Robotics, to do ahead of their keynotes at Automate 2025.
This request wouldn’t work for just anyone. But the pair are seasoned professionals who are primed to pedal through the suite of advanced AI-driven robotics solutions and accelerated computing platforms their respective companies have collaborated on.
READ MORE: AI Gains Physical Intelligence and Transforms Robotics & Automation Design
Talla was gearing up for his keynote the following morning, where he would discuss how physical AI and NVIDIA’s ecosystem are transforming automation “from a static system to a dynamic, adaptive technology.” He would inform audiences that AI is reshaping industrial sectors, from manufacturing and logistics to autonomous robots and vehicles, with a potential impact of up to $50 trillion in global GDP.
NVIDIA means business. The supplier of AI hardware and software, including platforms and services for data science and computing, has earmarked up to half a trillion dollars to scale up AI infrastructure in the United States. The plan is to meet demand for AI chips and supercomputers through partnerships with TSMC, Foxconn, Wistron, Amkor and SPIL, according to press notes.
For his part, Kumar’s keynote would impress upon audiences that physical AI is the biggest opportunity in tech and that scalable automation solutions are ready to solve real-world manufacturing problems right now. He would tell the audience that manufacturers are not looking for a “five-month integration cycle,” nor searching for a “battalion of custom programmers,” but instead seeking out scalable, adaptable solutions, flexible deployment and, ideally, deployment by the operators who are on the floor today.
Pairing AI with Robotics and Hardware
Talla opened the impromptu interview with the first prompt: “Tell me about the new announcements UR made today.”
“This morning we announced our long-awaited UR15 product, which has the next generation of motion control precision and industry leading speed on any cobot platform,” said Kumar. “This will drastically improve the throughput and the performance you can get from any collaborative platform. It also brings in AI capabilities. Most of our robots are either going on an AI application or along with our AI Accelerator, which we developed jointly with NVIDIA. I see a lot of excitement in our end-customer community from automotive to metal and machining, logistics and other segments around this new product launch.”
As a technology platform company, NVIDIA has been testing the types of AI applications it can bring to the industrial market for the past 10 years, but the technology has been slow to take off. According to Talla, however, market readiness and technological advancements are changing that scenario. With challenges such as labor shortages and reshoring, AI-powered robotics and automation solutions can provide bridging solutions.
NVIDIA partners with companies like Teradyne (UR and Mobile Industrial Robots) to integrate its technologies into their platforms and solutions. Among pioneering companies attending Automate, Teradyne, Vention, KUKA and Standard Bots were there not only to showcase their hardware and products designed to automate and optimize production lines, but also to demonstrate simulation tools and solutions that could build virtual robotic worlds, simulate sensor data, generate synthetic data and build the underlying API backbones—all powered by NVIDIA’s Omniverse and Isaac platforms.
Asked whether Teradyne is seeing progress given the surge of activity in Teradyne’s developer and customer base around solving problems using AI, Kumar said that Teradyne had over the past two years been testing AI applications to bring to industrial applications to market.
“There is a crystallization around applications where we can develop as a standard solution,” Kumar said. “The AI Accelerator—if you look at what the NVIDIA and the UR engineers brought in—they have handpicked the most common ways in which we can solve some standard problems for our customers. And it is in metal and machining and automotive and logistics space, and it's preloaded, which is game changing for the industrial world.
“They get a standard, factory-tested set of hardware with a collaborative robotics platform,” he added, “as well as preloaded software to get up and running quickly, which accelerates the deployment and redeployment of robotics solutions.”
The AI Accelerator is an AI-powered toolkit designed by Universal Robots (UR)—a Teradyne Robotics company—in collaboration with NVIDIA. The solution accelerates UR’s next-generation software platform, PolyScope X. Powered by NVIDIA Isaac accelerated libraries and AI models and running on the NVIDIA Jetson AGX Orin system-on-module, the toolkit is an enabler for physical AI. (Physical AI refers to the use of AI techniques that learn about their environment from sensor data and actuators that interact with an environment and enable manipulation of physical objects in that environment.)
Simulation Makes Innovation Faster, Safer, Cheaper
Although NVIDIA has been pioneering simulation for at least a decade, Talla pointed out that until recently “the simulation was not accurate enough.” The technology has sufficiently matured, he said, to the point where simulation enables companies to design and test robots in simulation before deploying them in the real world.
READ MORE: It’s a Big Deal: Industrial Automation Acquisitions, Strategic Growth & Diversification
To apply this in context, Talla asked Kumar to comment on how the partnership between NVIDIA and Teradyne has enabled simulation in the workflow. Kumar responded that a simulation used in palette detection took only eight months to develop versus three years for a similar robotic solution that applied traditional testing methods.
Kumar added that industrial customers were realizing that creating an entire robotic solution “in the simulation world not only decreases the risk of deployment, but it drastically increases the speed with which you can deploy, redeploy, and use the simulation environment also for training of their workers.”
Simulation-First Development Unlocks Scalable Robotics
Reversing roles in the impromptu interview, Kumar asked Talla where he thought their solution could have the biggest potential for traction in the industrial world.
Talla said that two technologies have changed the trajectory in the past year: simulation and AI—that is, physical AI (perception and real-world interaction) and generative AI/Large Language Models (LLMs).
“That same technology that’s been used in the digital IT world, is now coming to the physical robotics world,” said Talla. “Because of these two technologies, we can speed up the development, testing and then obviously the deployment of all these robotic applications. We’re seeing so many of these companies now come and build foundation models that can generalize very well. Problems that we could not solve in the last 10 years for manipulation, gripping, pick and place, are getting much more accurate now. And that’s basically what we’re providing."
