58: NVIDIA’s Deepu Talla on the Next Era of Robotics

Physical AI is fundamentally reshaping robotics by transitioning from rigid, pre-programmed tasks to general-purpose intelligence capable of operating in unstructured environments. Deepu Talla, Vice President and General Manager of Robotics and Edge AI at NVIDIA, highlights how advancements in large language models and high-fidelity simulation are closing the "sim-to-real" gap, enabling robots to be trained safely and efficiently in virtual worlds. Humanoid form factors offer significant potential for integration into human-centric spaces, while synthetic data generation provides the necessary fuel for training these complex systems. As global manufacturing undergoes a period of reshoring, the integration of intelligent robotics promises to address labor shortages and boost productivity. However, scaling these AI-driven systems remains constrained by energy availability, necessitating architectural innovations that maximize computational performance within fixed power budgets to drive the next generation of industrial automation.