Nvidia advances graphics and simulation with agentic and physical AI at SIGGRAPH

8:00 AM · July 20, 2026

Nvidia research and engineering leaders presented a keynote at SIGGRAPH 2026 detailing advances in neural rendering, world foundation models, and AI driven simulation, according to the company's own blog. Neil Ashton, a distinguished engineer focused on AI physics and agentic AI for computational engineering, Ming-Yu Liu, who leads Nvidia's Cosmos Lab and works on world foundation models for physical AI, and Edward Liu, who directs applied deep learning research behind Nvidia's DLSS and neural rendering technology, jointly laid out how these threads are converging. Their central argument is that the field is moving from generating isolated static assets, a single image or a short video clip, toward building dynamic, agentic environments and full world models that can simulate how a space or system behaves over time and respond to interaction, a shift with direct implications for creative tools, industrial design, and robotics alike. The presentation ties together work Nvidia has been pursuing across seemingly separate product lines, gaming graphics, robotics simulation, and enterprise design tools, under a single technical thesis about where generative AI for visual and physical content is headed next. For an audience of graphics researchers and industry practitioners, the keynote functions as both a technical update and a signal of where Nvidia expects the next wave of platform investment in this space to be concentrated.