NVIDIA (NVDA) SIGGRAPH 2026 Keynote summary

Opening and industry context

• Computer graphics is at a pivotal moment, with the convergence of graphics, simulation, and AI shaping the next generation of creative tools and digital twins.

• Over three decades, major innovations like programmable GPUs, CUDA, and RTX have driven industry transformation, with AI now deeply integrated into graphics workflows.

DLSS 5: Technology and innovation

• DLSS 5 introduces a new paradigm by combining traditional rendering with generative AI to enhance real-time photorealism, building on the foundation of RT and Tensor cores.

• The system uses the rendered frame as a grounding signal, ensuring enhancements respect artistic intent and maintain narrative consistency.

• Three core challenges addressed: preserving artistic intent, achieving temporal coherence, and meeting real-time speed requirements through a compact diffusion transformer model.

• Developers gain granular control over enhancements, with options to adjust models, structure, tone, and apply selective masking to specific scene elements.

• DLSS 5 is positioned as an extension to existing pipelines, not a replacement, and will ship in the fall.

AI-driven simulation and science

• AI models trained on vast simulation data now enable weather and climate predictions in seconds, matching or exceeding traditional simulation accuracy.

• Data compression techniques allow massive simulation datasets to be distilled into compact models, enabling rapid, accurate predictions for unseen scenarios.

• Open-source frameworks like PhysicsNeMo and Earth-2 support the development and deployment of these AI-accelerated simulations.

• The approach is expanding from fixed-geometry domains like weather to complex engineering challenges in aerospace and automotive design.