NVIDIA (NVDA) investor relations material

Opening and industry context

• Computer graphics is at a pivotal moment as graphics, simulation, and AI converge, setting the stage for new creative tools and digital twins.

• Over three decades, major innovations like programmable GPUs, CUDA, and RTX have transformed graphics and enabled breakthroughs in AI and real-time rendering.

• The convergence of AI and graphics is driving the next era, where virtual worlds and robotics will be trained and tested in synthetic environments before real-world deployment.

DLSS 5: Technology and innovation

• DLSS 5 builds on the integration of RT and Tensor cores, aiming for real-time photorealism by combining traditional rendering with generative AI.

• The core approach uses the rendered frame as a dense, pixel-aligned signal for generative models to enhance realism while preserving artistic intent.

• Three main challenges addressed: maintaining narrative intent, ensuring temporal coherence, and achieving real-time speed with a compact diffusion transformer model.

• DLSS 5 offers granular artistic controls, allowing developers to direct enhancements at global, scene, or object levels, with adjustable structure and tone intensities.

• The technology is designed to extend, not replace, existing pipelines, and will ship in fall 2026.

AI-accelerated simulation and science

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

• Synthetic data generation and model compression allow for rapid, accurate predictions in complex domains like aerodynamics, with up to a million-fold data compression.

• Open-source frameworks and datasets, such as PhysicsNeMo and Earth-2, support the development and deployment of these AI-accelerated simulations.