"The GPT Moment in Graphics"! DLSS 5 Major Release Brings Hollywood-Grade Visuals to Enterprise Computing

From March 16 to 19, local time, NVIDIA (NVDA.US) GTC 2026 conference will be grandly held in San Jose, California. At 2:00 AM Beijing time on March 17, NVIDIA CEO Jensen Huang delivered the highly anticipated keynote speech at GTC 2026.

In his keynote, Huang officially announced DLSS 5, a new version of NVIDIA’s AI graphics technology designed to make video games more realistic while reducing computational resource usage. DLSS 5 combines traditional 3D graphics data with generative AI models to predict and fill in parts of images, enabling NVIDIA’s GPUs to generate detailed scenes and realistic characters without rendering every element from scratch.

DLSS 5 can run in real-time 4K, understanding scene semantics (such as characters, hair, subsurface scattering of skin, fabric gloss, etc.) and injecting a sense of “physicality” into each frame, while giving developers detailed control (intensity, grading, masks, etc.). This means game visuals are no longer just rule-based approximations but are enhanced in real-time by trained models while maintaining determinism.

Huang stated that this is the most significant breakthrough in computer graphics since real-time ray tracing in 2018: through real-time neural rendering models, injecting “cinematic” lighting and material details into pixels, aiming to achieve Hollywood-quality interactive visuals in games.

Huang compared DLSS 5 to a “GPT moment in graphics,” emphasizing that DLSS 5 seamlessly integrates manual rendering with generative AI, greatly advancing visual realism while preserving the artistic control needed for creative expression.

According to Huang, DLSS 5 will be available for mainstream games this fall and has already received support from major companies including Bethesda, CAPCOM, NetEase (NTES.US), Tencent, and Ubisoft.

Huang said, “We are combining controllable 3D graphics, the benchmark truth of virtual worlds, structured data… with generative AI and probabilistic computing.” “One is fully predictive, the other probabilistic, but highly realistic.” Combining these two concepts—structured data and generative AI—allows developers to create “beautiful, astonishing, and controllable” content. This integration of structured information and generative AI will recur across many industries. Structured data is the foundation of trustworthy AI.

Although gaming revenue now accounts for a smaller proportion of NVIDIA’s income than in the past, it is the gaming industry that made NVIDIA what it is today. Huang framed DLSS 5 as an example of a broader computing transformation, hinting that this approach can extend far beyond gaming and even into enterprise computing.

Using enterprise data platforms like Snowflake, Databricks, and BigQuery as examples, Huang explained that these structured datasets can be analyzed by future AI systems to generate insights. “In the future, these data structures will be used by AI, and AI will be much faster than us. Future intelligent agents will use structured databases as well as unstructured and generative databases. These databases represent the vast majority of the world.”