At SIGGRAPH 2025, Nvidia is unveiling a wave of new products, including compact Blackwell GPUs, enterprise servers, and advanced AI models. The goal is to merge simulation and reality for robotics, autonomous systems, and intelligent infrastructure.
At the SIGGRAPH 2025 computer graphics conference, Nvidia has made a series of announcements that all contribute to a central vision: so-called "Physical AI." According to Nvidia, this term describes the convergence of artificial intelligence and computer graphics to create systems capable of acting in the real world, such as robots, autonomous vehicles, and smart infrastructures.
"AI is advancing our simulation capabilities, and our simulation capabilities are advancing AI systems," explained Sanja Fidler, Vice President of AI Research at Nvidia. The company presented a comprehensive ecosystem, ranging from new hardware based on the Blackwell architecture to simulation platforms and specialized AI models for logical reasoning.
New Blackwell hardware for AI in workstations and data centers
Nvidia has introduced new hardware to serve as the foundation for compute-intensive AI workloads. For data centers, the Nvidia RTX PRO 6000 Blackwell Server Edition GPU is coming to mainstream enterprise servers in the widely adopted 2U form factor. System partners like Cisco, Dell Technologies, HPE, Lenovo, and Supermicro are set to offer corresponding servers. According to Nvidia, these systems are intended to drive the transition from traditional CPU-based architectures to accelerated computing platforms. The company states that the servers can achieve up to 45 times higher performance and 18 times better energy efficiency than CPU-only systems. The GPUs feature fifth-generation Tensor Cores with support for the FP4 format, which is said to increase inference performance sixfold compared to the previous-generation L40S GPU.
For the desktop segment, two new compact graphics cards were announced: the Nvidia RTX PRO 4000 SFF Edition and the RTX PRO 2000 Blackwell. These are designed to bring AI acceleration to smaller, more energy-efficient form factors for fields like engineering, design, and 3D visualization. Compared to the previous generation, the RTX PRO 4000 SFF is said to offer up to 2.5 times higher AI performance at the same 70-watt power consumption. The RTX PRO 2000 is reported to deliver, among other improvements, 1.4 times faster performance in Computer-Aided Design (CAD). According to Nvidia, the new GPUs are expected to be available later this year.
Physical AI: Simulation as the key to robotics
The new hardware is designed to provide the computing power for Nvidia's vision of Physical AI. The core of the concept is the creation of highly realistic, physically accurate digital twins where AI systems like robots can learn safely through trial and error before being deployed in the real world. "Computer graphics and AI are converging to fundamentally transform robotics," said Rev Lebaredian, Vice President of Omniverse and Simulation Technologies at Nvidia.
The technological foundation for this is provided by the Nvidia Omniverse and Isaac platforms. Nvidia has announced new software libraries for Omniverse, including Omniverse NuRec, which enables the reconstruction of real-world environments from sensor data using 3D Gaussian splatting. The robotics simulation applications Isaac Sim 5.0 and Isaac Lab 2.2 are now available as open-source projects on GitHub and also integrate the new rendering techniques.
A concrete application example comes from Amazon Devices & Services with a "simulation-first" approach for so-called "zero-touch" manufacturing. In this process, CAD models of new products are loaded into Nvidia Isaac Sim to generate over 50,000 synthetic images for training AI models. These models then control robotic arms that autonomously perform quality checks or integrate new products into the production line—purely based on skills learned in simulation and without physical hardware modifications. Technologies like the FoundationPose pose-estimation model enable the robots to recognize even new objects without prior training.
New AI models designed to enable logical reasoning
To ensure that AI systems can not only perceive but also reason, Nvidia has introduced expanded AI model families. For enterprise applications, the Nemotron family has been extended with the Nemotron Nano 2 and Llama Nemotron Super 1.5 models. These are intended to enable AI agents to handle complex, multi-step tasks in areas like customer service or cybersecurity. According to Nvidia, the models offer high efficiency through a hybrid architecture and quantization (NVFP4). Companies such as CrowdStrike, Uber, and Zoom are reportedly already testing or planning to use the models.
Developed specifically for Physical AI is Cosmos Reason, a customizable 7-billion-parameter Vision Language Model (VLM). It is designed to enable robots and vision AI agents to interpret and act in the physical world by incorporating prior knowledge, an understanding of physics, and "common sense." Application areas include robot planning, automated annotation of training data, and video analytics. Partners like Uber are using the model to analyze the behavior of autonomous vehicles, while VAST Data and Milestone Systems are employing it for intelligent traffic monitoring.
Smart cities and factories with the Metropolis platform
To translate these technologies into concrete applications for intelligent infrastructure, Nvidia bundles many of the components into the Nvidia Metropolis platform. It has been enhanced with several new features, including the integration of the Cosmos Reason VLM, new vision foundation models in the TAO Toolkit, and extensions for Isaac Sim to generate rare training scenarios.
Partners are already using the platform for various solutions. Accenture and Belden are developing "smart virtual fences," simulated in Omniverse, to enhance worker safety around industrial robots. The company DeepHow is using the Metropolis VSS blueprint for a "Smart Know-How Companion" that transforms work instructions into visual guides. Beverage company Anheuser-Busch InBev is said to have reduced onboarding time for new employees by 80 percent using this solution.
