At CES 2025, NVIDIA showcased its latest advancements in AI, robotics and graphics processing including the new GeForce RTX series, world-class foundation models and advanced platforms for robotics. With a strong focus on enhancing performance and expanding AI capabilities, NVIDIA revealed innovations that are set to transform industries ranging from gaming to autonomous vehicles and robotics, providing solutions for developers, researchers and businesses worldwide. Below are the advanced innovations launched by NVIDIA at CES 2025:
GeForce RTX Blackwell 50 Series
NVIDIA’s new GeForce RTX 50 Series GPUs introduced at CES 2025, offer significant advancements in creative and AI-powered content creation. Built on the Blackwell architecture these GPUs provide up to twice the performance for video editing, livestreaming, 3D designing and generative AI. With up to 32GB of VRAM and FP4 support, the RTX 50 Series efficiently handles large, complex tasks. Enhanced video encoding features and AI-driven tools improve workflows for creators and streamers. The GeForce RTX 5090, RTX 5080 and RTX 5070 GPUs are set to deliver superior 3D rendering and AI performance.
Grace Blackwell NVLink72
NVIDIA's Grace Blackwell NVLink72 system features 72 Blackwell GPUs and 144 ports, providing an incredible 1.4 exaflops of AI performance. It includes 14 terabytes of high-speed memory and a bandwidth of 1.2 petabytes per second. With 103 trillion transistors and 2592 CPU cores, this powerful system is designed to support advanced AI, big data processing and high-performance computing.
Llama Nemotron Language Foundation Models
NVIDIA has introduced the Llama Nemotron family of open large language models (LLMs) designed for agentic AI. These models help enterprises develop AI agents capable of automating tasks such as customer support, fraud detection and inventory management. Optimized for efficiency and accuracy, the Llama Nemotron models are built on Llama foundation models and support real-time, high-performance applications across a range of devices. These models are available for easy deployment via NVIDIA's cloud infrastructure, providing businesses with scalable solutions.
Cosmos World Foundation Models
NVIDIA has introduced the Cosmos World Foundation Models (WFMs) to enhance the development of autonomous vehicles (AVs) and robotics. These generative models, integrated with NVIDIA’s full-stack systems, enable faster, more precise AI model development by amplifying training data quality. By generating synthetic driving scenarios and simulations, Cosmos accelerates data collection, saving time and resources for developers. Leading companies like Waabi, Wayve and Uber are leveraging Cosmos to build more efficient AV AI systems.
Hyperion Platform
NVIDIA's DRIVE Hyperion platform has achieved significant milestones in automotive safety and cybersecurity, setting new standards for autonomous vehicle (AV) development. It is the industry's first end-to-end AV platform, combining the DRIVE AGX system-on-a-chip, sensor suite and advanced safety features. Certified by TÜV SÜD and TÜV Rheinland, Hyperion ensures compliance with strict safety and cybersecurity requirements. With the latest DRIVE AGX Thor SoC, the platform is scalable, upgradeable and designed to support the next wave of autonomous vehicles and AI-driven functionalities.
Isaac GR00T
NVIDIA introduced the Isaac GR00T Blueprint to speed up humanoid robot development. This technology helps developers create large datasets through imitation learning, allowing robots to learn by mimicking human actions. With tools like Apple Vision Pro and NVIDIA platforms, Isaac GR00T makes robot training faster and more cost-effective, bringing us closer to advanced humanoid robots in industries like manufacturing.
Project Digits AI Supercomputer
NVIDIA introduced Project Digits, an advanced AI supercomputer designed to deliver high performance in a compact form. Powered by the Grace Blackwell Superchip, Project Digits aims to bring AI computing to a wide range of professionals, from engineers to creatives. It runs the full NVIDIA AI stack including DGX Cloud and it is expected to be available in May 2025, offering an innovative tool for AI development and integration.
