NVIDIA made a bunch of latest bulletins via a keynote speech from its CEO delivered at CES 2025 in Las Vegas on Monday.
The primary occasion was the GeForce RTX 50-series, the corporate’s first flagship desktop GPU providing in over two years. The graphics playing cards are highly effective sufficient to run AI fashions and may multiply gaming body charges by as much as eight instances with DLSS 4 neural rendering expertise.
Moreover, Jensen Huang, sporting his signature leather-based jacket, unveiled a brand new NVIDIA supercomputer for AI researchers, a modality for humanoid robots that enables them to study by imitation, a world basis mannequin platform, and a partnership with Toyota for autonomous automobiles.
GeForce RTX 50-series desktop GPUs
Essentially the most superior of the brand new GeForce sequence is the RTX 5090, with 92 billion transistors and delivering 3,352 trillion AI operations per second.
“The GeForce RTX 50 Series delivers breakthroughs in AI-driven rendering, including neural shaders, digital human technologies, geometry and lighting,” NVIDIA claims in a press launch.
The RTX 5090 will begin at $1,999, marking a $400 enhance over its predecessor, the RTX 4090, however additionally it is twice as quick. The RTX 5080 delivers 1,800 TOPS and can begin at $999. Each might be obtainable from Jan. 30.
Picture: NVIDIA
The RTX 5070 Ti gives 1,400 TOPS, just like the RTX 4090, and can price $749. In the meantime, the RTX 5070 presents 1,000 TOPS at $549. Each might be obtainable in February.
Acer, ASUS, Dell, HP, Lenovo, and Razer have designed laptops that embody the brand new RTX 50-series. Gadgets that includes all chips might be obtainable in March apart from RTX 5070 laptops, which is able to launch in April.
“Fusing AI-driven neural rendering and ray tracing, Blackwell is the most significant computer graphics innovation since we introduced programmable shading 25 years ago,” Huang stated within the speech.
Extra about Innovation
Private supercomputer for AI researchers
On high of the brand new chip sequence, NVIDIA launched the world to its private AI supercomputer; Undertaking DIGITS. That is meant for use by AI researchers, information scientists, and college students to “supercharge AI innovation,” in response to NVIDIA.
The DIGITS laptop accommodates the GB10 Grace Blackwell superchip, composed of a GPU and 20-core CPU, which delivers as much as a petaflop of computational energy. It will possibly run massive language fashions of as much as 200 billion parameters — the variety of parameters is usually proportional to efficiency capability.
Full with 128GB of reminiscence, as much as 4TB of flash storage, and NVIDIA’s DGX working system, the computer systems might be obtainable in Could. It’s going to begin at $3,000.
Picture: NVIDIA
Imitation functionality for humanoid robots
In March 2024, NVIDIA unveiled GR00T — a general-purpose basis AI mannequin for humanoid robots. At CES 2025, Huang introduced a brand new modality, or blueprint, for GR00T that enables it to study by imitation.
In different phrases, people will be capable of act out actions they need the robotic to carry out, it is going to seize them as a digital twin, after which execute them repeatedly. This will all be carried out utilizing an Apple Imaginative and prescient Professional headset.
The concept is that the expertise will make it simpler to teach massive fleets of robots deployed in manufacturing settings. “Physical AI will revolutionize the $50 trillion manufacturing and logistics industries. Everything that moves — from cars and trucks to factories and warehouses — will be robotic and embodied by AI,” Huang stated.
This new blueprint enhances the blueprint for Actual-Time CAE Digital Twins launched in November, which facilitates the creation of interactive simulations for engineering functions. Each leverage NVIDIA’s Omniverse platform for real-time 3D collaboration.
As well as, Huang unveiled different blueprints that may construct customized AI brokers for enterprises with NVIDIA NIMs: microservices containing the APIs, domain-specific code, optimized inference engines, and enterprise runtime wanted to run generative AI. He additionally introduced the Llama Nemotron household of open massive language fashions to help with creating and deploying brokers. TechRepublic predicted on the finish of the 12 months that the usage of AI brokers will surge in 2025.
Picture: NVIDIA
World basis fashions for robotics
Huang introduced a brand new platform for “world foundation models.” These AI fashions are impressed by how people kind their psychological fashions of the world, referred to as Cosmos. The primary Cosmos fashions launched are designed particularly for robotics, specializing in “physics-based simulation and synthetic data generation.”
“Cosmos WFMs are purpose-built for physical AI research and development, and can generate physics-based videos from a combination of inputs, like text, image and video, as well as robot sensor or motion data,” in response to NVIDIA.
SEE: NVIDIA: Blackwell Structure Will Speed up AI Merchandise in Late 2024
They may all be educated on 9,000 trillion tokens, or items of video information, together with 20 million hours of knowledge from autonomous driving, robotics, human interactions, and different domains. Firms reminiscent of Uber and Wayve have already dedicated to utilizing Cosmos fashions to be used circumstances together with self-driving automobiles.
The Cosmos fashions might be obtainable publicly via NVIDIA, GitHub, and Hugging Face.
Autonomous driving partnership with Toyota
Lastly, Huang introduced a brand new partnership between NVIDIA and Toyota to develop next-generation autonomous automobiles. Toyota’s upcoming fashions will incorporate NVIDIA’s DRIVE AGX Orin supercomputer and DriveOS working system.
“The autonomous vehicle revolution has arrived, and automotive will be one of the largest AI and robotics industries,” he stated in a press launch.
