The era of robots is approaching, but developers and experts lack accessibility to develop them. Nvidia will introduce the world's first integrated platform that can reduce the time and capital invested in developing not only robots but also autonomous vehicles.
Jensen Huang, CEO of Nvidia, said this during a keynote speech at the 'CES 2025' held at the Mandalay Bay Convention Center in Las Vegas, United States, on Jan. 6. Huang presented Nvidia's AI blueprint, which includes next-generation graphics processing units (GPUs) for PCs, a development platform for robots and autonomous vehicles, and an in-development personal artificial intelligence (AI) supercomputer. This marks his first appearance at CES in eight years since 2017.
On this day, CEO Huang unveiled a new robot development platform called 'Cosmos.' The strategy aims to increase accessibility to development for robots and autonomous vehicles based on Nvidia's semiconductor performance and software development platform. As of now, no platform has been released worldwide that supports the development of both robots and autonomous vehicles simultaneously. This is interpreted as a strategy to dominate the market, similar to how Nvidia has seized market share using GPUs that run AI and the CUDA platform used to develop AI-driven software services.
Nvidia's Cosmos creates a three-dimensional (3D) environment that allows robots and autonomous vehicles to learn in virtual reality exactly as they would in real situations. AI automatically generates events that could occur in a virtual environment, and robots and autonomous vehicles learn through training that responds to their surroundings in this virtual environment. CEO Huang noted, "When going into the real world, errors are inevitable, and if a robot breaks, there are losses," and highlighted the necessity of a platform to develop robots in virtual reality, stating that post-learning is the most challenging part of robot development.
Nvidia also announced plans for developing high-performance personal supercomputers optimized for AI development, including the semiconductors and software used in AI development. Nvidia unveiled the 'Personal AI Supercomputer' being developed based on GB10 (Grace Blackwell). CEO Huang introduced the development process of the personal AI supercomputer, which he referred to as 'Project Digit,' explaining that various AI functions could be developed using an AI model of 200 billion parameters on a PC.
Nvidia unveiled the next-generation graphics processing unit (GPU) 'RTX 50' series equipped with the latest artificial intelligence (AI) accelerator 'Blackwell.' Nvidia stated that it has lowered the price while enhancing performance. The previous model, the RTX 4090, which shows the highest performance among its predecessors, is priced at $1,599, while the RTX 5070, with similar performance, is sold for $549. CEO Huang remarked, "It has three times the performance compared to the previous Ada accelerator-based GPU and is cheaper," adding that "AI enabled cost reductions."