NVIDIA Introduces AI Robot Training Tech At CES 2025

NVIDIA CEO Jensen Huang introduced an array of new technologies developed by his company at this year’s edition of the annual tech conference Consumer Electronics Show (CES) in Las Vegas, which included artificial intelligence (AI)-driven robot training tools, as well as cutting-edge gaming chips. 

Speaking at CES 2025, Huang commented on the pace of change in AI, saying, “It started with perception AI — understanding images, words, and sounds. Then generative AI — creating text, images, and sound. Now, we’re entering the era of physical AI, AI that can proceed, reason, plan, and act.” 

One of the key products Huang showcased was the NVIDIA Cosmos world foundation model platform, which was billed as “a game-changer for robotics and industrial AI.” 

In a statement, Cosmos has been described as a platform that can integrate generative models, tokenizers, and a video processing pipeline to power physical AI systems like autonomous vehicles (AVs) and robots. 

Huang explained that Cosmos models ingest text, image, or video prompts and generate virtual world states as videos, adding, “Cosmos generations prioritize the unique requirements of AV and robotics use cases like real-world environments, lighting, and object permanence.” 

Huang likened the potential impact of Cosmos – which is available under an open license on GitHub and has been adopted by leading companies such as Uber, Hyundai, and XPENG – to the transformative role of large language models in generative AI. “The ChatGPT moment for general robotics is just around the corner,” he declared. 

In gaming, Huang introduced the GeForce RTX 50 series, based on NVIDIA’s Blackwell architecture. The flagship RTX 5090 GPU features 92 billion transistors and performs 3,352 trillion AI operations per second (TOPS). The new chips aim to enhance video game graphics by adding realistic details such as imperfections on objects and lifelike human faces. 

At CES 2025, Huang also unveiled AI foundation models for RTX PCs, accelerated by NVIDIA’s new GPUs. These models enable developers to create digital humans, podcasts, images, and videos. Additionally, Huang introduced NVIDIA’s first desktop computer, Project DIGITS, aimed at software developers. “Every software engineer, every creative artist — everybody who uses computers today as a tool — will need an AI supercomputer,” Huang said.  

Huang also presented developments in autonomous vehicles. NVIDIA DRIVE Hyperion, built on the new AGX Thor system-on-a-chip, provides safety and autonomous driving features. Huang highlighted the role of synthetic data in training AV systems. “Using Omniverse [a real-time 3D graphics collaboration platform created by NVIDIA] and Cosmos, we can scale hundreds of drives into billions of effective miles,” he said. Toyota, Hyundai, and other automakers are integrating NVIDIA DRIVE AGX systems into upcoming vehicles. 

In robotics, NVIDIA’s Isaac GR00T Blueprint is set to support developers in training humanoid robots using synthetic motion data created in NVIDIA Omniverse. The Mega blueprint enables robot fleet simulation, with companies such as KION and Accenture using it for warehouse operations. 

NVIDIA also announced a partnership with Toyota that will see the latter integrate the former’s Orin chips and automotive operating system into several models for advanced driver assistance. While specific models weren’t disclosed, the deal represents a significant expansion into the automotive sector. NVIDIA expects to generate $5 billion in automotive hardware and software revenue by fiscal 2026, up from $4 billion this year.