On the 7th, local time, CES 2025 opened in Las Vegas, USA, and 12,000 people filled the convention center as NVIDIA CEO Jensen Huang gave a keynote speech.

In his keynote address that day, CEO Jensen Huang mentioned the history of NVIDIA GPUs, which enabled the beginning of computer graphics, and introduced GPUs and NVIDIA's solutions that are achieving today's limitless AI innovation.

The newly released NVIDIA Blackwell-based GeForce RTX 50 series GPUs, as well as Grace Blackwell NVLink72, AI Foundation Model for RTX AI PC, Rama Nemotron model family for Agentic AI, Blueprints for AI agents, and Cosmos WFM platform for physical AI development were unveiled.

In particular, Jensen Huang showed off NVIDIA's extensive influence in the AI market by listing NVIDIA's diverse portfolio, which is represented by AI technology. The contrast was punctuated by a project for a tiny personal AI supercomputer.

Through Project Digits, we unveiled a personal AI supercomputer that brings the power of the NVIDIA Grace Blackwell platform to AI researchers, data scientists, and students worldwide.

This monstrous system-on-a-chip, combining an Arm-based Grace CPU, a Blackwell GPU, and HBM, was designed in secret by Nvidia in collaboration with MediaTek and is called the "GB10 Superchip." Priced at $3,000, it can run massive language models with up to 200 billion parameters.

“AI will become mainstream across every application in every industry,” said CEO Jensen Huang. “Through Project Digits, we will put an AI supercomputer on the desk of every data scientist, AI researcher, and student, empowering them to participate in and shape the AI era.”
Following on-device AI, the next hot topic in 2025 will be "AI agents." NVIDIA supports Blueprints and NIM microservices for building agent AI applications. Blueprints dramatically improves productivity by providing developers with tools to build and deploy AI agents that can analyze large amounts of video and image content.

As camera monitoring and environmental management, including AI CCTV and control centers, become increasingly widespread worldwide, the percentage of real-time human viewing is estimated to be less than 1%. To address this, AI agents can support early accident detection and prediction to supplement operational needs. Blueprint, accelerating this development, provides developers with tools to build AI agents capable of analyzing massive amounts of video and image content.

Additionally, the generative AI model system has been enhanced with the NVIDIA Cosmos Nemotron Vision Language Model, the NVIDIA Lamar Nemotron Giant Language Model, and the NVIDIA Nemo Retriever. This release supports developers' AI product development by distributing open AI models required for building enterprise-grade AI applications.

Jensen Huang predicted that robotics would be at the forefront of these AI advancements. He explained, "The era of robots is just around the corner, and robots will be able to learn behaviors generated through AI and omnibus to enhance their task performance." He predicted that agent AI would create trillions of dollars in future robotics opportunities.