Cigarette Pack-Sized AI Powerhouses and More: NVIDIA's Bold Leap at CES 2025

Speaking at CES on 6 January 2025, Jensen Huang, CEO of NVIDIA, delivered the keynote address from the digitally recreated NVIDIA headquarters, painting a fascinating picture of how artificial intelligence is making impact across industries and everyday life.

AI in Your Pocket: Introducing NVIDIA Project DIGITS

One of the most intriguing announcements was NVIDIA's Project DIGITS, a new AI-powered personal supercomputer. Huang was impressed by its compact design, similar to the 'size of a pack of cigarettes', but capable of delivering a petaflop of computing power. This innovation aims to democratize AI supercomputing by making it accessible to small labs, lone professionals, and even home users (NVIDIA Project Digits).

Consider a scenario where researchers have access to significant computational resources directly from their workstations. This eliminates dependencies on cloud computing or traditional supercomputers, which can be limited by availability and cost. In this setting, a scientist working in a smaller lab could conduct their research with greater autonomy, utilizing a powerful, localized supercomputer. This accessibility to high-performance computing would enable individual scientists and smaller teams within startups to engage in AI research more readily. This shift could lead to increased research and development activities, as barriers to computational resources are reduced. Data analysis and model training could be performed more quickly and iteratively. Consequently, this future could see more individuals and organizations participating in scientific and technological AI advancements.

From Drug Discovery to Climate Solutions: The Power of AI Tokens

"This is how intelligence is created,” Huang said, introducing the new era of AI development. He explained how AI tokens, the building blocks of intelligent systems, open up endless possibilities. Imagine a future in which AI-assisted drug discovery accelerates the development of life-saving medicines, climate change is mitigated through energy optimization and renewable energy solutions, and personalized education is made available to all as AI tutors adapt to individual learning styles.

The RTX 50 Series Unveiled

Moving on to hardware, Huang introduced the RTX 50 series, a new powerful GPU designed to accelerate AI workloads. With 92 billion transistors and offering 4,000 TOPS (terabytes of operations per second) for AI, it is three times more powerful than its predecessor, the Ada generation. Huang emphasized its usefulness in both gaming and professional applications: “This RTX 50 Series, Blackwell architecture... the GPU is just a beast,” he said, highlighting its role in real-time computer graphics and AI processing.

Driving the Future: NVIDIA's Three-Computer Model for Autonomous Vehicles

Huang made important announcements in the area of automotive AI, highlighting NVIDIA's commitment to the automotive industry. He stated: "The autonomous vehicle revolution has arrived. After so many years, with Waymo's success and Tesla's success, it is very, very clear autonomous vehicles have finally arrived.' He predicted that the autonomous vehicle industry is likely to become the first multi-trillion-dollar industry.

Huang presented NVIDIA's approach to automotive AI with a 'three-computer' model required to create autonomous vehicles:

1. DGX for training AI models – where AI is trained on huge data sets.
2. Omniverse for modelling and generating synthetic data – allows you to create digital twins and test scenarios that simulate real driving conditions.
3. DRIVE AGX, an automotive supercomputer, is designed to deploy AI models for real-time decision-making in cars.

Thor: The Next-Gen Automotive Processor

A next-generation processor for cars called Thor was announced at CES 2025, which is already in mass production and offers 20 times the processing power of the previous generation Orin. Not only is Thor designed for autonomous vehicles, but it is also a general-purpose robotics computer suitable for a variety of applications, from AMR (autonomous mobile robots) to humanoids.

Safety First: NVIDIA Drive OS Meets Highest Safety Standards

Huang also highlighted the certification of NVIDIA Drive OS to the highest automotive safety standard, ISO 26262 ASIL-D, which is a significant milestone in automotive safety and AI integration, proudly stating:

Drive OS is now the first software-defined programmable AI computer that has been certified up to ASIL D, which is the highest standard of functional safety for automobiles.

Partnerships with major automakers such as Toyota, Mercedes, BYD and others were highlighted, demonstrating NVIDIA's integral role in the automotive sector. Toyota was highlighted as a new partner in developing next-generation autonomous vehicles.

NVIDIA Cosmos: Understanding the Physical World

He also introduced NVIDIA Cosmos, a fundamental model designed to understand the physical world. The model, trained on 20 million hours of video of dynamic physical actions, aims to revolutionize robotics and industrial AI by providing a framework for understanding physical dynamics, with applications ranging from robotics to autonomous vehicles (NVIDIA at CES 2025).

Democratizing AI with GeForce

Huang emphasized the democratization of AI with NVIDIA hardware, making it accessible to gamers, creators, and enterprises. He said: 'GeForce enabled AI to reach the masses, and now AI is coming home to GeForce,' highlighting NVIDIA's role in the spread of AI.

The presentation delved into the evolution of AI from perception to generative capabilities, which are now moving towards 'agent AI' and 'physical AI'. Huang discussed the three laws of scaling that guide the development of AI—scaling before learning, after learning, and during testing—and how they help develop AI capabilities. For example, the law of scaling after learning uses techniques such as reinforcement learning and human feedback to improve AI models.

Robotics Reimagined: NVIDIA's Isaac Groot Platform

In addition, he introduced NVIDIA's Isaac Groot platform for developing general-purpose robot models, highlighting its use for creating digital twins for robots. “The ChatGPT moment for general robotics is just around the corner," Huang predicted, emphasizing the role of modeling and synthetic data in robotics (NVIDIA Isaac Groot).

In conclusion, NVIDIA’s groundbreaking innovations unveiled at CES 2025 mark a bold step toward an AI-driven future, from cigarette pack-sized supercomputers to the evolution of autonomous vehicles. Whether you’re a developer, researcher, or tech enthusiast, these advancements open up new frontiers in computing, robotics, and beyond.

Want to stay ahead of the curve? Join the conversation in the comments to discuss how NVIDIA’s latest technologies can transform your industry or project!