AI’s new horizons: highlights from NVIDIA’s CES 2025 keynote

With announcements ranging from cutting-edge GPUs to advancements in physical AI, the keynote offered a glimpse into the future of computing and the role NVIDIA aims to play in shaping it.

Here are the key highlights and themes:

Blackwell architecture: a new standard for GPUs

NVIDIA unveiled the RTX Blackwell architecture, powering the next generation of GeForce RTX 50-series GPUs. With 92 billion transistors, 4,000 TOPS (trillion operations per second), and groundbreaking energy efficiency, these GPUs promise to redefine gaming and AI workloads.

Huang emphasised how the new architecture leverages Deep Learning Super Sampling (DLSS) to predict and generate frames using AI. The RTX 5070, delivering RTX 4090-level performance at $549, represents a significant leap in making high-performance GPUs more accessible.

Democratising AI with personal computing

A key part of NVIDIA’s strategy is bringing AI capabilities to everyday devices. By integrating NVIDIA’s AI stack with Windows PCs through WSL 2 (Windows Subsystem for Linux), Huang promised to transform personal computers into “world-class AI PCs.”

This development aligns with NVIDIA’s goal to “democratise AI,” enabling developers and users to run sophisticated AI models directly on their devices. Huang revealed plans for downloadable AI models tailored for PCs, making generative AI accessible to a broader audience.

Physical AI: the Cosmos platform

One of the most exciting announcements was NVIDIA Cosmos, the world’s first physical AI foundation model. Designed to understand physical dynamics such as gravity, friction, and object permanence, Cosmos aims to revolutionise robotics and industrial AI.

Huang described Cosmos as a platform to “train robotics and generate physically plausible scenarios of the future.” This foundation model, trained on 20 million hours of dynamic video, is already being used for synthetic data generation, robotics training, and industrial applications.

Autonomous vehicles: a trillion-dollar industry

Huang called autonomous vehicles the “first multi-trillion-dollar robotics industry” and announced key partnerships with leading automotive companies, including Toyota. NVIDIA’s Thor Robotics Processor, delivering 20x the processing power of its predecessor, was highlighted as a game-changer for AV development.

Huang also unveiled NVIDIA’s three-computer system for AVs: DGX for AI training, Omniverse for simulation and synthetic data generation, and AGX for deployment in vehicles. “We can take thousands of drives and turn them into billions of miles,” Huang said, underscoring the role of synthetic data in scaling AV development.

Agentic AI: transforming enterprises

NVIDIA’s enterprise AI ecosystem took centre stage with the introduction of NeMo and NEM microservices, designed to simplify AI deployment for businesses. These tools allow enterprises to create digital AI agents tailored to specific tasks, such as customer service, data analysis, or logistics management.

Huang shared: “In the future, IT departments will become the HR departments for digital agents,” predicting that AI agents will become integral to workplace productivity.

NVIDIA’s collaboration with Meta on the Llama 3.1 model was also announced, delivering fine-tuned AI models for enterprise use.

Robotics: the next frontier

NVIDIA sees robotics as the next transformative industry. The Isaac Groot platform, introduced during the keynote, provides tools for developing humanoid robots and generating massive synthetic motion datasets from a few human demonstrations. Huang described the potential for breakthroughs in general-purpose robotics as a “ChatGPT moment for robotics” just around the corner.

A vision for the future

Huang’s keynote was a celebration of NVIDIA’s 30-year journey, highlighting how AI has reshaped computing. From GPUs and gaming to robotics and enterprise AI, NVIDIA is positioning itself as a cornerstone of innovation. As Huang aptly concluded: “The age of AI is here, and it’s only the beginning.”

Stay tuned on Electronic Specifier for detailed reports on each of NVIDIA’s announcements at CES 2025.

Images courtesy of NVIDIA/YouTube.