NVIDIA CES 2025

Key Takeaways:

• RTX 50 Series: Launching January 2025, the RTX 5090 and 5080 GPUs deliver up to 8x performance with DLSS 4 and Neural Shaders.
• Project DIGITS: A $3,000 AI supercomputer for developers, launching May 2025.
• Cosmos Platform: Powers robotics and autonomous vehicles with scalable Nemotron AI models.
• AI PCs & Robotics: Generative AI for RTX PCs and tools accelerating automation in robotics.

Ushering in a New Era of AI and Computing

At CES 2025, NVDA (NVIDIA) CEO Jensen Huang unveiled innovations poised to redefine industries ranging from gaming to robotics and autonomous vehicles. During the 90-minute keynote, Huang reflected on the company’s three-decade journey and shared bold projections for the future of AI and computing.

Huang highlighted how AI has rapidly evolved—from perception AI to generative AI and now to physical AI. “We are entering an era where AI can perceive, reason, plan, and act”, he stated, emphasizing NVIDIA’s pivotal role in this transformation across industries.

Redefining Graphics

At the heart of NVIDIA’s announcements was the new GeForce RTX 50 Series GPUs, powered by the Blackwell architecture. The GeForce RTX 5090, with its staggering 92 billion transistors and 3,352 trillion AI operations per second (TOPS), is NVIDIA’s most powerful GPU yet. Key features of the GeForce RTX 50 Series include:

• DLSS 4 with Multi Frame Generation: Enables up to an 8x performance boost.
• RTX Neural Shaders: Enhances textures, lighting, and materials in real-time.
• RTX Neural Faces and RTX Hair: Uses generative AI for hyper-realistic digital characters.
• RTX Mega Geometry: Increases ray-traced triangles by up to 100x for unparalleled detail.
• NVIDIA Reflex 2: Reduces PC latency by up to 75%.

The RTX 5090 ($1,999) and RTX 5080 ($999) will launch on January 30, 2025, with the RTX 5070 Ti and RTX 5070 following in February. Laptop GPUs, featuring up to 24GB of GDDR7 memory, will roll out in March. Huang described the series as a milestone for gaming and AI, stating, “GeForce enabled AI to reach the masses, and now AI is coming home to GeForce”.

A Personal AI Supercomputer

NVIDIA unveiled Project DIGITS, a compact AI supercomputer powered by the GB10 Grace Blackwell Superchip. This system brings enterprise-level AI capabilities to individual developers, creators, and engineers.

• Compact Design: Small enough to fit on a desk but powerful enough to run the entire NVIDIA AI stack, including DGX Cloud.
• Powered by GB10: The GB10 features an NVIDIA Blackwell GPU with the latest-generation CUDA cores and fifth-generation Tensor Cores, connected via NVLink-C2C chip-to-chip interconnect to a high-performance NVIDIA Grace CPU. The CPU includes 20 power-efficient cores built with the Arm architecture.
• Availability: Expected to launch in May 2025 at a starting price of $3,000.

Huang emphasized its accessibility, stating, “Every software engineer, every creative artist—everyone who uses computers today as a tool—will need an AI supercomputer”.

Advancing Physical AI

Huang introduced NVIDIA Cosmos, a world foundation model platform designed for physical AI applications like robotics and autonomous vehicles. This platform integrates generative models, tokenizers, and a video data pipeline to help AI systems perceive, simulate, and act in real-world environments.

Cosmos enables simulations of infinite scenarios, equipping robots and autonomous vehicles with foresight to optimize decision-making. Huang likened it to the “ChatGPT moment for general robotics”, signaling its transformative potential. Adopters of Cosmos include industry leaders like Uber, Hyundai Motor Group, and XPENG, who are leveraging it to develop safer vehicles and advanced robotics. NVIDIA DRIVE Hyperion, built on Cosmos, uses synthetic data to scale autonomous vehicle training data from hundreds of drives to billions of effective miles.

New AI Model Families

To support AI deployments across scales, NVIDIA introduced the Llama Nemotron and Cosmos Nemotron model families, available in three sizes:

• Nano: Optimized for real-time applications with low latency, ideal for deployment on PCs and edge devices.
• Super: A high-accuracy model offering exceptional throughput on a single GPU.
• Ultra: The highest-accuracy model, designed for data-center-scale applications demanding the highest performance.

Innovations in Autonomous Vehicles

NVIDIA continues to lead in autonomous vehicle (AV) technology with the introduction of the DRIVE Hyperion platform, built on the NVIDIA AGX Thor system-on-a-chip (SoC). This platform integrates generative AI capabilities with advanced safety systems, marking a new era for AVs. Key highlights include:

• Synthetic data generation via Omniverse and Cosmos to accelerate AV training.
• Partnerships with Toyota, Mercedes-Benz, and Volvo for next-gen vehicle development.
• A full-stack solution combining NVIDIA DGX, Omniverse, and DRIVE AGX for AI model training, testing, and deployment.

“The autonomous vehicle revolution is here”, Huang said, projecting that advancements in AV technology will accelerate at an unprecedented pace in the coming years.

AI Foundation Models and NIM Microservices

NVIDIA introduced AI foundation models and NIM microservices for RTX PCs, enabling developers to build and deploy generative AI applications such as digital humans, podcasts, videos, and more. Key features include:

• FP4 Computing: Doubles AI inference performance and reduces memory requirements.
• AI Blueprints: Customizable workflows for tasks like customer support, fraud detection, and media management.
• NIM-Ready PCs: Top PC manufacturers are rolling out NIM-ready RTX AI PCs, making advanced AI tools accessible to a broader audience.

“AI PCs are coming to a home near you”, Huang stated, signaling a future where AI-enhanced computing becomes commonplace.

Robotics and Digital Manufacturing

NVIDIA is extending AI’s impact into robotics and manufacturing with tools like the Isaac GR00T Blueprint for synthetic motion generation. Using NVIDIA Omniverse, developers can generate vast synthetic datasets for training humanoid robots.

The Mega Blueprint enables large-scale simulations of robot fleets, streamlining warehouse automation, and adopted by leaders like Accenture and KION. Huang highlighted the importance of these tools in driving rapid breakthroughs in general robotics.

Future Projections: A Year of Breakthroughs

As Huang concluded his keynote, he emphasized that 2025 has been a year of monumental progress for NVIDIA. From the GeForce RTX 50 Series GPUs to Project DIGITS and Cosmos, these innovations represent just the beginning of a transformative era.

Huang predicted rapid advancements in physical AI, agentic AI, and autonomous systems, stating, “All of the enabling technologies I’ve been talking about are going to make it possible for us in the next several years to see very rapid breakthroughs, surprising breakthroughs, in general robotics”. NVIDIA’s vision for the future is clear: to lead the charge in AI-driven innovation across industries, revolutionizing how people live, work, and interact with technology.