NVIDIA CEO Jensen Huang took the stage at CES 2025 to deliver a keynote that was as bold as the company’s vision for the future. Fresh off record-breaking highs in NVIDIA’s stock performance, Huang showcased how the company’s relentless focus on AI and accelerated computing is reshaping industries, from automotive and robotics to gaming and beyond.
“Our mission is to advance AI and computing to new heights, empowering industries to solve the unsolvable,” said Huang during the keynote. His remarks set the tone for an event that not only highlighted NVIDIA’s technological leadership but also underscored its role in driving the AI revolution.
A New Era of Computing: Physical AI Takes the Spotlight
Huang introduced the concept of Physical AI, marking a significant leap forward in how machines interact with the physical world. Unlike generative AI, which creates text or visuals, Physical AI enables robots and autonomous vehicles to perceive their surroundings and make decisions based on sensory input and intent.
“Physical AI will transform the backbone of the global economy,” Huang said, referring to its potential to automate industries like manufacturing, logistics, and transportation. With 10 million factories, 200,000 warehouses, and 1.5 billion vehicles worldwide, the scope of this transformation is staggering.
To achieve this, NVIDIA relies on what Huang called the AI Factory, a three-pronged system consisting of:
- DGX for training AI models,
- Omniverse for testing and generating synthetic data, and
- AGX Thor for running these models in real-world environments.
Huang emphasized the importance of simulation in this process, noting that “real-world testing is expensive, slow, and sometimes impossible.” Synthetic data and virtual environments, powered by NVIDIA Omniverse, make it feasible to train robots and autonomous systems on scenarios they might rarely encounter in real life.
Introducing NVIDIA Cosmos: A Platform for Physical AI
One of the most anticipated announcements was NVIDIA Cosmos, a world foundation model development platform aimed at accelerating Physical AI. Cosmos allows developers to train intelligent robots and autonomous vehicles faster, more safely, and more cost-effectively than ever before.
“Cosmos is game-changing,” Huang said. “It bridges the gap between simulation and reality, enabling developers to generate exponentially larger datasets and achieve breakthroughs in robotics and self-driving cars.”
The platform includes advanced diffusion models for creating synthetic data, cutting-edge video tokenizers for compressing training data, and integration with NVIDIA’s CUDA for lightning-fast video processing. According to Huang, Cosmos is poised to do for robotics what large language models like ChatGPT have done for generative AI.
Key Features of Cosmos
- Foundation Models: Cosmos uses advanced diffusion and autoregressive models to create synthetic data and predict real-world scenarios.
- Video Tokenization: Achieving compression and performance levels unmatched by other platforms, Cosmos enables rapid processing of billions of hours of video data.
- Seamless Integration with Omniverse: Developers can use Omniverse to create 3D simulations and then enhance them with Cosmos, bridging the gap between simulation and reality.
With Cosmos, NVIDIA aims to revolutionize training pipelines for robotics and autonomous vehicles, turning months of effort into mere days.
Omniverse: Building the Digital Twin Universe
NVIDIA also unveiled significant updates to Omniverse, its platform for creating 3D digital twins and virtual worlds. With the addition of tools like Isaac Groot for synthetic motion generation and Mega for simulating industrial robot fleets, Omniverse is helping developers scale their workflows while reducing costs.
Huang likened Omniverse to a “multiverse playground,” where robots and autonomous vehicles can simulate countless future paths, much like a scene from Doctor Strange. This capability is critical for industries looking to optimize everything from warehouse logistics to autonomous vehicle navigation.
Highlights from the Keynote
- Isaac Groot: A blueprint for scaling humanoid robot training. By leveraging digital twins, developers can create thousands of synthetic motion trajectories—no physical robots needed.
- Mega: A tool for simulating robot fleets in industrial environments, optimizing workflows in warehouses and factories.
- Big-Name Partnerships: Siemens, Cadence, and Fertelix are already integrating Omniverse into their operations, ranging from industrial design to autonomous vehicle simulation.
Automotive: A $5 Billion Cloud-to-Car Opportunity
The keynote also highlighted NVIDIA’s strides in the automotive sector. Huang announced that Toyota, the world’s largest automaker, is building its next-generation vehicles on NVIDIA’s Drive AGX platform, running the newly certified DriveOS.
“This is not just about cars—it’s about creating an entire ecosystem where the cloud and the car are seamlessly connected,” said Huang.
Key Innovations
- Drive AGX Thor: NVIDIA’s next-generation supercomputer for autonomous vehicles, built on the Blackwell architecture. Early adopters include Toyota, BYD, and Volvo.
- DriveOS: A secure, ASIL-D-certified operating system for self-driving vehicles. Mercedes-Benz and Toyota are among the first to integrate it into their vehicles.
- Cosmos for Automotive: By generating rare, long-tail scenarios, Cosmos equips automakers with the data they need to train vehicles for even the most unexpected situations.
He revealed that NVIDIA’s automotive business is expected to grow into a $5 billion operation by 2026, driven by partnerships with companies like Volvo, BYD, and Mercedes-Benz. With DriveOS now certified to the industry’s highest safety standard (ASIL-D), NVIDIA is setting a new benchmark for secure, software-defined vehicles.
What This Means for the Future and Why It Matters
NVIDIA’s keynote wasn’t just about showcasing cutting-edge technology—it was a roadmap for the future of AI. From billions of autonomous cars to humanoid robots transforming industries, NVIDIA is leading the charge.
The announcements at CES underscored NVIDIA’s position as a leader in the AI-driven future. From enabling humanoid robots to automating warehouses and revolutionizing automotive safety, the company is expanding the boundaries of what’s possible.
- Industries can now solve previously insurmountable challenges using synthetic data and advanced simulations.
- The collaboration between tools like Omniverse and Cosmos is unlocking new efficiencies in AI training and deployment.
- Autonomous systems are inching closer to mainstream adoption, promising safer, smarter, and more sustainable solutions.
Huang closed the keynote with a vision of a world reshaped by AI: “We are at the dawn of an incredible era where every industry will be redefined by accelerated computing and AI. This is just the beginning.”
With tools like Cosmos and Omniverse and partnerships that span the globe, NVIDIA isn’t just leading the AI revolution, it’s defining it and FS Studio is proud to be a partner with NVIDIA delivering innovative digital solutions using NVIDIA tools for our customers.
For those who missed it, the full keynote is available on demand and in this post, showcasing NVIDIA’s bold vision for a smarter, more connected world.