Nvidia Cosmos Reason 2: Leading the Global Transition to Physical AI
Nvidia unveils Cosmos Reason 2 and new Nemotron models at CES 2026, marking a pivotal shift toward Physical AI and generalist robots with 10x faster speech processing.
Can robots truly reason? Nvidia thinks so, and they’ve just released the brain to prove it. At CES 2026, the tech giant signaled a massive shift towards 'Physical AI,' where AI models move beyond chat interfaces and directly drive intelligent actions in the real world. CEO Jensen Huang's vision of embodied intelligence is no longer a future concept—it's arriving in the form of a sophisticated software-hardware ecosystem.
Nvidia Cosmos Reason 2 Physical AI: The Rise of Generalist Robots
The centerpiece of the announcement is Cosmos Reason 2, the latest vision-language model (VLM) designed for embodied reasoning. Building on the success of Cosmos Reason 1, which currently leads physical reasoning leaderboards, the new version allows physical agents to plan their next moves with the same fluidity that digital agents navigate workflows. This flexibility is crucial for enterprises looking to deploy robots in unpredictable human environments.
"Robotics is at an inflection point," stated Kari Briski, Nvidia vice president for generative AI software. She noted that the industry is moving from specialist robots limited to single tasks to "generalist specialist systems." These new robots combine broad foundational knowledge with deep proficiency in complex, task-specific skills, making them far more versatile than their predecessors.
Expanding the Nemotron Family for Agentic AI
Beyond physical movement, Nvidia is accelerating how AI agents communicate and stay safe. The new Nemotron Speech model is reportedly 10 times faster than existing alternatives, enabling real-time low-latency speech recognition for live applications. Alongside this, Nemotron RAG enhances multimodal data insights, while Nemotron Safety ensures that sensitive personally identifiable data isn't accidentally leaked during autonomous operations.
Authors
Related Articles
Nvidia shipped roughly a billion RISC-V cores in 2024, then announced it would run CUDA on the open standard. We break down how royalty-free instruction sets and open software stacks are trying to route around CUDA's lock-in. Part 2 of the Semiconductor Sovereignty series.
US AI-chip export controls split into three layers in the first half of 2026 — January easing, a May crackdown on circumvention, and a pending bill. Nvidia erased China from its guidance and still posted a record $81.6 billion quarter. A look at the export policy that both shields and cages it.
AMD's MI325X matches or beats Nvidia on memory and bandwidth — yet Nvidia's 86-92% share holds. The real moat is CUDA, 20 years in the making. Part 1 of 4.
Snowflake's new $6 billion AWS contract is about more than cloud spending. It signals a shift in AI infrastructure—away from Nvidia GPUs and toward cheaper, homegrown chips for the agent era.
Nvidia shipped roughly a billion RISC-V cores in 2024, then announced it would run CUDA on the open standard. We break down how royalty-free instruction sets and open software stacks are trying to route around CUDA's lock-in. Part 2 of the Semiconductor Sovereignty series.
US AI-chip export controls split into three layers in the first half of 2026 — January easing, a May crackdown on circumvention, and a pending bill. Nvidia erased China from its guidance and still posted a record $81.6 billion quarter. A look at the export policy that both shields and cages it.
AMD's MI325X matches or beats Nvidia on memory and bandwidth — yet Nvidia's 86-92% share holds. The real moat is CUDA, 20 years in the making. Part 1 of 4.
Snowflake's new $6 billion AWS contract is about more than cloud spending. It signals a shift in AI infrastructure—away from Nvidia GPUs and toward cheaper, homegrown chips for the agent era.
Thoughts
Share your thoughts on this article
Sign in to join the conversation