The Robotics Revolution: NVIDIA Unveils Physical AI and the Jetson T4000 at CES 2026

At CES 2026 this year, NVIDIA shifted its focus decisively from the digital realm of generative AI toward “Physical AI”—systems designed to interact directly with the real world. NVIDIA CEO Jensen Huang declared that the robotics industry has reached its own “ChatGPT moment,” unveiling in one sweeping announcement a complete ecosystem spanning foundation models, simulation software, and edge-computing hardware. The ambition is clear: to transform robots from single-purpose automation tools into intelligent physical agents endowed with perception, reasoning, and general-purpose capabilities.

To accelerate development, NVIDIA announced the release of a series of open-source foundation models on Hugging Face, enabling developers to bypass the costly pretraining phase and build robots with advanced reasoning abilities from the outset.

On the software and model front, NVIDIA introduced the all-new NVIDIA Cosmos platform, an infrastructure purpose-built for Physical AI.

• Cosmos Reason 2: A vision-language model (VLM) with open reasoning capabilities, allowing robots to “understand” the physical world much like humans do, with highly precise interaction awareness.
• Isaac GR00T N1.6: A vision-language-action (VLA) model designed for humanoid robots. By integrating the reasoning capabilities of Cosmos Reason, it enables full-body coordinated control and contextual understanding, resulting in far more fluid and natural movement.
• Alpamayo: A family of models tailored for autonomous driving, combining reasoning-capable VLA models with large-scale driving datasets.

In parallel, NVIDIA updated its Nemotron model family—enhancing speech recognition and security for agentic AI use cases—as well as its Clara models, which focus on protein design and drug synthesis planning in the healthcare domain. To address fragmentation in robotics development workflows, NVIDIA unveiled the open-source Isaac Lab-Arena framework, a system for evaluating robotic strategies at scale within simulation environments, integrated with benchmarks such as Libero and Robocasa.

At the orchestration level, NVIDIA formally launched OSMO, a cloud-native framework that unifies workflows ranging from synthetic data generation to model training. OSMO allows developers to seamlessly orchestrate compute resources across local workstations and the cloud, managing complex pipelines with command-like precision.

In a particularly significant move, NVIDIA announced the integration of Isaac and GR00T technologies into Hugging Face’s LeRobot framework, further strengthening its commitment to the open-source robotics community. On the hardware side, NVIDIA finally brought its latest Blackwell architecture to the edge.

The newly introduced Jetson T4000 module delivers an astonishing 1,200 FP4 TFLOPS of compute performance within a 70-watt power envelope, paired with 64 GB of memory. Offering four times the performance of its predecessor and priced at USD 1,999 (per 1,000-unit volume), it is positioned as a core engine for next-generation high-end autonomous machines.

For humanoid robots with even greater computational demands, the Jetson Thor system has emerged as the industry’s platform of choice. Companies including Boston Dynamics, 1X, NEURA Robotics, and LG—following the recent announcement of its home robot—have already adopted Jetson Thor to enhance navigation and manipulation capabilities. NVIDIA’s Physical AI technologies are now rapidly permeating multiple industries:

• Industrial: Caterpillar is leveraging NVIDIA technology to automate construction and mining operations.
• Healthcare: LEM Surgical and XRlabs are using Isaac and Jetson AGX Thor to train surgical robots and deliver real-time AI-driven analysis.
• Data Analytics: Salesforce employs Cosmos Reason models to analyze robot-captured video, dramatically reducing incident response times.

Together, these developments underscore NVIDIA’s intent to redefine robotics as a foundational pillar of the next AI era—where intelligence is no longer confined to screens, but embodied in machines that act, perceive, and reason in the physical world.

Related Posts:

• Nvidia Flaws Expose Jetson AI & Robotics Platforms to RCE and Data Theft
• Leaked Documents Reveal NVIDIA’s Secret AI Training Practices
• CVE-2024-0112: NVIDIA Patches High-Severity Vulnerability in Jetson and IGX Orin Platforms
• NVIDIA Critical AI Patch: Isaac Lab and NeMo Framework Flaws Risk Full Code Execution