NVIDIA announced JetPack 7.2 and NemoClaw support for Jetson edge devices at COMPUTEX on June 4, 2026, bringing agentic AI capabilities to robotics and industrial automation. The release includes a 20% performance boost through Super Mode, Multi-Instance GPU support for isolated workloads, and one-command deployment of the NemoClaw agentic AI stack. Early adopters include Cadence, Dassault Systèmes, Siemens, and Synopsys for autonomous AI engineering workflows.
JetPack 7.2 Delivers 20% Performance Increase with Super Mode
JetPack 7.2 introduces substantial performance improvements for edge AI workloads on Jetson devices:
- Super Mode increases AGX Orin 32GB from 200 to 241 TOPS by raising GPU frequencies from 930 MHz to 1.3 GHz
- Multi-Instance GPU (MIG) support on Jetson Thor creates two isolated GPU instances: 12 SMs for inference and graphics, 8 SMs for robotics control
- CUDA 13 support on Jetson Orin devices
- Yocto project integration for custom embedded Linux distributions
The performance gains enable more complex AI models to run locally on edge devices without cloud connectivity, addressing latency and privacy requirements for industrial applications.
NemoClaw Enables One-Command Deployment of Agentic AI
NVIDIA JetPack 7.2 supports one-command deployment of NemoClaw, an open source stack built on the OpenClaw agent platform. The NemoClaw stack includes:
- NVIDIA Nemotron models for local inference
- OpenShell secure runtime for privacy and policy controls
- CUDA-X libraries for domain-specific agent skills
- NemoClaw blueprints for agent orchestration
Security features include network and filesystem isolation managed by OpenShell, real-time policy approval for external access, and full local inference to ensure data never leaves the device during agent operation. Developers can deploy the entire stack through a single curl command in JetPack 7.2's preconfigured environment.
Enterprise Adoption Spans AI Engineering and Cybersecurity
Major technology companies are already integrating NemoClaw into production workflows:
- Cadence, Dassault Systèmes, Siemens, and Synopsys use NemoClaw for autonomous AI engineering in simulation and verification
- Microsoft is collaborating with NVIDIA to offer a native Windows experience for personal agents
- Canonical and Red Hat are integrating the OpenShell runtime into their enterprise platforms
- CrowdStrike and Palantir employ Nemotron models for cybersecurity and operational decision-making
The new Nemotron 3 Ultra model launches on June 4 with 550 billion parameters, delivering five times faster inference with a 30% reduction in cost compared to previous generations.
Availability and Production Readiness
NemoClaw is available now for production deployment, while OpenShell is currently in early preview. According to NVIDIA, "With NVIDIA JetPack 7.2 and NVIDIA NemoClaw support, Jetson is agentic-ready, giving developers a production-grade stack to deliver next-level intelligence to robotics, inspection and industrial automation."
The release positions Jetson as a complete platform for deploying agentic AI at the edge, with security controls and performance capabilities suitable for enterprise production environments.
Key Takeaways
- NVIDIA released JetPack 7.2 with NemoClaw support at COMPUTEX on June 4, 2026, enabling agentic AI on Jetson edge devices
- Super Mode increases AGX Orin 32GB performance from 200 to 241 TOPS (20% boost) by raising GPU frequencies to 1.3 GHz
- NemoClaw deploys via a single curl command and includes Nemotron models, OpenShell secure runtime, and CUDA-X libraries for agent orchestration
- Enterprise adopters include Cadence, Dassault Systèmes, Siemens, Synopsys, Microsoft, CrowdStrike, and Palantir for applications ranging from AI engineering to cybersecurity
- The new Nemotron 3 Ultra model launches June 4 with 550 billion parameters, delivering 5x faster inference and 30% cost reduction