News
Mar 9, 2026
baCta Raises €7M to Build AI-Powered Molecular Factories for Sustainable Industrial Ingredients
Tech Updates
Enterprise
Artificial Intelligence
Americas
NewDecoded
3 min read

Image by Nvidia
NVIDIA announced the NemoClaw stack today at GTC, offering a single-command installation for the OpenClaw agent platform. The new software suite integrates NVIDIA Nemotron models and the OpenShell runtime to provide enterprise-grade privacy for autonomous AI assistants. This launch aims to make self-evolving agents, known as claws, more trustworthy and accessible for both home and office use. During his keynote, NVIDIA CEO Jensen Huang described OpenClaw as the operating system for personal AI. He noted that while traditional operating systems manage personal computers, OpenClaw manages the next generation of software: proactive agents. NemoClaw provides the necessary infrastructure to scale these agents across platforms ranging from RTX PCs to the newly announced DGX Spark supercomputer.
The primary innovation within NemoClaw is NVIDIA OpenShell, a secure runtime designed to solve OpenClaw's biggest security hurdles. Early versions of the open-source project often left host systems vulnerable because agents required deep file and network access to function effectively. OpenShell creates an isolated sandbox that enforces strict, policy-based guardrails on every action an agent takes. Privacy is further enhanced through a built-in privacy router that manages data flow between local and cloud models. Sensitive tasks are automatically routed to local NVIDIA Nemotron models running on the user's hardware. For more complex reasoning, the system can tap into cloud-based frontier models without compromising the defined security perimeter.
To support these always-on agents, NVIDIA introduced the DGX Spark AI supercomputer for just $3,999. Powered by the Grace Blackwell GB10 superchip, this desktop unit allows developers to run 200-billion-parameter models entirely on-premises. This localized approach ensures that corporate intellectual property and personal data never leave the owner's physical control. The collaboration with OpenClaw creator Peter Steinberger highlights a growing ecosystem focused on secure, autonomous computing. Developers attending GTC can participate in a build-a-claw event to deploy their own customized assistants using the new stack. This initiative marks a pivotal step in turning experimental AI tools into hardened software for the global workforce.
The launch of NemoClaw signifies NVIDIA's transition from a hardware provider to an orchestrator of the AI deployment layer. By securing OpenClaw, which previously suffered from significant privacy vulnerabilities, NVIDIA is standardizing the runtime for personal AI. This move effectively counters the risks of unmanaged autonomous agents while locking developers into the NVIDIA hardware ecosystem, specifically via the new DGX Spark and RTX platforms.
Related Articles