Nvidia positions OpenClaw as the foundation of agentic AI, signalling a shift to open, multi-agent systems and Agent-as-a-Service.
Nvidia is placing open source at the centre of its AI strategy with OpenClaw, positioning it as the new application layer for self-evolving, multi-agent systems. Designed to orchestrate autonomous “claws” that create sub-agents, execute workflows, and collaborate.
continuously, OpenClaw signals a structural shift beyond traditional large language models.
Karri Briski, VP of Generative AI Software for Enterprise at Nvidia, said: “OpenClaw is likely the single most important software release in history.” She added, “We used to prompt with ‘what’, ‘how’ or ‘why’, but for claws, we now prompt with ‘build’, ‘create’ or ‘make’. Claws are the new application layer for AI.”
Nvidia is reinforcing this with a full open-source stack, including the Nvidia Agent Toolkit and NemoClaw, which enables one-command deployment of OpenClaw, NemoTron AI models, and OpenShell. OpenShell adds sandboxing, policy enforcement, and privacy guardrails, while NemoTron 3 Super leads the PinchBench rankings, strengthening open-model credibility.
CEO Jensen Huang framed OpenClaw as a defining shift: “Every company in the world today needs to have an OpenClaw strategy, just as they did for Linux and Kubernetes,” signalling a move from SaaS to Agent-as-a-Service.
Underpinning this is the Vera Rubin platform, integrating GPUs with Groq LPUs to deliver over 500 tokens per second and up to 35× higher throughput via CUDA and Dynamo. This aligns with Nvidia’s “agentic scaling” push, where inference, reasoning, and continuous collaboration redefine AI workloads.
With expanded infrastructure and industry partnerships, Nvidia is positioning itself not just as a chipmaker, but as the architect of an open AI ecosystem.
