Jensen Huang Believes Every Software Company Needs an OpenClaw Strategy
By Siddharth Jindal ·
At GTC 2026, the premier AI conference hosted by NVIDIA, CEO Jensen Huang made it clear that the future of software is no longer apps but agents.
At the centre of this shift is OpenClaw, an open-source project created by Peter Steinberger earlier this year, which Huang described in striking terms.
“OpenClaw is number one. It’s the most popular open source project in the history of humanity, and it did so in just a few weeks. It exceeded what Linux did in 30 years,” Huang said.
OpenClaw crossed more than 250,000 stars on GitHub in early March 2026, surpassing projects such as the Linux kernel and growing rapidly from around 200,000 stars in mid-February.
OpenClaw, as Huang explains, is an agentic system that connects to large language models, tools, and file systems, while handling scheduling, workflows, and communication. It can break down prompts into steps and even create sub-agents to complete tasks.
“It could decompose a problem… step by step by step. It could spawn off and call upon other sub-agents,” Huang said.
OpenClaw (formerly Clawdbot and Moltbot) became an internet sensation earlier this year for its ability to run locally on a user’s own hardware. It was envisioned as a long-running AI assistant capable of taking actions independently. Users connect it to an AI model provider, such as OpenAI, Anthropic, Google, or a LLM, and enable different connectors, including APIs, browser automation, email and calendar access, or command-line tools.
OpenClaw is also the backbone of Moltbook—a social network designed exclusively for AI agents.
It has also seen rapid adoption in China, with SimilarWeb reporting a nearly 12% share in site visits. In the country, it is being widely used for everyday tasks like writing reports, managing emails, and booking travel.
The platform has gone viral among developers and businesses, with companies such as Alibaba, Tencent, and ByteDance integrating it into their products.
Some of the firms have even launched their own versions of OpenClaw, such as KimiClaw, AutoClaw, QClaw, and MaxClaw.
The OS of Agentic Computing
In Huang’s view, OpenClaw is not merely a framework but the foundation of a new computing layer.
“It’s an operating system,” he said. “OpenClaw has open-sourced essentially the operating system of agentic computers.”
The comparison is intentional. Just as Windows ushered in personal computing, Huang believes OpenClaw will shape the era of personal agents.
He also compared OpenClaw to major technologies like Linux, HTML, and Kubernetes that changed how the software industry works. “Every software company needs to have an OpenClaw strategy.”
Enter NemoClaw
OpenClaw quickly got popular for making it easy to build AI agents, but they can access sensitive information, run code, and transmit data externally, making them hard to control in enterprise settings.
To tackle this, NVIDIA introduced NemoClaw. It allows users to install Nemotron models and the OpenShell runtime in a single command, with built-in privacy and security controls for autonomous AI agents.
NVIDIA Nemotron family of open-weight models is designed for fast, accurate coding, math, and agentic workloads.
The company said the stack can support self-evolving AI agents, referred to as ‘claws’, by adding policy-based controls, sandboxing, and network boundaries to improve trust, scalability, and accessibility.
NemoClaw integrates NVIDIA’s Agent Toolkit software to deploy OpenClaw environments with a single command. It installs OpenShell, which provides an isolated runtime environment for running models and agents with privacy safeguards.
NVIDIA said this creates an infrastructure layer that allows agents to access tools and data while enforcing security and policy constraints.
Several users on social media projected that NemoClaw could make OpenClaw viable for enterprise deployment.
Jordy Coltman, Marketing Lead at Sport.Fun, who has been experimenting with AI agents for some time, said in a post on X, that, unlike OpenClaw, NemoClaw doesn’t insert a fear of permissions. “ My agent has access to my files, my browser, and my APIs. One bad instruction and it could wipe something important,” he wrote.
NemoClaw fixes that in one command with sandboxed execution, where “your agent does its job without access to anything it doesn’t need,” he added.
jensen huang just called openclaw "the most popular open source project in the history of humanity."
the CEO of a trillion dollar company just put openclaw above linux.
then nvidia dropped nemoclaw - their own secure runtime that installs openclaw in one command with sandboxed… pic.twitter.com/aBw1pP0Ab7
— jordy (@jordymaui) March 16, 2026
Similarly, Kevin Simback, COO of Delphi Labs, in a post on X, said that he is excited to try out NemoClaw.
“Having spent a good bit of my career in enterprise, I’ve been pretty vocal about OpenClaw not being enterprise-ready. But the concept of an agentic workforce is a killer, and enterprises are going to want it, so this may be what really kicks it off,” he wrote.
Nemotron and the Open Model Push
Alongside NemoClaw, NVIDIA is building its open model ecosystem, led by Nemotron.
“The world is diverse. No single model can serve every domain,” Huang quipped. Notably, the company recently announced Nemotron 3 Super, a 120 billion parameter model that can run agentic systems at scale.
“It is going to be the best base model the world’s ever created. This allows us to help every country build their sovereign AI,” Huang said.
To accelerate development, the company announced the Nemotron Coalition, bringing together AI companies globally, including India’s Sarvam AI. The coalition also includes Black Forest Labs, Cursor, LangChain, Mistral AI, Perplexity, Reflection AI, and Thinking Machines Lab as inaugural members.
The group will build open models that developers and organisations can adapt for specific industries and regions.
The first project will be a base model co-developed by Mistral AI and NVIDIA, trained on NVIDIA DGX Cloud and later open-sourced. It will serve as the foundation for the upcoming Nemotron 4 family of models.
Members will contribute data, evaluation frameworks, and domain expertise to support post-training and ongoing development, to create a shared model for different use cases.
“We want to create the foundation model so that all of you can fine-tune it and post-train it into exactly the intelligence you need,” Huang said.
Framing the shift as a broader transformation, Huang called it “a renaissance of the enterprise,” and added that the industry could become a multi-trillion-dollar market built around specialised AI agents.
He said companies may allocate token budgets to engineers to increase productivity. “Every engineer that has access to tokens will be more productive,” he said, adding that such access is becoming a factor in hiring.
Huang added that future enterprises will not just consume AI but also produce it. “Every single software company of the future will be agentic,” he said, describing them as both token users for internal teams and token manufacturers for customers.
If Huang’s vision holds, OpenClaw is not just another open source success but the foundation of agentic computing.