Nvidia is also "cooking lobster"? The open-source platform NemoClaw is expected to be unveiled next week.

On March 9th, local time, according to media reports, sources revealed that Nvidia plans to launch an open-source AI agent platform called NemoClaw, designed to allow businesses to deploy AI agents within their workflows to replace employees in performing specific tasks.

Sources stated that regardless of whether these companies’ products run on Nvidia chips, they will be able to access the platform.

For a long time, Nvidia’s software strategy has heavily relied on its proprietary CUDA platform—this system deeply binds developers to Nvidia GPUs, forming a core competitive barrier for the company. This extension into an open-source platform indicates Nvidia is seeking to attract a broader range of enterprise customers through its software ecosystem, rather than relying solely on hardware lock-in.

Nvidia has already begun seeking partnerships with several major software companies for this product, targeting clients including Salesforce, Cisco, Google, Adobe, and CrowdStrike, and plans to officially unveil the product around next week’s annual developer conference in San Jose.

Due to the open-source nature of the platform, reports suggest that partners may contribute code in exchange for early free access. Nvidia also plans to incorporate security and privacy tools into the platform.

The launch of NemoClaw is set against the backdrop of the rise of “Claw-type” intelligent tools, which run locally on user devices in open-source form, capable of autonomously executing continuous tasks and even possessing self-learning abilities.

The earliest and most popular among these is OpenClaw.

At the Morgan Stanley Technology Conference on March 6th, Jensen Huang highly praised OpenClaw, calling it “the most important software release of this era.”

He emphasized that the Linux operating system, which took thirty years to permeate the infrastructure of global servers, smartphones, and supercomputers, has achieved widespread adoption, and OpenClaw surpassed this in just three weeks. On GitHub, OpenClaw’s stars quickly exceeded 250,000, surpassing even the Linux kernel and ending the long-standing dominance of the React frontend framework, becoming the most popular open-source foundational software project in history.

Huang believes that intelligent agents only need a series of prompts to perform tasks that previously required significant time and expertise. This has led to a roughly 1,000-fold increase in token consumption, creating a “computing power vacuum”: as AI agents continue to penetrate human work, existing hardware deployments—regardless of scale—will face computational limitations.

This week, OpenClaw’s popularity has continued to rise, evolving from a developer craze into a nationwide learning wave.

Major internet companies are actively involved, quickly localizing OpenClaw and packaging it into their own platform capabilities. ByteDance launched ArkClaw, Tencent Cloud introduced WorkBuddy, Alibaba developed CoPaw as a benchmark, and Xiaomi began testing MiClaw.

Meanwhile, several local governments have acted swiftly. After the “Lobster Ten Rules” in Longgang, Shenzhen, Wuxi High-tech Zone announced the “Lobster Twelve Rules,” both offering substantial subsidies.

Nvidia’s decision to step into this space at this time is not surprising. However, their products are clearly aimed at consumers, and the application of such intelligent agents in enterprise environments remains controversial.

Earlier media reports indicated that companies like Meta have instructed employees to avoid using OpenClaw on work devices, citing unpredictable agent behavior and potential security risks. Last month, an employee responsible for safety and alignment at Meta’s AI lab publicly described an “out-of-control” incident involving an AI agent, which resulted in mass deletion of emails.