Nvidia's Jensen Huang calls OpenClaw the "next ChatGPT" for the age of AI agents

Nvidia CEO Jensen Huang has compared OpenClaw to ChatGPT, effectively announcing the next phase of the AI market: from bots that answer to agents that act. For Nvidia, this is more than a catchy comparison. It is a bet on a new software layer on top of the AI computing boom.

Why this is an important signal

When a statement like this comes not from a startup or a venture investor but from the head of Nvidia, the market listens especially closely. The company has become one of the biggest beneficiaries of the AI boom, and Huang himself has turned into one of the industry's most influential voices in recent years.

In an interview with CNBC, he explained OpenClaw's popularity as a shift in user behavior: people are no longer satisfied with getting an answer from a model. They want to hand the system real actions, from searching for information to carrying out a chain of tasks without constant supervision.

"It's definitely the next ChatGPT."

What matters in this comparison is not a literal similarity between the products but the scale of the possible effect. ChatGPT made the very format of talking to a language model mainstream. OpenClaw and similar platforms aim for the next step: making work with autonomous agents mainstream, agents that don't just offer suggestions but open websites, write files, run commands and carry a task through to a result. If this model catches on, the AI market will shift again, from assistants to executors.

What OpenClaw can do

OpenClaw is an open-source platform for autonomous AI agents and, at the same time, a personal assistant that can run on the user's computer or in the cloud. Unlike an ordinary chatbot, it connects to large language models and gets access to tools: messengers, a browser, the file system, a terminal and external APIs. As a result, OpenClaw can not only respond to a request but actually do work in the background, keeping context and memory between sessions.

• Works through Telegram, WhatsApp, Slack and other familiar channels
• Reads and writes files, runs commands and scripts
• Autonomous background tasks and persistent workflows
• Local deployment with greater control over data
• Connects to different LLMs instead of being locked to a single model

Huang separately stressed the low barrier to entry: according to him, you can create your own agent with literally one line of code and then ask it to do the work you need. As an example, he described kitchen design: the agent first studies other people's solutions and tools, then proposes options and iteratively improves them. The idea is that the user states a goal, and the agent assembles the intermediate steps itself and moves toward the result.

Why Nvidia is backing it

Nvidia's interest in OpenClaw quickly went from compliments to products. At GTC 2026, the company introduced NemoClaw and Agent Toolkit, tools that make it easier to launch agent systems and add enterprise mechanisms for security, privacy and control. NemoClaw deploys OpenClaw together with Nvidia Nemotron models and the OpenShell environment with a single command. In essence, Nvidia is offering not a competitor but an infrastructure layer on top: take a virally growing open-source project and make it production-ready on RTX PCs, DGX workstations and server clusters.

For Nvidia, this is a logical move. The more popular autonomous agents become, the higher the demand not only for models but also for compute, local GPUs, enterprise guardrails and orchestration tools. OpenClaw generates interest among developers and enthusiasts, while Nvidia is trying to become the supplier of the full stack, from hardware and runtime to models and secure deployment. That is why Huang's words about the "next ChatGPT" sound less like an outside observation and more like strategic positioning for an entire market.

What this means

The AI market is moving ever faster from "ask the model" mode to "assign the system a task" mode. If OpenClaw and similar projects keep up their pace, the main competition will no longer be the quality of a single answer but an agent's ability to act reliably, securely and without unnecessary human involvement. For business, this is a signal to look not only at new models but also at the tools that turn them into real operational capacity.

*Meta has been designated an extremist organization and banned in Russia.