Three months ago, almost no one had heard of a lobster-themed AI project by an Austrian indie developer. This week, it dominated Nvidia's GTC — the biggest AI conference in the world.

OpenClaw, an open-source platform for building autonomous AI agents, took center stage at GTC 2026 in San Jose. Nvidia CEO Jensen Huang dedicated a major portion of his keynote to it, calling it "the most popular, open-source project in the history of humanity" and telling CNBC's Jim Cramer it is "definitely the next ChatGPT."

What OpenClaw Does

Unlike traditional chatbots that wait for prompts, OpenClaw agents connect tools — email, calendars, messaging apps, web browsers — and act autonomously. Users spin one up with a short shell command, then delegate tasks ranging from bidding on eBay auctions to iterating on design projects. "Every carpenter can now be an architect," Huang said in his keynote.

Nvidia's Response: NemoClaw

Recognizing both the opportunity and the risk, Nvidia announced NemoClaw at GTC — a free, enterprise-grade security and oversight layer built on top of OpenClaw. The goal is to make autonomous agents safe enough for large organizations to deploy at scale.

The Commoditization Concern

The success of a free, indie-built framework is fueling an uncomfortable question in the industry: are the large, expensive frontier models from OpenAI and Anthropic becoming commoditized? OpenClaw routes tasks across whatever LLM is cheapest or best-suited, treating the underlying models as interchangeable infrastructure.

That dynamic is drawing more scrutiny than celebration in some quarters — even as it marks a genuine milestone for open-source AI.