Nvidia is on the verge of announcing its own Claw, according to Wired. Can you believe it? An Nvidia claw!
Which claw do you use for your agentic AI tasks? The lightweight Nanoclaw? The security-focused IronClaw? Oh, I can tell from your fashion sense you’re retro, and you prefer the O.G.—OpenClaw. Sometimes I get nostalgic for six weeks ago too. Simpler times!
If you don’t know what the hell I’m talking about, you’re not a depraved AI freak, which is fortunate for you. Please be aware, however, that this whole “claw” trend is moving very quickly. Nvidia’s position as the premier developer of AI chip architectures, and of CUDA, the underlying proprietary software platform behind much of the AI world, could mean Nvidia is looking to set standards for an important new tech category by getting into the claw game.
Bought a new Mac mini to properly tinker with claws over the weekend. The apple store person told me they are selling like hotcakes and everyone is confused 🙂
I’m definitely a bit sus’d to run OpenClaw specifically – giving my private data/keys to 400K lines of vibe coded…
— Andrej Karpathy (@karpathy) February 20, 2026
Claws, a hardware and software trend that started with the release of OpenClaw (formerly known as Clawdbot and Moltbot) last year, are normally wrappers for LLMs like Claude Code and OpenAI’s Codex that ostensibly function as personal assistants that can perform tasks that involve writing code and browsing the internet. Users typically set up a dedicated computer to run a claw, plug an expensive LLM subscription into it, give it access to their personal data and accounts, and then communicate with it over a messaging app like WhatsApp (Claws are also, famously, a security nightmare).
The creator of OpenClaw, Austrian software engineer and former entrepreneur Peter Steinberger, was hired by OpenAI last month. OpenAI CEO Sam Altman wrote that his mission at OpenAI is “to drive the next generation of personal agents,” and that he expects that what Steinberger creates there will “quickly become core to our product offerings.”
According to Wired, whose reporting on Nvidia comes from anonymous leaks—or “people familiar with the company’s plans,” to use Wired’s phrase—Nvidia has been approaching enterprise software companies to discuss it’s claw platform, which is named (for now?) NemoClaw. Enterprise software platforms have been subjected to an all-out stock price assault from investors lately, whose market behavior suggests they believe the software-as-a-service (SaaS) model is overvalued due to impending automation made possible by tools like OpenClaw.
Nvidia is apparently allowing these companies free early access to NemoClaw, meaning they can source actual work tasks from claw-style AI agents, whether their systems run on Nvidia chips or not, in exchange for contributions to Nvidia’s claw project. Google, Adobe, Salesforce, Cisco, and CrowdStrike, are the potential partner companies mentioned by Wired, although they have been silent so far on whether or not they are partnering with Nvidia.
NemoClaw is reportedly open-source, and its moniker suggests that it’s meant to be powered by the Nemotron family of open-source models, like Nemotron 3 which was announced last year. The press release for Nemotrom 3 says these models are “designed to power transparent, efficient and specialized agentic AI development across industries.”
According to Wired’s reporting, NemoClaw will be announced at Nvidia’s GTC developer conference next week. That would mean, if the Wall Street Journal’s reporting is accurate, that it will coincide with the release of a new inference chip Nvidia plans to release as well.
