Announcement
Mar 18, 2026
NVIDIA Goes All-In on OpenClaw: Jensen Huang Declares the Agentic OS As Big as Linux
The Keynote That Just Rewrote the Rules of Enterprise Computing
There are product launches, and then there are moments that redraw the map of an entire industry. On stage this week, Jensen Huang — the CEO of a company now worth more than most countries' GDP — delivered one of the latter.
In a keynote that lasted nearly twenty minutes, Huang did not announce a new chip. He did not reveal a new data center. Instead, he did something far more consequential: he declared that an open-source project called OpenClaw is the most important infrastructure development since Linux, and that NVIDIA is going all-in to make it enterprise-ready.
"OpenClaw is the most popular open-source project in the history of humanity," Huang told the audience, his voice carrying the weight of someone who has spent decades watching technology cycles come and go. "It exceeded what Linux did in 30 years. And it is that important."
Let that sink in. Linux took three decades to achieve its level of adoption. OpenClaw did it in weeks.
Understanding What OpenClaw Actually Is
Before we get into what NVIDIA announced, it is worth understanding why Huang considers this so significant. OpenClaw is not a chatbot. It is not a copilot. It is not another wrapper around a language model.
OpenClaw is an operating system for AI agents.
Huang broke down its architecture with the precision of someone who has been thinking about operating systems for forty years:
Resource management — it manages computational resources, memory, and tool access
Tool integration — it connects to any tool, any API, any file system
LLM orchestration — it calls and connects to large language models from any provider
Scheduling and automation — cron jobs, task decomposition, step-by-step execution
Sub-agent spawning — it can decompose a problem and delegate parts to specialized sub-agents
Multi-modal I/O — "You could wave at it and it understands you. You could talk to any modality you want. It sends you messages. It texts you. Sends you email."
Then Huang made the connection explicit: "I have just used the same syntax that I would describe an operating system. OpenClaw has open-sourced essentially the operating system of agentic computers."
This is not metaphor. This is architectural equivalence. Just as Windows abstracted hardware complexity and gave users a graphical interface to their personal computer, OpenClaw abstracts AI model complexity and gives users — and businesses — an interface to their personal agents.
"It is no different than how Windows made it possible for us to create personal computers," Huang continued. "Now OpenClaw has made it possible for us to create personal agents."
The Stories That Made It Real
What elevated this keynote from technical announcement to genuinely compelling narrative were the stories. Not hypotheticals. Not demos. Real people, using OpenClaw right now.
Huang mentioned Peter Steinberger, the creator of OpenClaw, being present in the audience. He spoke about what the platform has already enabled:
"I really love what my stuff enables that person to do," Huang said, relaying a story about a 60-year-old father who installed OpenClaw. "He told me like he installed it as a 60-year-old dad and they made beer, connected the machine via Bluetooth to OpenClaw. And then we automated everything including the whole website for people to order."
A father and son, brewing beer. Connected their equipment via Bluetooth. Built an entire ordering website. Automated the whole operation. With an open-source tool anyone can install from a terminal.
Then there was the lobster story: "Hundreds of people are queuing up for lobsters" — a small business that automated their ordering system and saw immediate, tangible demand.
These stories matter because they illustrate something the enterprise world often forgets: the most powerful technologies are the ones that regular people can use to build real things. That is what made the web transformative. That is what made smartphones transformative. And that, Huang argued, is what makes OpenClaw transformative.
"Believe it or not, there is already a ClawCon," Huang added with visible amusement. "Incredible."
NemoClaw: Why Enterprise Security Changes Everything
Here is where the keynote shifted from inspirational to urgent.
Huang paused, and asked the audience to consider something uncomfortable:
"Agentic systems in the corporate network can have access to sensitive information. It can execute code and it can communicate externally. Just say that out loud."
He repeated it for emphasis. Access sensitive information. Execute code. Communicate externally.
"You could of course access employee information, access supply chain, access finance information, sensitive information and send it out, communicate externally. Obviously, this cannot possibly be allowed."
This is the tension at the heart of the agentic AI revolution. The same capabilities that make AI agents extraordinarily useful — autonomous execution, data access, external communication — are precisely what make them dangerous in an enterprise context without proper governance.
A personal AI agent that can read your email, execute code on your servers, and send messages to anyone is either the most productive tool ever built or the most dangerous security vulnerability ever introduced. The difference is entirely a function of the security architecture wrapped around it.
That is where NemoClaw enters the picture.
What NVIDIA Actually Built
"We worked with Peter," Huang explained. "We took some of the world's best security and computing experts and we worked with Peter to make OpenClaw enterprise-ready."
The result is what NVIDIA calls NemoClaw — a reference design for enterprise OpenClaw deployments. It includes several critical security layers:
Open Shell — a technology NVIDIA developed that has now been integrated directly into OpenClaw's core. This provides the foundational security sandbox within which agents operate.
Policy engine connectors — every major enterprise already has governance and compliance systems. NemoClaw connects to these existing policy engines, so companies do not need to rebuild their security infrastructure from scratch. "Your policy engines are super important, super valuable," Huang noted. "The policy engines could be connected."
Privacy router — controls the flow of data between agents and external systems. Decides what an agent can see, what it can share, and what channels it can communicate through.
Guard rails — behavioral boundaries that prevent agents from taking actions outside their approved scope, even when prompted to do so.
"As a result we could protect and keep the claws from executing inside our company and do it safely," Huang said.
The naming convention is deliberate. Nemo — NVIDIA's broader AI platform — combined with Claw — the agent framework. NemoClaw. The enterprise wrapper around the open-source engine.
What makes this strategically significant is not just the technology itself, but the partnership model. NVIDIA did not fork OpenClaw. They did not acquire it. They contributed security infrastructure back into the open-source project and built a reference design on top of it. This means every enterprise that deploys NemoClaw is also deploying OpenClaw — reinforcing the open standard rather than fragmenting it.
This is the Linux playbook, executed perfectly. An open-source kernel. Enterprise distributions built on top. The ecosystem grows together.
Every SaaS Company Will Become a GaaS Company
Huang then made what may be the most consequential business prediction of 2026:
"Every single SaaS company will become a GaaS company. No question about it. Every single SaaS company will become a GaaS company, an Agentic as a Service company."
GaaS — Generative AI as a Service. Not a buzzword. A structural transformation of the entire enterprise software industry.
To understand why Huang is so confident, you need to understand how he sees the current enterprise IT landscape — and how fundamentally OpenClaw changes it.
The Old Model: Tools for Humans
Huang described enterprise IT before OpenClaw with unusual clarity: "The reason why it is called data centers is because these large buildings held data, held the files of people, the structured data of business. And we pass through software that has tools and systems of records and all kinds of workflow that is codified into it and that turns into tools that humans would use."
Data centers. Files. Software. Tools. Humans. That is the stack. That is the $2 trillion enterprise IT industry in five words.
And then Huang added the ecosystem around it: "GSIs — consultants that help companies figure out how to use these tools and integrate these tools. These tools are incredibly valuable for governance and security and privacy and compliance and all of that continues to be true."
He was careful to acknowledge that existing enterprise infrastructure is not going away. The governance, security, privacy, and compliance layers that companies have built over decades remain essential. What changes is everything above them.
The New Model: Agents Instead of Tools
"Post OpenClaw, post agentic — this is what it is going to look like," Huang declared. "Every single IT company, every single company, every SaaS company will become a GaaS company."
The transformation is structural. Where companies currently sell tools that humans operate, they will sell agents that operate autonomously within human-defined boundaries. The value proposition shifts from "here is a tool you can use" to "here is an agent that does the work for you."
This is not incremental. It is categorical. And Huang drew explicit parallels to the three previous platform shifts that reshaped enterprise computing:
HTML — "just as HTML showed up, it made it possible for the entire industry to grab onto this open-source stack and go do something with it." The web did not replace computing. It created an entirely new layer of value on top of it.
Linux — "just as Linux gave the industry exactly what it needed at exactly the time." Linux did not replace proprietary operating systems overnight. It created a shared foundation that enabled an ecosystem.
Kubernetes — "just as Kubernetes showed up at exactly the right time." Container orchestration did not replace servers. It abstracted them, enabling cloud-native computing at scale.
"OpenClaw gave us, gave the industry exactly what it needed at exactly the time," Huang said. The fourth platform shift. The one that transforms every software company from a tool vendor into an agent vendor.
The math behind this prediction is staggering. The enterprise IT industry is currently valued at roughly $2 trillion. Huang called what is coming "a multi-trillion dollar industry" — not just growth within the existing category, but the creation of entirely new categories of value that did not previously exist.
The Neotron Coalition and the Open Model Frontier
No platform shift succeeds alone. Huang knows this — he built NVIDIA's dominance not by going solo, but by assembling ecosystems. And that is exactly what he announced here.
The Coalition Partners
The Neotron Coalition is a group of companies partnering with NVIDIA to advance the Neotron 4 model and integrate the NemoClaw reference design. The names speak for themselves:
Black Forest Labs — the imaging company behind some of the most advanced image generation systems
Cursor — "the famous coding company," as Huang called them. "We use lots of it." When the CEO of NVIDIA publicly endorses your developer tool, it says something about where software development is heading.
LangChain — with billions of downloads, the de facto standard for building custom AI agents. Their integration means NemoClaw agents can leverage the entire LangChain ecosystem.
Mistral — the frontier model company. Huang acknowledged Arthur Mensch in the audience: "Incredible company."
Perplexity — "Everybody use it. It is so good. A multimodal agentic system." When Jensen Huang gives an unsolicited product endorsement on stage, it is worth paying attention to.
Reflection, Sarv, Thinking Machine, Mirror, Morardi's Lab — companies spanning reasoning systems, Indian AI development, and advanced research.
The Enterprise Integration Wave
Beyond the coalition, Huang revealed that enterprise software companies across the board are integrating NemoClaw: "One company after another. There are so many. And we are partnering with all of you."
His message to the enterprise world was unambiguous: every company needs an OpenClaw strategy. Not eventually. Now. "Just as we needed to have a Linux strategy, we all needed to have a HTTP HTML strategy which started the internet, we all needed to have a Kubernetes strategy — every company in the world today needs to have an OpenClaw strategy and an agentic system strategy."
Neotron 3 and the Open Model Frontier
Alongside the platform and partnership announcements, NVIDIA showcased their open model ecosystem — and the numbers are remarkable. The NVIDIA open models initiative now encompasses nearly 3 million open models spanning six families:
Neotron — reasoning models for language, visual understanding, RAG, safety, and speech. Neotron 3 is now ranking among the top three models in the world when deployed inside OpenClaw.
Cosmos — frontier models for physical AI, world generation and understanding. Already on generation 2.
Groot — general purpose robotics foundation models, also at generation 2.
Alpayo — "the world's first thinking and reasoning autonomous vehicle AI"
Bioneo — open models for biology, chemistry, and molecular design
Earth 2 — models for weather and climate forecasting rooted in AI physics
Huang made a commitment that matters enormously for long-term planning: "Our models are valuable because number one, they are on the top of the leaderboard, world class. But most importantly, it is because we are not going to give up working on it. We are going to keep on working on it every single day. Neotron 3 is going to be followed by Neotron 4. Cosmos 1 was followed by Cosmos 2."
This is the enterprise promise: you are not building on an abandoned foundation. NVIDIA is committing ongoing engineering resources to every model family. Vertical integration with horizontal openness — their models, open for everyone to fine-tune and deploy.
The Neotron 3 Ultra announcement was particularly significant: "It is going to be the best base model the world has ever created. This allows us to help every country build their sovereign AI." A foundation model designed explicitly for national-scale customization — enabling countries and industries to build domain-specific AI without starting from scratch.
The Token Economy and What It Means for Every Business
Perhaps the most provocative section of the keynote was not about technology at all. It was about compensation.
Huang made a prediction that, if it proves correct, will fundamentally change how companies think about employee investment:
"Every single engineer in our company will need an annual token budget. They are going to make a few hundred thousand a year, their base pay. I am going to give them probably half of that on top of it as tokens so that they could be amplified 10x."
Read that again. Half of an engineer's salary — potentially $150,000 or more — allocated not as cash compensation, but as AI compute tokens. The explicit goal: 10x amplification of every individual's productive output.
"Of course we would," Huang continued. "It is now one of the recruiting tools in Silicon Valley. How many tokens come along with my job."
This is already happening. Not in five years. Not as a pilot program. As a recruiting differentiator today. Companies that provide generous token budgets are attracting better talent because those engineers know they will be dramatically more productive.
The AI Factory Model
Huang connected the token economy to a broader infrastructure vision. Every company currently sits on top of file systems and data centers. In the post-OpenClaw world, every company becomes three things simultaneously:
Token users — consuming AI compute to amplify their workforce
Token manufacturers — producing specialized AI compute for their customers
Agent operators — deploying autonomous agents that create value continuously
"Those tokens as you know will be produced by AI factories that all of you and us we partner to build," Huang explained. The physical infrastructure — GPUs, data centers, cooling systems — becomes an AI factory. The output is not products or services in the traditional sense. The output is tokens. Intelligence on tap.
The companies that understand this shift earliest will have a compounding advantage. An engineer with $150K in compute tokens does not just work 10% faster. They work categorically differently — delegating entire workstreams to AI agents, reviewing and steering rather than executing manually, focusing human judgment where it matters most.
Multiply that across an organization of thousands, and the productivity gap between token-rich and token-poor companies becomes unbridgeable within a few years.
What This Means for Every CEO
Huang's closing challenge was direct and unambiguous:
"What is your OpenClaw strategy?"
This is not a rhetorical question. It is the strategic equivalent of asking "What is your internet strategy?" in 1998 or "What is your cloud strategy?" in 2012. Companies that treated those questions as optional are, in many cases, no longer companies.
The enterprise IT industry today is valued at roughly $2 trillion. Huang believes the agentic transformation will push it to "multi-trillion" — not by replacing what exists, but by creating entirely new categories of value. Companies will not just sell software tools. They will rent specialized AI agents. They will not just store data. They will manufacture intelligence from it.
The infrastructure is open source. The enterprise security layer is available. The coalition is assembled. The models are frontier-class. The question is no longer whether this transformation will happen. It is whether your organization will lead it or be disrupted by it.
The Bottom Line
NVIDIA just did for AI agents what they did for GPUs in gaming, for CUDA in deep learning, and for DGX in data centers: they identified the defining technology of the next decade, created the reference architecture, open-sourced the critical infrastructure, built enterprise security around it, and assembled the largest coalition of AI companies ever to drive adoption.
OpenClaw is the operating system. NemoClaw is the enterprise layer. Neotron is the intelligence. The Neotron Coalition is the ecosystem. And every company in the world just got put on notice.
The age of personal agents is not a future prediction. It is a present reality. A 60-year-old father is already using it to automate his beer business. Small businesses are already using it to serve hundreds of customers. And now, with NVIDIA's full backing and enterprise security integration, every Fortune 500 company has no excuse not to start.
Jensen Huang ended with a word he used three times: renaissance. "This is a reinvention. This is a renaissance of enterprise IT."
Renaissances do not ask permission. They just begin. This one already has.
This article is based on Jensen Huang's keynote announcing NVIDIA's NemoClaw reference design and partnership with OpenClaw. Watch the full keynote: NVIDIA NemoClaw Launch
Changelog