Jensen Huang took the stage at GTC 2026 last week and posed a question that should be echoing through every boardroom: "What's your OpenClaw strategy?" It is the same question that should have been asked twenty years ago about the internet. And ten years ago about the cloud. The companies that did not listen back then no longer exist.
TL;DR — At GTC 2026, Jensen Huang declared that every company on earth needs an OpenClaw strategy, just as companies once needed an internet and cloud strategy. OpenClaw is an autonomous AI agent platform that independently executes tasks. NVIDIA simultaneously launched NemoClaw, the enterprise version with built-in security. Companies not thinking about AI agents now risk making the same mistake as the retail giants that missed the internet wave: moving too late in a shift that shows no mercy.
The man in the leather jacket
If you do not know Jensen Huang, you certainly know his company. NVIDIA, founded in 1993 by Huang together with Chris Malachowsky and Curtis Priem, started as a maker of graphics chips for gamers. Thirty years later it has become one of the most valuable companies in the world. The reason: those graphics chips turned out to be exactly the hardware needed to train AI models.
Huang is one of those rare CEOs who understands both the technology and the business at boardroom level. He is known for his keynotes in his inseparable leather jacket, where he sketches the future of computing for two to three hours with infectious enthusiasm. At GTC 2025 it was about AI data centers. At CES 2026 about robotics. But his keynote on March 16, 2026 at GTC in San Jose was different. This time he had a message meant not just for technologists.
This year CUDA, NVIDIA's platform for parallel computing on graphics chips, celebrates its twentieth anniversary. What began as an architecture few people believed in now has hundreds of millions of installations worldwide. It is the foundation upon which the entire AI revolution was built. But Huang looked beyond the hardware in his keynote. He outlined a future in which data centers transform into what he calls "AI factories," factories whose product is not electricity or steel but tokens. The number of tokens you can produce per watt becomes the most important benchmark for your future revenue.
The core of that message: every company in the world now needs an OpenClaw strategy. Just as every company once needed an internet strategy. Just as every company needed a cloud strategy. Those who do not think about how AI agents fit into their operations are making the same mistake that Dutch retailers Blokker, V&D, and Kijkshop made when the internet arrived.
What is OpenClaw and why does every company need a strategy for it?
To understand why Huang was so excited, you first need to understand what OpenClaw is and who is behind it.
Peter Steinberger is an Austrian software developer who spent thirteen years building PSPDFKit, a company that made PDF tools for developers. The company grew into a globally used product and Steinberger sold it for around one hundred million euros. Then he burned out. He could no longer write code, stared at his screen, and felt empty. He booked a one-way ticket to Madrid and disappeared.
But while Steinberger was recovering in Spain, the AI world was changing at a breathtaking pace. When he cautiously tried to build a Twitter analytics tool, he discovered that AI had undergone a paradigm shift. The technology could now take over the tedious plumbing work of programming, freeing him to focus again on the creative building itself.
In November 2025 Steinberger built a prototype of what would eventually become OpenClaw in a single hour. It started as Clawdbot (a nod to Anthropic's Claude), was renamed Moltbot under legal pressure, and ultimately landed on OpenClaw. The project exploded. Within sixty days it had accumulated more than 250,000 GitHub stars, making it the fastest-growing open source project in history. Faster than Linux. Faster than React.
What makes OpenClaw different from a regular chatbot? A chatbot answers questions. OpenClaw does things. It is an autonomous AI agent platform that runs locally on your own hardware. It can manage your email, control your browser, and automate work processes, all through messaging services like WhatsApp or Telegram. It is always on. It does not wait for a prompt; it keeps working.
Steinberger described it as a digital employee. In February 2026 he joined OpenAI to advance the project further, while OpenClaw itself continues as an independent open source foundation.
NemoClaw: OpenClaw ready for the enterprise
Huang saw in OpenClaw what he had once seen in Linux: an open source foundation around which an entire industry can organize. And so NVIDIA also launched NemoClaw during that same keynote. That is the enterprise version of OpenClaw, built specifically for companies that do not want to simply unleash an open source tool on their network.
Where regular OpenClaw runs on your own hardware and can in principle do anything, including sending sensitive company data outside the organization, NemoClaw adds a layer of security and control. Think sandboxing so agents cannot operate outside their assigned environment, a privacy router that prevents internal data from leaving the corporate network, and a policy layer that lets you precisely configure which actions an agent may and may not take. Installation happens with one command. It runs on your own hardware, in the cloud, or on an NVIDIA laptop.
For executives the message is simple: OpenClaw is the technology, NemoClaw is what you can safely deploy as a company. NVIDIA has deliberately lowered the barrier. The question is no longer whether the technology is ready. The question is whether you are.
"Every company needs an OpenClaw strategy"
Let us pause on the exact comparison Huang made. He said on stage: every company needed a Linux strategy. Every company needed an HTML strategy, which marked the beginning of the internet age. Every company needed a Kubernetes strategy for the cloud. And now every company needs an OpenClaw strategy.
What Huang is actually saying, and this is important to understand, is not that every company must specifically start using OpenClaw. What he means is that every company needs an agentic AI strategy. A plan for how to deploy AI agents that can reason independently, break down tasks, use tools, and execute actions in your digital environment.
He sketched a future in which traditional SaaS companies transform into what he called "Agentic as a Service." Instead of delivering software that people operate, you deliver specialized AI agents that operate the software. That is a fundamentally different way of thinking about what a software company is.
And then he said something that made everyone in the room take notice. At NVIDIA, every engineer will soon receive an annual token budget alongside their salary. A base salary of a few hundred thousand dollars, plus an allocation of roughly half of that in tokens, so that engineer can be ten times as productive. The question in Silicon Valley job interviews has already shifted from "how much vacation do I get?" to "how many tokens are in the offer?"
That is not future music. It is happening now.
The internet strategy that never came
To understand why Huang's warning is so urgent, you do not need to look to Silicon Valley. You only need to look at the Dutch high street.
Blokker (a Dutch household goods retailer founded in 1896) was a fixture of the Dutch retail landscape for 128 years. Nearly four hundred stores, billion-euro turnover at its peak, and a brand name every Dutch person knew. In November 2024 the curtain fell definitively. The bankruptcy affected 3,500 employees.
What went wrong? The list is long, but one factor stands out. The Blokker family did not believe in the internet. They thought consumers would keep coming to stores to touch and feel products. The webshop they eventually launched was a halfhearted afterthought. While online retailers Bol.com and Coolblue dominated the online market, Blokker clung to a strategy that was already ten years behind. The painful truth is that Blokker did not fail because the product was bad. It failed because management did not understand that the rules of the game had fundamentally changed.
De Kijkshop (a Dutch showcase-concept retailer) was actually an innovator in the 1970s and 1980s. The showcase concept, where you viewed products behind glass and wrote down numbers, was revolutionary at the time. But when the internet offered the possibility of viewing and ordering products from home, the showcase concept shifted from innovation to anachronism. In 2018 the last seventy stores closed. Four hundred people lost their jobs. An attempt to continue as a webshop ended in a second bankruptcy in 2021.
V&D (Vroom and Dreesmann, a major Dutch department store chain), which served generations of Dutch consumers, went bankrupt at the end of 2015 and closed its doors definitively at the start of 2016. More than ten thousand employees lost their jobs. The causes were complex, but the core was the same: moving too late, holding too long to a model that no longer worked in a digital world.
Free Record Shop is perhaps the most poignant example. Hans Breukhoven built a record store in Schiedam from 1971 into more than four hundred branches with annual turnover of half a billion euros. But when iTunes sold its ten billionth song in 2010, Breukhoven still had no answer to the digital revolution. The stores tried everything, from internet cafes to gift items, but it was a losing battle. By 2013 the first bankruptcy came. In 2014 the definitive end.
All these companies had one thing in common: they had no internet strategy. Or more precisely, they had no strategy that accounted for the fundamental shift in how consumers behaved.
The cloud strategy nobody took seriously
The next wave was the cloud. Around 2010-2015 the IT world shifted from its own servers to cloud solutions. Companies that had not invested in their IT systems during the financial crisis were running three to four software versions behind. They were working with outdated on-premise systems while their competitors were already in the cloud.
This pattern was broadly visible in the Netherlands. SMEs lagged behind. In 2024 only 81.5 percent of Dutch SMEs applied a basic level of digital technologies. That may sound high, but consider that "basic level" already includes having a website and using email.
The companies that did embrace the cloud transition, a Coolblue, a Bol.com, an Adyen, grew into dominant players. The companies that stayed stuck in their old systems lost ground step by step. Not in one dramatic bang, but in a slow erosion of competitive strength.
The lesson of the cloud was more subtle than that of the internet. With the internet it was clear: if you are not online, you miss customers. With the cloud it was about efficiency, scalability, and speed of innovation. Companies without a cloud strategy could not move fast enough with the market. They could not launch new services, analyze data, or facilitate remote work. When COVID hit, it was precisely these companies that were hit hardest.
Philips is a good example of a Dutch company that handled the cloud transition well. By building a digital health platform together with Microsoft, the company transformed from a traditional electronics manufacturer into a digital health partner. Adyen is another example: it built a cloud-native payment infrastructure from day one and grew into one of the most valuable tech companies in Europe.
BCC (a Dutch consumer electronics chain) is a striking example of the other side. Once part of the same group as Blokker, it went bankrupt in 2023. While Coolblue invested in an advanced logistics platform, its own delivery service, and data-driven personalization, BCC held on to a traditional retail model with a webshop that looked more like a digital flyer than a modern e-commerce experience. The difference was not the product range. The difference was the technological backbone on which the business ran.
Why the agentic AI wave is different
Now we get to the point where things get interesting. The agentic AI wave Jensen Huang describes is not just the next incremental step. It is a fundamentally different way of working.
The internet was about reach. The cloud was about infrastructure. Agentic AI is about capacity. AI agents literally multiply what a person, a team, or a company can do.
Jensen Huang used the example of kitchen design. With one prompt, an OpenClaw agent can study images, learn to use design tools, iterate on ideas, and improve its own output. All autonomously. His conclusion was as simple as it was revolutionary: every carpenter can now become an architect. Every plumber can become an architect. We are going to amplify the capabilities of everyone.
That sounds like a Silicon Valley promise, but the technology is here already. OpenClaw runs on a laptop. Claude Code, Cursor, and similar tools are already daily workhorses for developers. And it is not just about code. AI agents schedule meetings, analyze data, write reports, manage customer relationships, and execute complete workflows.
The question is not whether this will affect your business. The question is when, and whether you are ready.
What an agentic AI strategy looks like for a trade contractor
Let us make it concrete. Take a heating and plumbing contractor with 75 employees. Boilers, heat pumps, solar panels, climate installations. A typical SME with an owner-manager, a small administrative team, a few project coordinators, and fifty to sixty field technicians.
Such a company today faces an enormous amount of administration: quotes, work orders, invoices, purchase orders, certifications, safety protocols. On top of that there is a chronic shortage of technicians, rising material costs, and customers who expect faster and faster responses.
An agentic AI strategy for this company might look as follows.
Phase 1: Automate the paper flow. An AI agent that automatically processes incoming quote requests. The agent reads the request, checks material availability with suppliers, calculates the price based on historical quotes and current material prices, and generates a draft quote. The project coordinator only needs to review and approve. This saves eight to ten hours per week.
Phase 2: Smart scheduling and dispatching. An agent that optimizes technician scheduling based on location, specialization, availability, and traffic information. If a job runs over, the agent automatically adjusts the rest of the day's schedule and informs affected customers. The planner shifts from executor to supervisor.
Phase 3: Knowledge management and support. Field technicians run into technical problems. An AI agent with access to all technical documentation, previous service history, and manufacturer information can provide direct answers via a messaging app. What error code is this? What part do I need to order? How do I solve this with this specific boiler model? The knowledge of your most experienced technician becomes available to the entire team.
Phase 4: Proactive customer management. An agent that monitors the maintenance database and automatically contacts customers when a service visit is due, a warranty expires, or a new subsidy scheme becomes available for energy efficiency improvements. From reactive to proactive customer contact, without needing extra staff.
Phase 5: Insight and prediction. An agent that recognizes patterns in breakdowns, predicts seasonal peaks, and advises on material purchasing based on historical data and market prices. The director receives dashboards produced not by an intern in Excel but by an agent combining real-time data.
None of these steps requires you to set up an IT department. The technology is available as a service. The investment is a fraction of what comparable automation would have cost five years ago. And it starts with one use case, one agent solving one problem.
The freelancer and the agentic revolution
Where the trade contractor has the advantage of scale, the freelancer has the advantage of agility. And it is precisely for freelancers that the impact of agentic AI is potentially greatest.
In the Netherlands there are around 1.2 million self-employed professionals. The most common profession among service-providing freelancers is business advisor or organizational consultant. Average hourly rate in 2026: 83 euros. Most consultants spend a significant part of their time on non-billable tasks: acquisition, administration, research, writing proposals, formatting reports.
Research and preparation. For every new assignment the consultant currently does research manually. With an AI agent that monitors the client's sector, conducts competitive analyses, and collects relevant benchmarks, the consultant can deliver a deeper analysis in a fraction of the time.
Content production. Many consultants build their brand through content: LinkedIn posts, articles, white papers. An AI agent that knows your writing style, tracks relevant sources, and delivers drafts doubles your thought leadership output without sacrificing quality.
Administration and acquisition. An agent that maintains your CRM, schedules follow-ups, generates proposals based on intake notes, and prepares your bookkeeping for the accountant. The average consultant gains five to eight hours per week from this. At 83 euros per hour that is four hundred to six hundred euros per week in freed-up capacity.
Quality improvement. An agent that reviews your advisory reports for consistency, spots data errors, and adds benchmarks you would not have found yourself. You deliver better work, faster. And that leads to satisfied clients and repeat business.
But it goes beyond efficiency alone. The freelance consultant with a good agentic AI strategy can offer services that were previously only feasible for larger firms. Think continuous monitoring of KPIs for your client, automated monthly reports, or an always-on AI assistant that gives your client direct answers between advisory sessions. Your service offering shifts from periodic projects to continuous value creation. And that changes not only your business model but also your pricing model. From hourly rates to subscriptions. From one-off assignments to ongoing relationships.
Steinberger himself is the ultimate example of what an individual can achieve with AI agents. He built OpenClaw, a project of 300,000 lines of code, largely alone. The AI wrote the code, the AI ran the tests, and he clicked confirm. That sounds reckless, but it is the reality of what Steinberger calls "ambient programming." And the result was the fastest-growing open source project in history.
Jensen Huang put it aptly: we are going to amplify the capabilities of everyone. For the freelancer this means that as a one-person operation you can have the firepower of a small agency. Not by working harder, but by working smarter alongside AI agents.
The mistake you cannot afford to make
Let us return to the core of Huang's message. The companies that missed the internet wave, Blokker, Kijkshop, V&D, Free Record Shop, all made the same mistake. They saw the technology as something for others. As a hype that would blow over. As something their customers would not embrace.
By the time they understood that the world had changed, it was too late. Not because the technology was not available, but because their competitors had already built an insurmountable lead.
The cloud wave was more subtle. There it was not about sudden death but about a gradual loss of competitive strength. Companies without a cloud strategy became slower, more expensive, and less flexible than their competitors who were in the cloud.
The agentic AI wave combines both patterns. It is both an existential risk and a competitive risk. Companies that deploy AI agents become exponentially more productive. They serve more customers with fewer people, they respond faster, they make fewer mistakes, they innovate more quickly.
And the window to act is smaller than you think. OpenClaw grew to 250,000 GitHub stars in sixty days. By comparison, Linux took thirty years to get there. The adoption curve of agentic AI is not linear. It is exponential.
What you can do tomorrow
You do not need to overhaul your entire business tomorrow. But you do need to start thinking. Here are three concrete steps:
Step 1: Identify your biggest time wasters. Which processes in your business cost disproportionate amounts of time relative to the value they deliver? Administration? Customer service? Reporting? Scheduling? Those are your first candidates for an AI agent.
Step 2: Experiment small. Give one employee the space to work with AI tools for a week. Not on a big project, but on daily drudgery. Let that person discover what is possible. The surprises will come on their own.
Step 3: Make it strategic. Put it on the executive agenda. Not as an IT project but as business strategy. Just as you once formulated an internet strategy and a cloud strategy, you now need an agentic AI strategy. Who will lead this? What is the budget? What are the first three use cases?
The point of no return
Jensen Huang is not known for false modesty. But his comparison of OpenClaw with Windows, Linux, HTML, and Kubernetes is not just a sales pitch. It is an observation about how technological transitions work.
Every major technological shift has a tipping point. A moment at which the technology becomes available, affordable, and usable enough to trigger mass adoption. For the internet that was the browser. For the cloud that was AWS. For agentic AI it is OpenClaw and the platforms being built around it.
We are at that tipping point right now. Jensen Huang's question to every CEO, including you, is simple: what is your OpenClaw strategy?
Or more precisely: what is your strategy for deploying AI agents as full-fledged digital employees in your organization?
The Blokker family thought their customers would keep coming to the store. Hans Breukhoven thought people would keep buying CDs. Kijkshop thought showcases were the answer.
They were all wrong. And they paid the highest price for it.
The question is: are you going to make the same mistake?
Or will you ask yourself today the question Jensen Huang posed to thirty thousand people in San Jose: what is my OpenClaw strategy? The answer does not need to be perfect. It does not need to be complete. But it needs to exist. Because history teaches us one thing with absolute certainty: the companies that start thinking too late are the companies we will remember in ten years as cautionary examples.
Sources
- NVIDIA Blog, NVIDIA GTC 2026: Live Updates on What's Next in AI, March 16, 2026. blogs.nvidia.com
- TechCrunch, Nvidia's version of OpenClaw could solve its biggest problem: security, March 16, 2026. techcrunch.com
- Fortune / Yahoo Finance, Who is OpenClaw creator Peter Steinberger?, February 2026. fortune.com
- AllClaw.org / Fortune: exit of approximately 100 million euros for PSPDFKit. allclaw.org
- Lex Fridman interview with Steinberger, February 2026. yahoo.com
- PANews / AllClaw.org: prototype built in November 2025 in one hour. allclaw.org
- Bosio Digital, NVIDIA's CEO Asked Every Company a Question, March 2026. bosio.digital
- CNBC, Steinberger joined OpenAI last month, March 16, 2026. cnbc.com
- TechCrunch / Atlan GTC 2026 recap: NemoClaw architecture with three security layers. techcrunch.com
- SDxCentral, NVIDIA GTC 2026: Jensen's victory lap, March 2026. sdxcentral.com
- Fortune, Jensen Huang thinks $1 trillion won't be enough, March 17, 2026. fortune.com
- Hart van Nederland / NOS, November 13, 2024: Blokker declares bankruptcy. hartvannederland.nl
- Wikipedia NL, Kijkshop. nl.wikipedia.org
- NOS / various sources: V&D bankrupt end of 2015. nos.nl
- Various sources: Free Record Shop, first bankruptcy 2013, definitive end 2014.
- Eurostat / CBS SME digitalization 2024: 81.5% basic level digital technologies. cbs.nl
- CNBC, Nvidia CEO Jensen Huang says OpenClaw is 'definitely the next ChatGPT', March 17, 2026. cnbc.com
- CBS StatLine 2025: approximately 1.2 million self-employed in the Netherlands. cbs.nl
- CBS / ZZP Barometer 2026: average hourly rate for advisory and consultancy approximately 83 euros.
- The-AI-Corner.com, OpenAI Acquires OpenClaw Creator, February 2026. the-ai-corner.com