Jensen Huang has a simple observation about his own workforce. His software engineers would rather construct AI agents than sit down and write another line of Python. The remark, delivered in a recent NVIDIA-published interview, captures a transformation already underway inside the chipmaker that powers much of the world’s artificial intelligence infrastructure.
“These agentic systems are new skills, and now we have a lot of software engineers building agents,” Huang told interviewers. “If you ask me, every one of my software engineers prefers to be building agents than to be writing Python code.” The comments, highlighted by Business Insider on July 9, 2026, arrived alongside NVIDIA’s latest push into enterprise agent technology. But the CEO’s point runs deeper than preference. It signals a fundamental redefinition of what engineering work means in the age of capable AI systems.
Coding, in Huang’s view, has started to resemble typing. Mundane. Repetitive. Something better left to the machines. Engineers instead focus on imagination. On creativity. On the architecture that lets agents handle complex, multi-step tasks while humans set direction and review outcomes. “You’re taking all the mundane work, and you’re trying to get this agent to do it,” he explained. “That requires imagination, that requires creativity, a lot of technology.”
The timing feels deliberate. Just hours before the interview surfaced, NVIDIA and LangChain unveiled a new blueprint called NemoClaw for LangChain Deep Agents. The stack combines NVIDIA’s Nemotron 3 Ultra model with LangChain’s orchestration tools and a secure runtime environment. Early benchmarks show it delivers strong performance at a fraction of the cost of closed models. Enterprises can now tune these agents on their own data, deploy them across cloud or on-premise infrastructure, and maintain governance without starting from scratch.
Yet Huang’s comments stretch back months. At the company’s GTC conference in March 2026, he painted an even bolder picture. In ten years, he predicted, NVIDIA could operate with 75,000 human employees supported by 7.5 million AI agents. That works out to one hundred agents for every person. “In 10 years, we will hopefully have 75,000 employees, as small as possible, as big as necessary,” Huang said during a media Q&A, according to Fortune. “Those 75,000 employees will be working with 7.5 million agents.”
The agents, he added, would labor around the clock. They would shoulder grunt work. Humans would stay busy tackling problems once considered unsolvable. Drug discovery reframed as engineering. Life extension discussed seriously. “We will all feel superhuman,” Huang declared. The vision contrasts sharply with warnings from other AI leaders about job losses. Huang has consistently argued the opposite. AI, he insists, creates far more work than it removes.
“The amount of work that we have to do to bring AI into the world is really quite incredible,” he said in the recent interview. “So it’s creating a whole bunch of jobs. And, my software engineers love this.” He doubled down in a May television appearance, telling viewers that AI represents “the United States’s best opportunity to re-industrialize ourselves.”
This stance has drawn skeptics. Anthropic CEO Dario Amodei and Amazon’s Andy Jassy have voiced concerns about white-collar displacement. Huang dismisses the fear. For him, the evidence sits inside his own company. Engineers once chained to editors now design agent behaviors, establish benchmarks, and erect guardrails. The creative spark moves upstream.
But building effective agents proves no small feat. These systems must plan, use tools, maintain memory, recover from failure, and operate securely at scale. Most university computer science programs still ignore the topic. Bootcamps rarely touch agent memory architecture or production harness engineering. The skills gap is real. And NVIDIA wants to close it.
Enter the partnership with LangChain. Their newly released blueprint, detailed on the LangChain blog yesterday, offers enterprises a complete open stack. Nemotron 3 Ultra provides the reasoning core. LangChain Deep Agents handles orchestration. NVIDIA’s OpenShell runtime adds security boundaries. The combination reportedly achieves benchmark-leading results at roughly one-tenth the cost of some proprietary alternatives.
Harrison Chase, LangChain’s co-founder and CEO, joined Huang for a fireside discussion around the launch. The pair explored what it takes to move agents from prototypes to production systems that enterprises can trust. Their conversation, available on YouTube via LangChain, underscores a shared belief. Open, customizable agent platforms will dominate. Closed systems may dazzle in demos but struggle with data privacy, customization, and cost at scale.
Huang has urged every company to develop what he calls an “OpenClaw strategy.” The term refers to an open-source agent control layer created by researcher Peter Steinberger that gained explosive popularity. At GTC, Huang likened it to the operating system for a new era of personalized AI. “Every company in the world today needs to have an OpenClaw strategy,” he said, per a Reuters report from March. “This is the new computer.” NVIDIA followed up with NemoClaw, a more enterprise-ready variant developed in partnership with Steinberger.
The implications stretch across industries. Software development teams that once measured output in commits may soon track agent fleets managed per engineer. One builder recently noted on X that every engineer could soon handle hundreds of agents. The most valuable skills in 2026, he argued, involve designing systems that survive real-world production pressures.
Wall Street has taken notice. NVIDIA’s stock has ridden the AI wave to become one of the world’s most valuable companies. Yet the narrative is shifting. No longer just a supplier of graphics processors, the firm positions itself as the full-stack provider for agentic systems. Chips still generate revenue today. The software layer, the blueprints, the runtimes, those create switching costs tomorrow.
Critics wonder whether the enthusiasm outruns reality. Current agents still stumble on complex, long-horizon tasks. Hallucinations persist. Integration with legacy enterprise systems remains messy. Guardrails can be brittle. Huang acknowledges the challenges. He simply believes the trajectory points one direction. More agents. Fewer manual keystrokes. Greater human focus on strategy and invention.
Inside NVIDIA itself, the experiment continues. If Huang’s prediction holds, the company will one day employ 100 agents for every human. Those agents will run experiments, analyze data, draft reports, and iterate without sleep. Engineers will orchestrate rather than implement. Managers will govern rather than micromanage. The organizational chart of the future may list both people and their digital counterparts.
Not everyone buys the timeline. Ten years can feel distant in technology. Breakthroughs arrive unevenly. Yet the direction of travel looks clear. Engineers already show preference for agent construction over traditional coding. Companies race to assemble their own agent strategies. And NVIDIA supplies both the compute and the reference architectures.
Huang’s message carries a note of optimism that feels increasingly rare. AI need not hollow out the workforce. It can amplify it. The mundane work disappears. The interesting problems multiply. Creativity gains room to breathe. His engineers, it seems, have already voted with their time. They would rather build the future than type instructions for it.
Whether the rest of the industry follows remains an open question. But the tools have arrived. The examples exist. The CEO who bet his company on accelerated computing now bets it on accelerated agency. So far, the market likes the wager.
WebProNews is an iEntry Publication