Nvidia’s software engineers are writing less code than ever, and according to its CEO, that’s exactly how they like it.
Jensen Huang said this week that his engineers have come to prefer building AI agents to writing Python, a change he sees as a promotion rather than a threat.
“These agent systems are new skills, and we now have a lot of software engineers creating agents,” Huang said in an interview published by the company on Wednesday.
“If you ask me, every one of my software engineers would rather be construction agents than write Python code.” It is a message that has been sharpening for months, from a Carnegie Mellon Commencement toward Computex Stage.
The distinction he draws is between coding as a task and engineering as a craft. By his account, the typing part of the job, the mundane work of turning an idea into syntax, is now something an agent can take on.
The 💜 of EU technology
The latest rumors from the EU tech scene, a story from our wise founder Boris and some questionable AI art. It’s free, every week, in your inbox. Register now!
“You’re taking all the mundane work and you’re trying to get this agent to do it,” he said. “That requires imagination, creativity and a lot of technology.” Instead of churning out lines of Python, its engineers now spend their days building agents, writing benchmarks, and designing the barriers that keep those systems in check.
An AI agent breaks down a large goal into a sequence of smaller steps, each handled in turn, so that the software can plan and act instead of simply answering a single question.
Huang, co-founder of Nvidia in 1993, has become one of the strongest advocates of putting AI assistants to work within companies. He has repeatedly raised a future in which Nvidia deploys agents across all divisions to increase productivity, an ambition that sits comfortably with selling chips and platforms who direct them.
That business logic is never far from the surface. Nvidia spent the last year positioning itself as the infrastructure behind the agent economy, and a workforce that builds agents is also a workforce that increasingly consumes compute.
‘Lots of jobs’
Where Huang distances himself from several of his peers is in what all this means for employment. He rejected the increasingly common warning that AI will hollow out white-collar work, arguing that the technology is creating roles rather than erasing them.
“The amount of work we have to do to bring AI to the world is truly incredible.” said. “So it’s creating a lot of jobs. And my software engineers love this.”
The optimism distinguishes him from Anthropic CEO Dario Amodei and Amazon boss Andy Jassy, both of whom have warned that AI could wipe out many entry-level and white-collar jobs.
Huang has consistently rejected that framework, even as other companies report that agent progress It is slower than the advertising suggests.
“This is the part of AI that people don’t realize. The first thing AI is doing right now is creating a huge number of jobs,”he said in a television interview in May.
“AI creates jobs. AI is America’s best chance to reindustrialize.”
Whether that holds across the economy is a separate question from what’s happening inside Nvidia, where the engineers Huang describes still have many employees. His claim is narrower and harder to argue: Giving better tools to trained people makes much of the monotony disappear.
For now, the proposal reaches both investors and staff. A company whose own engineers seek out agents by default is a persuasive advertisement for the hardware Nvidia sells, and Huang knows it.
The real test will be whether your enthusiasm survives contact with the more complicated parts of software work, where the judgment, debugging, and responsibility still fall to people. After all, creating an agent is its own kind of coding.
