Huang argues AI will transform engineering roles, not eliminate them—but a timeline for junior developers remains unclear

The engineering purpose persists#

Nvidia CEO Jensen Huang rejects the prevailing interpretation of predictions that AI will write 90% of software code within a short timeframe—namely, that engineers as a profession will become obsolete. Speaking with The New York Times' Ezra Klein at Nvidia's headquarters, Huang draws a distinction between the core purpose of engineering work and the specific tasks that make up that work today.

He argues that the fundamental role of software engineers—inventing products, solving problems, and connecting technology to social needs—will outlast the automation of coding itself. His analogy draws from radiology, where AI has automated the reading of scans without eliminating the radiologist's diagnostic function. Similarly, he suggests that as coding becomes automated, the essential engineering challenge will remain.

"The purpose of the software engineer is engineering," he states in the interview. "There was engineering before software. There will be engineering after software programming."

The AI-native class of 2028#

When pressed on how this shapes hiring for junior developers, Huang offers a straightforward answer: wait two years. His math hinges on the typical duration of undergraduate computer science programs. If capable coding agents are now emerging and undergraduates take four years to complete their degrees, the first cohort of graduates trained to collaborate with AI systems should enter the workforce around 2028. Huang expects these AI-native engineers to arrive with advantages their predecessors lack, having learned their discipline alongside agentic tools from day one.

Evidence for this optimism comes from Huang's observation that recent PhD and master's graduates in computer science are starting companies at notable rates. He compares the current moment to previous technological inflection points—calculators and personal computers—that transitioned from optional to mandatory skills. Students, he predicts, will soon be unable to graduate without learning to work alongside AI systems.

The skills fade problem#

The shift carries a cost that Huang acknowledges but downplays. Research into 26,000 Chinese students found that AI adoption improved homework scores by 18% while reducing monthly exam performance by 20% within six months, suggesting that certain intellectual capabilities atrophy when automation absorbs the foundational work. Huang accepts this trade-off, arguing that .