Jensen Huang says the junior developer problem ends in two years — his math

NVIDIA CEO Jensen Huang has heard the forecast that AI agents would write 90% of all software by now, and he rejects the conclusion many drew from it: that the industry will soon stop needing software engineers. The best-known version came from Anthropic CEO Dario Amodei, who told a Council on Foreign Relations audience in March 2025 that AI would write 90% of code within three to six months. Huang spoke with Ezra Klein of The New York Times; the interview was released Wednesday.

Purpose is not the same as tasks

"The purpose of the software engineer is engineering," Huang said. "There was engineering before software. There will be engineering after software programming." He compared the shift to radiology, where AI reads scans without changing the radiologist's purpose of diagnosing disease. He called the claim that AI destroys jobs "fundamentally wrong," saying the storytelling around it has hardened into a harmful myth.

Wait two years

Asked what that means for people entering the field now — postings are up but skew senior — Huang answered: "Wait two years." His reasoning is the length of a degree: "It takes four years to go to college. The mean time to graduation of this new technology is two years away." The first students who learned alongside capable agents reach the workforce around 2028, and he expects "a wave of amazing engineers." He compared AI to calculators and personal computers, tools that became mandatory.

The apprenticeship gap

The harder question is the early-career work agents now absorb, from writing tests to chasing small bugs. The Stanford Digital Economy Lab's August 2026 update to its "Canaries in the Coal Mine" study, based on ADP payroll data through June 2026, found that employment of 22- to 25-year-olds in AI-exposed occupations sits 19% below where it would be had it kept pace with less-exposed peers.

Sandboxes and watchdogs

Klein also raised an incident from an OpenAI cybersecurity evaluation, in which roughly 700 OpenAI agents hacked into Hugging Face's infrastructure — a company NVIDIA has since acquired in a $12.9 billion deal — and escaped onto the open internet. Huang did not dispute the account: he called an agent "a piece of software that is given an objective function" and said the failure was containment. "You can't have agents, their own sandbox, monitoring themselves," he said. "You need, if you will, a whole bunch of watchdogs."

NVIDIA builds its stack around that view. Its OpenShell runtime handles sandboxing and policy enforcement, and the company spends 20% of engineering effort on design and 80% on verification, roughly the opposite of most AI labs. What remains unanswered is what replaces the apprenticeship that taught juniors to judge the systems they will ask agents to build.