The New Talking Point: AI Builds, It Doesn't Destroy

NVIDIA CEO Jensen Huang has been making the rounds, and his message is consistent: stop panicking about AI taking your job.

At a World Economic Forum session alongside BlackRock CEO Larry Fink, Huang called AI a "five-layer cake" and described it as "the largest infrastructure buildout in human history," according to NVIDIA's own blog. On the Special Competitive Studies Project podcast "Memos to the President," he pushed the same line.

His headline number, quoted by The National: "AI has created more than half a million jobs in the last couple of years."

Fortune got a blunter version: "It is unlikely most people will lose a job to AI," Huang said.

NOTE: The phrase "Return on Intelligence" — a framing that circulated alongside Huang's comments — was coined by Brian Solis, ServiceNow's Head of Global Innovation, as his own analytical lens for Huang's argument. Solis introduced it, not Huang. An earlier version of the Let's Data Science report misattributed it. They corrected the record on May 9, 2026.

Meanwhile, the Layoffs Are Real

The same week Huang's optimism was making headlines, the actual job market was doing something different.

Pleasanton-based Workday announced it would cut 1,700 jobs and pivot to hiring AI-related talent, according to NBC Bay Area. Salesforce of San Francisco said it would eliminate 1,000 positions — but add 2,000 new AI-related jobs, per NBC Bay Area's Scott Budman.

In total, NBC Bay Area reported that critics blamed AI for nearly $1 trillion in stock market losses and close to 3,000 job losses within just one week.

Those are real people cleaning out their desks.

What the Pro-AI Crowd Gets Right

HR and tech analyst Josh Bersin published research in March 2026 pushing back hard on job-loss alarmism. His core argument: AI eliminates tasks, not jobs.

Bersin points to Lightcast and Draup labor market data showing that global software engineering job postings have remained largely flat despite years of AI advancement. OpenAI alone currently has 650 software engineering-related openings, according to Bersin.

What IS shifting, per Draup's analysis: entry-level coding and testing roles are slowly declining. But Bersin frames this the same way history frames the end of the steno pool in the 1970s or the office secretary answering phones in the 2000s. The job evolves. It doesn't disappear.

His emerging category: "Superjobs" — roles like "full stack AI engineer" that combine skills no single job title covered five years ago.

Joint Venture Silicon Valley CEO Russell Hancock put it plainly to NBC Bay Area: "You don't just wave your hand at AI. You have to create it, construct it, build it, insert it, implement it, and iterate on it."

What the Pro-AI Crowd Gets Wrong

The Huang-Bersin optimism is almost entirely focused on tech-sector jobs.

Software engineers transitioning from writing code to architecting AI systems? Fine. Those people have options, and the data supports that.

The 1,700 Workday employees who just got their walking papers aren't mostly engineers pivoting to AI roles. Neither are the 1,000 Salesforce workers out the door. These are operations staff, support roles, mid-level managers — people who don't have "full stack AI engineer" in their skill set and can't acquire it in 90 days.

The NYT ran a piece titled "A.I. Is a Job Creator." While that framing isn't wrong in aggregate, it obscures a critical distinction: the people getting the new jobs are not the same people losing the old ones. The skill gap between those two groups is enormous.

The Math Behind the Headlines

Google announced it would invest $75 billion in AI in 2025, according to NBC Bay Area. NVIDIA is printing money. Salesforce says it's hiring 2,000 people. OpenAI has 650 openings.

The people getting those 2,000 Salesforce AI jobs are NOT the 1,000 people Salesforce just laid off. The skill gap is enormous — and nobody in power is being honest about the retraining cost, the timeline, or who pays for it.

When a CEO says "AI creates jobs" at Davos, he's talking about his company's hiring pipeline. He is NOT talking about the 52-year-old customer support manager in Tulsa who doesn't know what MLOps means.

The Incomplete Picture

For workers in tech with adaptable skills, the data suggests the jobs are shifting, not vanishing.

For workers in routine, task-heavy roles anywhere in corporate America, the optimism coming from Jensen Huang's podium was not directed at them.

The "AI creates jobs" headline is technically defensible. It is also incomplete. The workers bearing the cost of this transition deserve more than a TED Talk about the future of work.