Is AI Upending Tech Jobs First? Is STEM Facing Unemployment?

ProTip: You were told to follow the data. Now Data has a layoff notice.

Focus on STEM has long been the gospel of career advice. Teachers, with their own easy-peasy education degrees, parroted the message. STEM was the safe career bet. Learn to code, too. This wasn’t just guidance; it was doctrine. Major in something useful. And by “useful,” they meant: Never Liberal Arts.

But just like housing, just like college itself, just like every bubble in American life before it—there’s a limit to how many people can crowd through the same door labeled “future-proof.”

There are more STEM grads than there are STEM jobs. In fact, the U.S. Census Bureau flat-out says it: 74% of STEM bachelor’s grads aren’t even working in STEM. So where are they? Selling SaaS? Waiting tables? Running the few help desks that haven’t been outsourced yet?

Worse still, unemployment for recent grads in computer engineering and physics is higher than the national average for all young workers—college or not. So much for STEM being the “sure bet.” It’s now just another overpriced raffle ticket with massive student debt.

Ron Hira makes the provocative claim that for decades, there’s been no real, lasting shortage of STEM workers. Employment projections have been wildly off for years, yet they still get cited like gospel. It's a policy mirage that continues to produce real debt and false hope.

Yet now, as the old Ronco ads used to say, But Wait There’s More.

Let’s say you did everything right. You majored in comp sci. You got the GitHub repo, the coding bootcamp certificate, the fridge full of Soylent. You’re ready to launch.

Enter AI.

And not just the vague “AI might take your job someday” kind of warning. Try this: the CEO of Anthropic recently told the Council on Foreign Relations that by the end of this summer, AI could be writing 90% of all new code. Meta is already scrapping mid-level coders. Google says AI writes over a quarter of its new code. Stripe is cutting engineers even as it grows its headcount elsewhere. Salesforce saw a 30% productivity boost from AI and responded by laying off 1,000 workers.

You can hear the narrative wheeze: “No, no—AI won’t replace you. It’ll make you more productive.” Sure. Until someone decides they don’t need 20% more productivity. They need 20% fewer salaries.

One recent headline laid it out plainly: “Heads up students, computer science is now among the top ten majors with the highest unemployment.” That came out the same week Mark Zuckerberg promised an “AI engineer to write code.” Meanwhile, the Federal Reserve Bank quietly updated its employment data for recent STEM grads: joblessness up, wages flat.

So what kinds of jobs are actually AI-safe?

That short list is shrinking fast—but it’s still there. Jobs that involve high levels of human interaction, creativity, complex problem-solving, and adaptability are generally considered less likely to be steamrolled by automation. In other words, the things humans do best when we’re not pretending to be machines. Emotional intelligence, critical thinking, unpredictability—basically all the soft skills that don’t show up on a resume but determine whether someone gets invited back to the room.

Take healthcare. Doctors, nurses, therapists—these aren’t just technical roles. They require empathy, trust, reading between the lines of a conversation and a lab result. Not easy for an algorithm.

Or skilled trades. The guy fixing your furnace when it dies at 2 a.m. isn’t being replaced by ChatGPT anytime soon. Plumbers, electricians, mechanics—these are people who solve real-world problems in non-standard conditions. No two water heaters are ever quite the same.

Ironically, some of these AI-resistant fields still demand strong STEM foundations—especially in healthcare. But many don’t. What they do demand, however, are the very skills we’ve been devaluing: empathy, communication, and improvisation. In other words, the stuff you learn not by memorizing formulas, but by getting kicked off the monkey bars and learning how to apologize.

We told kids to learn calculus. Turns out they also needed the strong interpersonal skills learned on a playground at recess that teach people how to get along.

After telling a generation to follow the data, now Data has a layoff notice.