• The 1983 report “A Nation at Risk” launched a 40-year push toward STEM that measurably worked: US high school physics-taking nearly doubled, and STEM degrees rose.

• But the US still lags peer nations. China produces nearly 10x more engineering graduates, and Germany’s STEM share far exceeds America’s.

• The AI moment doesn’t flip the script, but it does help clarify that the answer isn’t STEM or liberal arts. It’s both.

My older sister was just a few months old in April 1983 when a blue-ribbon commission delivered a warning that changed American education.

“The educational foundations of our society are presently being eroded by a rising tide of mediocrity that threatens our very future as a nation and a people,” declared A Nation at Risk, a report commissioned by the Reagan administration. 

The language was charged: “If an unfriendly foreign power had attempted to impose on America the mediocre educational performance that exists today, we might well have viewed it as an act of war.”

Within a year, they’d distributed 6 million worldwide. As government reports go, that’s a bestseller.

It explicitly invoked Sputnik, the 1957 Soviet satellite launch that had sparked the original push toward science education. “We have squandered the gains in student achievement made in the wake of the Sputnik challenge,” the report warned. The US government printing office received over 400 requests for copies in a single hour; within a year, they’d distributed 6 million worldwide. As government reports go, that’s a bestseller.

What followed was a sustained, bipartisan effort to push American students toward science, technology, engineering, and mathematics. Introduced just as my older sister graduated high school in 2001, the STEM acronym itself became ubiquitous. 

States raised graduation requirements. The standards movement reshaped curricula. The Bush administration’s divisive No Child Left Behind mandated testing. The America COMPETES Act directed billions toward STEM education. 

Forty years later, we can measure the results. And in 2026, as AI reshapes the labor market and pundits proclaim the return of “soft skills,” a fair question emerges: Did it work? And does AI change everything?

The US STEM push worked…

By the metrics that matter most to policymakers, the STEM push succeeded. More federal funding was available for primary school clubs and curriculum tied to math and science. 

The brainier of the two of us, my sister took part in many, as did American school children nationwide. By high school, she and millions others were ready for more advanced learning.

In 1990, half of American high school graduates completed Algebra II; by 2019, nearly 90% did. Precalculus participation tripled (from 13% to 40%). Physics-taking nearly doubled to more than a third. Chemistry went from fewer than half of graduates to nearly three-quarters.

At the postsecondary level, STEM bachelor’s degrees grew faster than total degrees. Using the National Center for Education Statistics definition — biological sciences, computer science, engineering, math and physical sciences — STEM’s share rose from 16% of all bachelor’s degrees in 1990-91 to 22% in 2021-22. 

In absolute terms, that’s a jump from 175,000 STEM graduates to 436,000.

The defining structural change was computing. Computer and information sciences degrees rose from 25,000 in 1990-91 to over 108,000 in 2021-22, a more than fourfold increase that tracks almost perfectly with the rise of the internet economy. 

The economic returns followed. Bureau of Labor Statistics data shows the STEM wage premium is substantial and persistent. The annual mean wage for STEM occupations in May 2023 was $108,330, compared to $58,720 for non-STEM occupations. AI chatter be damned, STEM employment is projected to grow at 8.1% from 2024 to 2034, compared to 3.1% for all occupations. 

My sister mirrors this too: She went to a prestigious private engineering school and later moved to Silicon Valley. In summer 2018, she was awarded her first patent (for the Diagnosis of Network Anomalies Using Customer Probes), and the family group chat lit up. I’ve always been proud of my sister, but gosh, I was especially impressed by this culmination of the years of hard work I witnessed. 

By these measures, the STEM push worked exactly as intended.

…But not relative to other countries 

Here’s the problem: The rest of the world was moving faster.

China now produces approximately 1.96 million engineering graduates annually, which is roughly 40% of its university degrees. The United States produces about 203,000. That’s nearly a 10-to-1 ratio, even accounting for population differences.

Germany awards 37% of its bachelor’s degrees in STEM fields, the highest rate in the OECD, a group of rich countries. South Korea and the UAE exceed 30%. The OECD average is 23.4%. The United States, after four decades of deliberate emphasis, sits at 21.6%. Below average.

This gap matters more in the context of a now-telling 1991 report from an influential Chinese political theorist, which advised: “If you want to overwhelm the Americans, surpass them in science and technology.”

Last year, in his bestselling book “Breakneck: China’s Quest to Engineer the Future,” technology analyst Dan Wang argued that US-China competition can be understood through a simple frame: “America is run by lawyers, and China is run by engineers.” The contrast is structural, not metaphorical. China, Wang writes, is an “engineering state” that brings “a sledgehammer to problems both physical and social.” America is a “lawyerly society” that brings “a gavel to block almost everything, good and bad.” 

The United States has roughly one lawyer for every 248 people — about four lawyers per 1,000 residents. China has approximately one lawyer per 2,500 people. That’s a 10-to-1 ratio in the opposite direction from engineering.

Wang traces this to America’s founding. 13 of the first 16 presidents were lawyers, he says. The Declaration of Independence, he notes, “reads like a lawsuit against the United Kingdom.” Elite law schools remain the easiest path into the top ranks of American government and business. 

The result, Wang argues, is a “litigious vetocracy” that works well for the wealthy and well-connected but struggles to build things.

China builds too much, too fast. The United States builds too little, too late.

His prescription is measured: he wants America to become “20% more engineering” and China to become “80% more lawyerly.” Neither extreme is desirable. But the current imbalance explains why America’s STEM push, however successful on its own terms, hasn’t closed the gap.

Enter AI (and the great reversal?)

Which brings us to the present moment. As generative AI tools proliferate, a counternarrative has emerged. Suddenly, the hot take is that STEM was oversold. 

Coding is being automated. Software developers are being laid off, and Technical.ly has mothballed our long-standing tech hiring events. The skills that matter now, the argument goes, are the ones AI can’t replicate: critical thinking, ethical reasoning, communication, creativity. The traditional domain of the humanities.

There’s something to this. A peer-reviewed study in Science estimated that 1 in 50 jobs could have over half their tasks affected by large language models with simple interfaces. The exposure is concentrated in cognitive and communication tasks, precisely the work that STEM graduates without broader skills may struggle to adopt.

Meanwhile, research on actual AI adoption tells a more complicated story. A large-scale Danish study found essentially no effect on earnings or hours within two years of chatbot adoption. A study of AI in customer support found productivity gains (roughly 14% on average) without displacement, with the largest benefits going to novice workers. 

The lesson from the internet era is instructive: the fastest displacement episodes — newspapers losing 27% of jobs from 2007-2010, video rental losing 62% from 2008-2011 — happened when technological substitution met macroeconomic collapse. The Great Recession accelerated what broadband enabled. Technology alone didn’t cause the job losses; the recession did.

AI’s displacement, if large, may look different: slower hiring and reduced entry-level demand rather than mass layoffs, especially under the tighter monetary conditions of 2022-2024. That’s harder to measure, harder to see, and harder to panic about.

So does AI change everything? No. The American STEM focus matters, and remains important. But it clarifies something important.

The STEM push was never about turning everyone into an engineer. It was about building a workforce that could participate in an increasingly technical economy. By that measure, it worked, even if it didn’t close the gap with China or Germany.

What the STEM push didn’t do was adequately account for the complementary skills that make technical work valuable. The research on job satisfaction and well-being is clear: “fit” matters more than field. A STEM degree without interest in STEM work correlates with lower satisfaction than a humanities degree with good job fit. Pushing students into mismatched careers isn’t a workforce strategy. It’s a recipe for churn.

As ever, our focus should be on encouraging kids to love to learn — of STEM and poetry both. My sister excelled at math, and would pore over novels too.

The answer isn’t to abandon STEM. It’s to recognize that America’s innovation edge depends on two things the STEM-vs.-humanities debate often obscures: Immigration helped accelerate our STEM savvy, and integrating hard science and soft communication is necessary.

I say this as a liberal arts graduate who has spent 17 years covering the innovation economy. The builders I’ve profiled who succeed long-term are rarely the narrowest specialists. They’re the ones who can explain what they’re building and why it matters — to investors, to policymakers, to the public.

More recently, my patent-holding, STEM-trained sister spoke at an industry conference, attempting to explain complex concepts to a broader audience. Her math couldn’t help. She charmed them instead.