That deal just collapsed. The industry’s own numbers killed it.
Robert Reich’s new column connects two facts AI companies have worked hard to keep apart. Workers’ share of US GDP dropped to 52.9% in mid-2026. That’s the lowest figure since tracking began in 1947. Meanwhile, data center construction keeps accelerating, funded by the same firms telling Congress that AI will eventually create more jobs than it destroys.
Reich reads this as a political opening — proof that populist anger finally has one target Bernie Sanders and Steve Bannon can both stand behind. That’s a real story. It’s not the most useful one for anyone watching the AI sector itself.
A Prediction That Came True, Right on Schedule
In May 2025, Anthropic CEO Dario Amodei warned publicly that AI could eliminate half of entry-level white-collar jobs. He said unemployment could hit 10 to 20%.
That warning was meant to signal responsibility. A company being honest about its own disruption, ahead of policy solutions that would presumably follow.
Sixteen months later, the disruption arrived on schedule. The policy response didn’t. For the workers actually navigating that shift, the practical question isn’t whether the warning was accurate — it’s what skills still hold value on the other side of it, a question this piece on staying relevant as AI reshapes hiring digs into directly.
That’s the real tipping point in this story. Not that AI is hurting workers — that was already priced in. It’s that the industry’s credibility window has closed. A prediction that comes true without a corresponding fix stops looking like foresight. It starts looking like a warning nobody planned to act on.
Trust Is Falling Faster Than Expected
New Bentley-Gallup research found that 39% of Americans now think AI does more harm than good. That’s up from 31% a year ago. Fewer than one in ten think the reverse.
Among 18-to-29-year-olds, distrust in how businesses deploy AI jumped from 29% to 41% in the same stretch.
Compare that to past tech backlashes. Public skepticism toward social media took years to build. It needed specific scandals — Cambridge Analytica, congressional hearings, whistleblowers — to move opinion.
AI’s trust collapse needed none of that. People can watch the wage-share number and the data-center number move in opposite directions without a journalist connecting the dots for them. Part of the problem is which numbers get repeated versus which get buried — a gap this rundown of AI statistics and their missing context lays out well.
Call it ambient distrust. It’s erosion driven by macro data lining up with lived experience, not by one disclosed failure. It’s harder to manage than scandal-driven distrust. There’s no single narrative to correct. No apology fixes it.
Data Centers Are Where the Argument Gets Physical
About 70% of the public opposes data center construction near them. Most coverage treats this as a NIMBY story about power grids and water use.
Read it differently. This is the first place where “AI progress” and “AI cost” become visible to people who don’t work in tech.
A layoff blamed on AI is debatable. Was it really the AI, or a bad economy, or bad management? A data center’s water and power draw isn’t debatable in the same way. It shows up on a utility bill. Anthropic’s own $50 billion data center buildout makes the point directly — 800 permanent jobs promised, against infrastructure demands that reshape entire regional grids.
That’s likely why data center opposition is outpacing broader AI skepticism. It’s the one part of the AI economy that resists company messaging, because the evidence sits in people’s own infrastructure.
What This Means for Anyone Building AI Products
The takeaway isn’t “public opinion turned, so AI is doomed.” It’s that the industry’s go-to PR move — emphasize future benefits, downplay present costs — now works against companies instead of for them.
Every capability announcement lands next to a wage-share chart moving the wrong direction. That gap between pitch and data is generating the populist energy Reich describes. The technology itself isn’t the driver.
Reich frames the open question as political: will Democrats seize this moment? The more consequential question might be simpler. Can AI companies find a message that survives contact with their own economic data?
Right now, none of them have one.
Related: Jensen Huang Says AI Extinction Risk Is 0%. Can Anyone Defend That Number?
