Nvidia’s VP of Applied Deep Learning Bryan Catanzaro stated publicly that the cost of compute at his company now exceeds the cost of employees. Uber’s CTO told The Information that the company burned through its entire 2026 AI budget on token costs before the year reached its midpoint.
The entire justification workers have been given for their displacement is that AI is cheaper and more efficient than people.The workers who lost jobs to that cost assumption have no mechanism for recourse while the economics get worked out.
Amazon made the most consequential single announcement of the week for middle-American workers. The company launched Connect Talent, an agentic AI system that conducts voice job interviews around the clock, scores candidates on competency, and prepares recruiter notes entirely without human intervention. Amazon built the system to handle 250,000 seasonal hires.
It is now selling that system to retailers, manufacturers, logistics operators, and hospitality companies. The human job interview is now an optional feature. The human recruiter is the first professional role Amazon has automated and packaged for sale at scale to the industries where most middle-American service workers are employed.
Also listen on Apple Podcasts
The macro picture around those two stories is built on a set of numbers that do not fit together easily. The BEA reported Q1 2026 GDP growth of 2.0%. Jobless claims dropped to 189,000, the lowest level since 1969. Corporate earnings across the Magnificent Seven were strong.
Also, the same quarter produced the largest single-week cluster of AI-attributed layoff announcements in corporate history, with 92,000 tech workers cut so far in 2026 and nearly half of those cuts attributed to AI.
The economy is growing. Workers are being cut. The headline statistics look healthy because the workers absorbing the displacement are contractors, gig workers, and early-exit buyout takers — populations the unemployment system was never designed to count.
Get the eBook that exposes the Narratives Tech uses to build its AI Empire. $4.95 SPECIAL SALE PRICE for PDF or ePUB formats; no subscription required. 3.5-hr reading time.
The people building artificial intelligence did not invent their ideas. They inherited them.
The Citrini Research bit of speculative fiction that moved markets in February called this “Ghost GDP”: growth that shows up in national accounts but never circulates through the people who used to earn it. This week’s data made that speculative scenario visible in actual earnings filings and government statistics, simultaneously.
Meanwhile, Anthropic’s March 2026 labor market research measured actual AI task coverage in real workplace deployments. Computer programmers show 74.5% observed task coverage. Customer service representatives are at 70.1%. The most exposed workers are more likely to be female, over 40, more educated, and better compensated.
The entire “get a degree and move into knowledge work” strategy a generation of middle-American workers followed is contradicted by the actual usage data from the largest AI deployment in the world.
Regardless, workers using AI daily are producing more output and losing underlying competency. A January 2025 controlled study found that participants who received AI assistance during training performed substantially worse on subsequent independent assessments.
Senior workers who built expertise before AI became ubiquitous can use AI as an augmentation layer because they understand their domain well enough to evaluate its output.
Get the eBook that examines how the AI invasion already happened. You just weren’t invited. $9.95 flat fee for PDF or ePUB; no subscription required. 2-hr reading time.
Junior workers building careers on top of AI from day one are building on a foundation they have never been required to develop independently.
What the week’s stories mean together
Three trends are now running simultaneously and reinforcing each other: The first is that AI costs more than projected, organizational cuts are made in anticipation of savings that have not yet arrived, and workers bear the cost of that gap.
Further, the measurement architecture — unemployment claims, GDP, monthly jobs reports — was built for a workforce that no longer describes the majority of American workers. The contractor economy absorbs displacement in silence.
Future Forwarded is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
And then, the workers most at risk are precisely the workers who followed the advice they were given: get educated, move into knowledge work, build credentials.
More practical advise would be: do what AI cannot do for at least a generation into the future.
Get full access to Future Forwarded at futureforwarded.substack.com/subscribe
