July 25, 2026, (Inside AI) — The long-feared AI jobs apocalypse is not arriving on schedule. A new analysis from Anthropic, the maker of chatbot Claude, finds no systematic rise in unemployment for workers highly exposed to AI since late 2022, despite bold predictions from its own co-founder that half of all entry-level jobs could vanish within one to five years.
Anthropic’s report reveals a stark gap between capability and deployment. While AI could theoretically handle nearly 100% of tasks in computer and math occupations, actual usage covers just 33%. “We find no systematic increase in unemployment for highly exposed workers since late 2022,” the report states, noting that deployment “remains a fraction of what’s feasible.”
This disconnect is fueling a broader reassessment. Productivity growth has been sluggish in the AI era, slower than during the 1990s IT boom, even as datacenter spending surges. OpenAI CEO Sam Altman recently walked back doomsday scenarios. “I don’t think we’re going to have the kind of jobs apocalypse that some of the companies in our space advocate or talk about,” he said in May.
The tech-heavy Nasdaq, driven largely by AI hype, has dropped about 8% since its June peak, reflecting cooling expectations. Nobel laureate economist Daron Acemoglu points to unsustainable economics: “They are losing hundreds of billions of dollars every year.”
Read: Entry-Level Employment Falls 19% in Most AI-Exposed Jobs, Stanford Study Finds
Why the O-Ring Theory Deflates Automation Panic
The Challenger disaster offers a powerful counter-narrative. In 1986, a failed rubber O-ring destroyed a multibillion-dollar spacecraft. Applied to AI, the analogy suggests that as long as machines cannot perform every task perfectly, the remaining human tasks gain value. This can boost demand for high-skill workers freed from routine work or elevate lower-skilled workers when AI takes on expert roles.
A recent study supports this view: “despite strong substitution at the task level, overall employment effects are modest, as reduced demand in exposed occupations is offset by productivity-driven increases in labor demand at AI-adopting firms.” In other words, firms using AI grow and hire more, offsetting job losses.
MIT economist David Autor, a leading voice on technology and work, notes that AI still cannot connect language to reality. “Not everything is a computational problem,” he said. “A lot of people have noticed that the world is not changing as fast as they predicted.”
Insiders Bet on AGI Even as Economics Buckle
Despite the tempered near-term outlook, AI insiders remain committed to a transformative future. Elon Musk still envisions a world where “AI+Robots will be able to do everything, resulting in universal high income. Work will be optional.” Daron Acemoglu observes, “Insiders are as gung ho as ever. They still believe artificial general intelligence is around the corner.”
Yet the path is littered with obstacles. Public opposition is hardening: seven in 10 Americans oppose building AI datacenters locally, driven by energy concerns. The International Energy Agency projects datacenter power demand will more than double by 2030 to about 945 terawatt-hours, exceeding Japan’s total consumption. And the breakneck pace of model obsolescence means AI investments depreciate rapidly, raising doubts about profitability.
Read: AI Transformation Requires Redesigning Work, Not Cutting Roles
Anthropic’s Dario Amodei still has nearly four years left in his prediction window, and the Federal Reserve notes fast-rising AI adoption across businesses. History offers a cautionary parallel: in the early computer age, economist Robert Solow famously quipped that you could see the computer age everywhere but in the productivity statistics. It took a decade for the gains to materialize. AI may yet follow that arc—if its economics and politics allow.
