What the New AI Layoff Data Actually Proves

Kind: captions Language: en The job market is being massacred. Since the beginning of 2026, over 477,000 jobs have been axed. AI is killing the livelihood of white-collar workers. Silicon Valley says this is just a temporary phase, a market correction. It's not. Industry insiders say AI is just moving the labor market, but the truth is the entry-level position is dead and Big Tech doesn't care. This is only the beginning. At the current rate, the country is set to lose more jobs than at any point in the last two decades. In 2019, before the pandemic and ChatGPT, American employers announced around almost 600,000 jobs lost across the entire year. Only 2009 had seen more, and that was in the fallout of the financial crisis. The current scale of cuts is like firing the entire working population of Atlanta, but we're being told things are looking up. The figures are terrifying, and so is the reason behind them. For decades, the firm Challenger, Gray & Christmas has tracked the reason behind every announced layoff, restructuring, cost-cutting, bankruptcy, closings. It's all there. In 2023, the firm added a new category to the list, artificial intelligence. In July 2026, AI was the most cited reason for job cuts for the sixth month in a row. By that point, AI had caused 113,000 job cut announcements. Since the category was created 3 years ago, it has ballooned to 184,000. And the cuts are pooling in one place. The tech sector has announced nearly 150,000 job cuts in 2026, a spike of 67% compared to the previous year. This is happening at the same time when every other sector is slowing job cuts. Tech now accounts for nearly a third of every announced layoff in the US. The AI cull is happening, but is it targeting specific roles? At Stanford's Digital Economy Lab, the economist Erik Brynjolfsson had spent 3 years running payroll data from ADP, the real monthly records covering roughly one in six American workers. The results showed a brutal generational divide. Employment among workers aged 22 to 25 in the most AI exposed occupations had fallen by about 16% since generative AI went mainstream. Over that same period, employment for experienced workers in the same roles stayed flat or actually grew. It's the younger workers that are taking the hit, and nowhere is that more evident than among the software developers. Since late 2022, the headcount for entry-level coders is down roughly 20%. All the while, senior developers writing code in the same repositories are finding even more job opportunities. A 23-year-old and a 43-year-old doing identical work are living in two different economies. The labor market isn't collapsing, but the entrance to it is. Brynjolfsson's researchers found the entry-level decline is driven almost entirely by a collapse in hiring rather than firing. Nobody is escorting 23-year-olds out of the building, companies simply stopped opening the door. The job was never posted in the first place. You just need to look at the graduate data to see the wreckage. In the second quarter of 2026, the Federal Reserve Bank of New York put unemployment among recent college graduates at about 5.7%. Underemployment was sitting at a massive 42%. That means four in 10 recent graduates were working in roles that did not need a degree. It is the highest score since 2020. But, for workers aged 25 and over with a bachelor's degree, unemployment sits at around 2.8%. >> [music] >> So, getting a degree is useful, just not right away. The gap between the new starts and the experienced workers keeps growing. This trend is not even confined to tech. Entry-level roles across law, consulting, and investment banking have fallen roughly 35% since 2023. PricewaterhouseCoopers said it'll hire 39% fewer audit associates by 2028. Law firm Baker McKenzie is cutting a thousand [music] roles. Hiring of paralegals at the 250 largest American law firms is down 26% since [music] 2018. The entire law industry is vulnerable to this trend. It's built on an explicit pyramid of hierarchy, and that's the only way it functions. Firms hire a large class of juniors, [music] hand them document reviews, legal research, boilerplate drafting, and the clients are billed for those hours, >> [music] >> and then the revenue is used to train the next generation of partners. AI wipes out the bottom of that pyramid. Not because it's smarter than a first-year associate, but because the tasks are mostly grunt work. AI can handle first-pass review and contract analysis faster and cheaper than any junior. Suddenly, the billable hours justify axing those juniors. And the firms took advantage of that. American law firms hired just over 24,000 associates in 2025. Only 38% came straight out of law school. The rest were lateral hires poached from rivals, billable on day one, and they required no training at all. The law industry is just [music] one of the easiest to measure examples. Across all major AI-affected industries, we can see the same trend. Companies are reporting increased profits, and yet they're hiring fewer and fewer juniors. None of this is a conspiracy. It is being announced and put on the record. In February 2026, [music] the payments company Block cut close to half of its 10,000 person workforce. Jack Dorsey told the staff [music] AI had made many of those roles unnecessary and predicted most companies would reach the same conclusion within a year. Then reports emerged that Oracle had roughly [music] 21,000 employees over 12 months. That's about 13% of its workforce. It was all part of a broad restructuring that included efforts to reduce [music] costs and increase efficiency. Amazon cut 30,000 corporate jobs over the course of 3 [music] months. The groundwork for that change had been laid a year and a half earlier when CEO Andy Jassy posted that as the company rolled out more AI agents, it would need fewer people in corporate positions. Meta did something subtler. [music] It moved about 7,000 employees into AI-focused roles while laying off another 8,000. That [music] is 15,000 more people affected by the change while on paper the reduction was only about 8,000 roles. So, where is the money going? In August 2026, [music] the Bureau of Labor Statistics published its productivity release for the second quarter. Buried in it was a number that is arguably the biggest predictor of how global capitalism works. It revealed that workers took home just 52.9% of the value they produced for the business. For scale, in the first quarter of 1947, [music] that figure was 65.8%. At the end of 2000, it was 62.8%. The August 2026 figure was a record low. In the same release, real hourly compensation, what workers [music] actually earn after inflation, fell 3.1% year-over-year. The remaining 47% isn't all profit. It includes depreciation, interest, rent, [music] and taxes. But, regardless of that, the portion ending up in the workers' hands has never been smaller. And when [music] you couple this with shrinking employment, the concentration of wealth gets even worse. Nvidia's fiscal 2026 [music] revenue came at $215.9 billion produced by about 42,000 employees. >> [music] >> That works out at about $5.14 million of revenue per person. Three years earlier, in 2023, that number sat at [music] 1 million. The company's output per employee has quintupled in the time it takes to get a bachelor's degree, but the average wage has barely moved. And their total workforce only grew by 50%. At the height of American industrial power, a Ford worker on the assembly line was producing a fraction of a percent of $5 million in inflation-adjusted terms. The Industrial Revolution made muscle more productive, but it still needed the muscle. AI does something different. It makes the output climb while the number of people attached to [music] it barely moves. In March 2026, Brett Falk of the University of Pennsylvania and Gerry Sukalac of Boston University published a paper titled The AI Layoff Trap. It is a theoretical model that asks, "If AI replaces the workforce, who is left to buy anything?" Let's just say the economy is one company, a monopoly that makes one product. Everyone works there. Then AI arrives. The CEO could automate 90% of the staff, but those workers are also the consumers, the ones who buy the product. Fire [music] them and demand disappears. So, the rational monopolist automates carefully, stopping well short of the cliff. But the moment there's competition, things change. Suddenly, the workers are only a sliver of the customer base. Automate the staff and the CEO pockets 100% of the cost savings. The economic damage gets spread across [music] every company in the market, including your rivals. The share of the damage is close to zero. But then the competition notices and decides [music] to automate. It's the only strategy. No matter what anyone else does, the best move is to cut. Every CEO is acting rationally, yet everyone is worse off. Falk and Sukalac found that the harm landed both on the workers and the owners of the firms doing the automating. Crucially, more competition makes it worse. [music] Better AI makes it worse still. The economic demand being destroyed is not evenly distributed. It's being stripped from Gen Z, the exact group of professionals who are supposed to spend the next 40 years earning money and driving the economy. That is the broken loop at the heart of the AI boom. The industry's pitch is that AI is a product ordinary professionals will buy, a subscription, a co-pilot, a tool that makes a knowledge worker more valuable. But the same technology is being deployed [music] to remove the entry-level knowledge jobs those buyers were supposed to be hired into. A 23-year-old who can't get a junior analyst position is not going to pay 200 bucks a month for an AI tool. They're more concerned about paying their utility bill, their rent, their mortgage, or their lunch. You can't simultaneously sell the world a productivity tool for professionals, and eliminate the same pipeline that produces professionals. That's not just a hypothesis. We're already seeing it. Moody's Analytics estimates that the top 20% of earners, households that earn over $175,000, account for almost 60% of all personal spending. The top 10% alone are responsible for almost half of the consumer economy. The highest share recorded since the series began in 1989. For the bottom 80%, it only grew by a measly 2.6% falling to match inflation. It's what economists call a K-shaped economy. >> [music] >> Massive wealth is shooting straight up like the top arm of a K, while the rest of the country is plunging downward. Some institutions like the Minneapolis Fed disagree. They argue the economic split is less dramatic than people assume. The keyword is less dramatic, not non-existent. It implies the trend is still happening. The entire economy is leaning heavily on the people at the top while thinning out at the middle class. The problem with that is that the middle is the engine that drives the economy in the first place. Kill the middle class and everything comes to a shuddering stop. But there's a serious argument that says all of this is wrong. Stanford's Institute for Economic Policy Research published a brief separating AI hype from reality. The paper showed hiring in AI exposed occupations began to fall after the Federal Reserve's rate hikes, and even before Chat GPT launched. It was the pandemic shift to remote work that eroded the value of hiring junior staff. On-the-job learning was harder to deliver over video. The New York Fed went further saying almost 64% of young graduate unemployment could be attributed to remote work rather than to AI. This is where Stanford economists and the Fed agree. Brynjolfsson and his co-authors found that when they added the broadest possible set of controls, the declines in AI exposed occupations became statistically significant from 2024 onward. The earlier drop was probably driven by something else. So the skeptics make a case that the timing is muddy. Interest rates and remote work are possible culprits. The aggregate labor market is the way it is because of factors unrelated to AI. But, there's a bigger problem here. Even if that is true, it doesn't really matter. Every one of those objections is about causation, whether AI is the reason the entry rung has disappeared. But, none of them dispute it disappeared. The hiring decisions have already been made. A firm that spent 2024 and 25 not hiring juniors still has a gap that it isn't filling, and it's part of the problem. If all firms agree to not hire juniors, associates won't exist. And if no one's hiring juniors, the fewer juniors there are to promote. Even if economists conclude in 2029 that rates and remote work were to blame, the damage has already been done. An entire generation of entry-level workers will have been wiped from the corporate ladder. Improving AI widens the gap [music] between what companies do and what would be collectively optimal. The IMF has been saying a version for a number of years. Around 60% of jobs in advanced [music] economies are exposed to AI. About half of those are likely to benefit and half could be displaced. Falk and Sukalac also tested the standard menu of fixes against their model. Universal [music] basic income, giving workers equality, capital income taxes, retraining and upskilling. And in their model, every one of them failed because none touched the incentive to replace a person with a machine. They just treat the aftermath. The one instrument that worked in the model was a tax on automation itself. It was costly to fully replace a worker, but cheap to augment one. The authors conceded that the word tax made it politically radioactive, so they offered a [music] softened version, subsidize the firms that keep people. But, both of those still relied on the government to step in. That same government is the one where companies have the most sway thanks to lobbying. 2025 saw over $5 billion spent on federal lobbying, and AI is a recent newcomer. In 2022, Anthropic, Nvidia, and OpenAI had no federal lobbyists at all. Today, these companies are spending millions all on preemptive federal regulations and frameworks. That's the main problem with how the AI layoff trap works. [music] The first firm to slow down loses to the ones that don't. While the economists are arguing about causation, and they will be for years, the same question remains. If AI does all the junior work, how does anyone ever get the experience [music] to become senior? AI is continuing to threaten the middle class, and it's Gen Z that's feeling the brunt of it. Find out how bad it is in how AI is causing a white-collar purge, or click on this.