Tuesday morning, the Bureau of Labor Statistics dropped a bombshell that should terrify economists and workers alike: They hallucinated nearly one million jobs that never actually existed between March 2024 and March this year. While some revisions are normal, this is a record-setting high in either direction. For the last eighteen months, we’ve been making policy decisions based on phantom positions that existed only in government spreadsheets.
Revisions to estimated job creation statistics are standard procedure – the BLS always adjusts preliminary estimates when more complete data arrives. But the scale of what’s been happening lately is unprecedented: May’s numbers were slashed by 125,000 and June’s by 133,000, a combined quarter of a million phantom jobs on top of yesterday’s numbers, that simply never existed.
Any doubt that the Federal Reserve will begin cutting interest rates has all but vanished in light of these revisions. This is their go-to playbook for recessionary environments: Cutting interest rates means that individuals and companies can borrow more freely, which stimulates economic expansion – historically necessitating hiring, which boosts employment and stimulates spending, and so on.
Or that’s been the theory historically, anyway. That link is now broken. Companies can now scale operations, write code, and expand their operations without adding a single employee – AI and automation have begun to sever the relationship between growth and employment. Lower rates will continue to stimulate investment, but that investment can now flow freely into automation and AI systems rather than creating new jobs.
Don’t believe me? CEOs of the largest AI companies are openly admitting this is the case: Dario Amodei, head of Anthropic, warns that AI could eliminate 50% of entry-level white collar jobs within three years, potentially spiking unemployment to 20%. In an economy where even a 2 percentage point increase historically triggers recession, we’re staring at economic devastation that would make 2008 look mild.
Honestly, he’s likely understating the scale of this issue given the extreme reliance on consumer spending that undergirds the U.S. economy. Personal consumption makes up 68% of our GDP and directly contributed to a third of last quarter’s growth alone — most of our expansion. With credit card delinquencies rising to 12.3%, the highest since 2011, and auto repossessions hitting 1.73 million in 2024 – the highest since 2009 – consumers are clearly already breaking. Pullbacks in employment from where we are today is only going to worsen this effect: Fewer people working with less income will mean less spending, which will ripple throughout the economy.
This isn’t something that’s coming years or even months down the pipeline. It’s something that’s here, and has been for some time. Google’s code is already written 25% by artificial intelligence; Microsoft cut 15,000 workers this year despite posting record profits; and Meta eliminated 5% of its workforce after admitting AI does mid-level engineering work. Salesforce announced they aren’t hiring any more engineers because AI handles half of its development, period.
The most dangerous delusion here is not false jobs numbers – it’s thinking that AI’s current limitations give us breathing room or security. Today’s AI that confidently presents misinformation or outright fabricates data and terrible, stilted prose is the worst it will ever be again.
Two years ago, ChatGPT was barely usable – now, it’s replacing broad swaths of entry level (and, honestly, mid-level) professionals. These systems improve exponentially, with capabilities improving at dramatic rates.
And yet despite all of this being so clear, our political class has refused to even acknowledge this crisis publicly. While the federal government has taken an interest in ensuring we do not fall behind China in the AI arms race, absolutely no attention has been paid to what happens to a consumption-based economy when millions lose their income.
While there may be no clear solutions, the complete absence of public discourse has real-world implications for a generation. College application season is approaching, and millions of 17-year-olds are about to sign up for tens or hundreds of thousands of dollars in debt for degrees their schools and parents insist are critical to success – computer science, engineering, even law.
How many of these jobs are going to be automated by the time they enter their freshman year in fall 2026, let alone when they graduate in spring 2030?
The Fed will cut rates, companies will use that cheaper capital to accelerate automation, and employment will continue its long-term decline. Washington will scramble when their tools fail, yet we seem poised to live through a period in which universities keep loading students with debt for careers that no longer exist. With the S&P at record highs, markets seem equally blind to the uncomfortable reality: Machines don’t need paychecks, but they also don’t buy iPhones or vacations or homes. We discovered yesterday that we hallucinated a million jobs. Soon we’ll discover the real cost of an economy built on consumer spending when the consumers can no longer afford to consume.
