← All articles

💼 Labor & AI

Announced AI Cuts Need 103 Years to Reach 15 Million

Employers cited AI for 121,401 cuts in ten months. At 145,681 a year, Goldman's 15 million takes 103 years.

An empty open-plan office at dusk, one desk lamp lit among dark monitors, with a city skyline beyond the windows.

One hundred and three years.

That is how long it takes to reach the 15 million workers Joseph Briggs named, if every future case shows up as an announced cut and the reason line keeps saying artificial intelligence, at the pace employers actually published through October.

Challenger, Gray & Christmas counted 121,401 of those announcements in ten months, inside 617,189 cuts of every stated reason, and 121,401 divided by 617,189 is 19.67 percent, beside the firm's "approximately 20 percent." Multiply by 12/10 and the pace is 145,681.2 announcements a year, so 15,000,000 divided by 145,681.2 is 102.96 years, which rounds to the 103 in the headline.

Briggs said it on 2 July 2026, on Goldman Sachs Exchanges, not in a November paper: under a 15 percent productivity uplift, "around nine percent of all workers in the US will be reallocated," and "nine percent of workers being displaced by AI would correspond to 15 million workers." Divide 15,000,000 by 0.09 and the implied base is 166,666,667 people, while Table A-1 puts October employment at 163,615,000, which is 3,051,667 lower, so 9 percent of that count is 14,725,350 and the same pace runs 101.1 years.

October is a different month. AI was cited for 1,265 cuts, the year's lowest monthly AI total, and 1,265 over 43,994 is 2.875 percent. Market and economic conditions were 16,407, or 12.97 times the AI line, and for the year they lead by 9,130, with 130,531 against 121,401. Andy Challenger called this "a low-fire environment," with announced hiring "largely short-term, seasonal work." The stated reasons shifted.

Cut counts are Challenger, Gray & Christmas, October 2026 report, released 5 November 2026. The 15 million is Joseph Briggs on Goldman Sachs Exchanges, 2 July 2026. Employed is BLS Table A-1, seasonally adjusted, October 2026. Annualizing multiplies ten months by 12/10 and does not seasonally adjust the announcement series.
WorksheetArithmeticResult
AI share, January-October121,401 / 617,18919.67%
AI share, October only1,265 / 43,9942.875%
Annualized AI-cited announcements121,401 × 12/10145,681.2
Years to Briggs's 15 million15,000,000 / 145,681.2102.96 years
Nine percent of October employed0.09 × 163,615,00014,725,350
Same pace, October base14,725,350 / 145,681.2101.1 years
Even 10-year slice, these 10 months121,401 / 1,250,0009.7% of the slice

Spread 15 million evenly over ten years and a year is 1.5 million people, so these ten months are a 1,250,000 slice and the announcements cover 9.7 percent of it. Since 2023, when AI became its own Challenger reason, the series holds 193,226 cuts, and 2026's 121,401 are 62.8 percent of that history, crammed into ten months: a burst beside a stock it does not fill.

October hiring plans were 444,692 and the year's are 655,304, of which 470,400 are seasonal, 71.8 percent, from five companies rather than eleven a year earlier. Amazon's 300,000 seasonal plans, 50,000 above the 250,000 of each of the prior three years, are nearly all of October's 300,252 technology hiring plans, leaving 252, and yearly technology plans of 322,613 leave 22,613 once that same 300,000 is removed. Technology cuts of 171,501 are 7.58 times that remainder, because Challenger files Amazon's warehouse seasons under technology.

Microsoft offered voluntary exits to 60,000 people in October, and Challenger leaves a voluntary program out until someone accepts or a target is stated, so those 60,000 never entered the 43,994. Divide 60,000 by 43,994 and the excluded pool is 1.36 times every cut counted that month, which means an offer is not a layoff and is also not zero.

Payrolls, unemployment, earnings, and part-time work are the 6 November Employment Situation. Household levels are Table A-1. U-6 is Table A-15. Claims are the Employment and Training Administration release for the week ending 7 November. JOLTS levels are the September release and Table A, which is one month earlier than the payroll report.
MeasureFigureWhat it is not
Nonfarm payrolls, October+165,000Not the household survey's +463,000 employed
Unemployment rate4.1% (7,068,000 people)Not U-6, which is 7.9%
Part time for economic reasons+339,000, to 4.8 millionNot a payroll-survey job loss
Initial claims, week of 7 November200,000; four-week average 195,250Not the year's 121,401 AI announcements
September layoffs and discharges1.6 million, rate 1.0%Not Challenger's 43,994 October announcements
September quits3.4 million2.13 times that month's layoffs

On 6 November the Bureau of Labor Statistics said payrolls rose 165,000 and unemployment "changed little" at 4.1 percent, and Table A-1's count is 7,068,000 unemployed in a labor force of 170,683,000, so 7,068 over 170,683 is 4.141 percent, the printed 4.1. Long-term unemployment "changed little" at 2.0 million and 27.6 percent of the unemployed, which is 0.276 times 7,068,000, or 1,950,768, a rounding that does not license dividing 2.0 by 7.1.

Part-time for economic reasons rose 339,000, to 4.8 million, which is 2.05 times the payroll gain and a household measure rather than a payroll loss. Table A-15 shows U-6 up from 7.6 percent to 7.9 while U-3 fell from 4.2 to 4.1, widening the gap from 3.4 points to 3.8, and household employment rose 463,000 against a payroll gain of 165,000, a gap of 298,000.

A dent, not the month. Briggs's July drag of about 10,000 to 15,000 jobs a month, in tech, consulting, and graphic design, is 6.1 to 9.1 percent of October's net 165,000, while professional and business services added 70,000 and health care 42,000, and information "showed little change."

Initial claims for the week ending 7 November were 200,000, up 9,000 from a revised 191,000, with the four-week average at 195,250. A year earlier that week was 228,000, so the print is 12.3 percent lower, 195,250 against a year-earlier four-week average of 227,000 is 14.0 percent lower, and insured unemployment of 1,697,000 against 1,946,000 is down 12.8 percent. Fortune says Goldman called that four-week average the lowest in the series' history, a ranking that belongs to Goldman, while the levels belong to the release.

September JOLTS put layoffs and discharges at 1.6 million, a 1.0 percent rate, quits at 3.4 million, and hires at 5.4 million. September job openings on that table were 7,523,000. Divide 1.6 million by October's 43,994 announcements and the ratio is 36.4 across neighboring months and different definitions; divide by the average announcement month, 61,718.9, and one month of discharges is 25.9 times the press-release flow, while quits were 2.13 times layoffs.

The 13 November Goldman note expects a small labor-market effect in 2027 and more notable effects in 2028 and 2029, even while it says Hatzius sees early signs that adoption is weighing on jobs, and it puts 2027 AI investment at $1.8 trillion, up from $1.1 trillion. Fortune, quoting a Sunday U.S. outlook that is not that public page, says the bank told clients the feared layoff wave did not arrive and that low-hire, low-fire conditions last one more year "before AI displacement begins to have a larger impact in subsequent years."

On that July exchange Daron Acemoglu expected some net job losses within five years, "less than two to four percent," and he sized today's most exposed customer-service and back-office tasks at "eight million in the United States, nine million," a stock that is neither a press release nor 15 million. Stanford, revised 12 August on ADP payrolls through June, finds no widespread displacement, and finds workers ages 22 to 25 in the most exposed jobs about 19 percent below the pace of less exposed peers the same age, versus a 15 percent shortfall at the July 2025 vintage, a gap the authors place in hiring rather than separations and label descriptive rather than causal.

What You Can Do

If a reduction is blamed on AI, ask for the headcount, the stated reason, and whether it is a counted cut or a voluntary offer, because the Microsoft pool of 60,000 never entered October's 43,994. If a model starts from 15 million, put 145,681 next to it before putting a year on it: 9.7 percent of an even ten-month slice is a pace, not a promise that the rest never arrives.

Watching the market means pairing 4.1 percent with U-6 at 7.9 and with the 339,000 rise in involuntary part-time work, treating claims of 200,000 and a four-week average of 195,250 as the firing tape and September's 1.6 million discharges as the wider one. For early-career office hiring, Stanford's 19 percent shortfall through June is the series with an age band, the October payroll report does not replace it, and the November Employment Situation is set for 4 December.

Limitations

Not a forecast. This does not prove a 103-year transition, a healthy market in every occupation, or that an employer's stated reason is the cause, because Challenger counts public announcements and misses silent separations, which is why 1.6 million September discharges can sit beside 43,994 October announcements. Briggs tied 15 million to a 15 percent productivity assumption spread over about ten years, with any single year's unemployment-rate increase staying under one point, and 9 percent of October employment is 14.7 million rather than 15. A clock made only of 2026 announcements answers a question Goldman has dated to 2028 and 2029.

Household and payroll surveys disagree by 298,000 on October's employment change, "changed little" is a judgment about a noisy sample, and the unemployed count moving from 7,109,000 to 7,068,000 can be signal or noise. JOLTS is still the September release and still preliminary, Stanford stops in June on ADP records, and the authors say the 19 percent gap is descriptive, shrinks once education is held constant, and shows some drift that predates generative AI. The 13 November Goldman page says it is an Executive Office summary, not Global Investment Research, Fortune's lines from the Sunday outlook are here because that note is not public, and nobody has published how many of the 60,000 Microsoft offers will be accepted.

The Strongest Case Against This Analysis

Grant the timing argument its full strength. Briggs said in July that the unemployment effect stays under a point a year if 9 percent of workers are reallocated over a decade, and that automation's labor damage shows up faster when the economy slows. A healthy 4.1 percent unemployment rate, claims at 200,000, and quits running 2.13 times layoffs are exactly the setting in which he would not expect the announcement tally to look like 15 million. Judging that path by dividing this year's press releases into 15 million is a category error: the harm they describe is a slowdown in hiring and a reorganization, not a stack of AI-labeled layoff notices.

Bring another dataset. Stanford's 19 percent shortfall for workers 22 to 25, and Acemoglu's eight to nine million in customer service and back office, never needed a layoff headline, so a 103-year figure bounds the press-release channel only, and anyone who wants the larger claim has to arrive with records this tally does not contain.

The Bottom Line

Payrolls rose 165,000, unemployment is 4.1 percent, U-6 rose to 7.9, and part-time work for economic reasons rose by 339,000. The firing tape is quiet. The announcement tape is a poor stand-in for a reallocation that is not supposed to be finished this year.

Related