Goldman Sachs Says AI Job Losses Are Spreading Beyond Tech as Spending Boom Nears Peak

Goldman Sachs now says artificial intelligence is cutting into hiring well beyond Silicon Valley, and the same research warns that the growth boost from AI spending is close to topping out. The bank's chief economist, Jan Hatzius, wrote in a client note that the lift from the AI boom is likely to peak in late 2026 or early 2027, a judgment he reached even before the past few weeks of investor anxiety about whether the spending can be sustained.

The labor-market findings are the more striking part. Goldman examined U.S. non-manufacturing industries and found that companies making heavier use of AI in everyday operations are posting weaker employment growth and higher unemployment than their peers. Both relationships hold up statistically, with the bank comparing the latest three-month readings against averages from 2015 through 2019.

A separate note from Goldman economist Joseph Briggs, dated October 22, put a number on the damage: employment in the industries most exposed to AI has fallen roughly 10,000 jobs below its long-term trend over the past three months. Hiring tied to AI-related construction has largely filled that hole, masking the underlying weakness.

The strain is also unevenly distributed. Research covering more than 800 occupations found that entry-level workers face the strongest AI-related headwinds, a finding that matters because college graduates and early-career employees are usually the first to be squeezed when companies slow hiring.

Corporate America is already responding. American Express said on October 23 that it would cut about 5% of its global workforce, with CEO Stephen Squeri saying the company is leaning on AI to become more agile and efficient. A card issuer rather than a chipmaker making that move is precisely the kind of broadening that Hatzius describes.

Hatzius did not sound an alarm about the overall labor market. He said developed economies are still holding up, and Goldman has kept its view that AI spending will eventually pay for itself through stronger productivity and profits. The awkward part is the mechanism: some of that payoff arrives because companies need fewer people, and the bank concedes the adjustment does damage before it delivers benefits.

Briggs has been tracking the drag for months. In July he estimated that AI tools were already shaving 10,000 to 15,000 jobs from monthly payroll gains, concentrated in technology, management consulting and graphic design, and he projected that AI could displace about 9% of American workers over the next decade. More recently, the research showed AI eliminating more positions each month than it creates in adjacent roles, though the net gap narrowed by June as data-center construction hiring accelerated.

That cushion may prove temporary. Only 23% of U.S. establishments currently use AI in regular workflows, which suggests most of the disruption lies ahead, and an economic downturn could speed up automation and layoffs. Goldman's base case remains a gradual increase in AI-related disruption rather than a sudden shock.

On the investment side, the bank said the growth impulse from AI was set to peak in late 2026 or early 2027 even before doubts about the boom intensified. AI investment is still climbing quickly, but its increase as a share of GDP is expected to slow over the next twelve months, and the fading effect of last year's larger U.S. depreciation allowances adds another drag on capital spending.

Goldman's strategists expect the five largest U.S. hyperscalers to keep raising AI spending, but they project capital expenditure growth will fall sharply in 2027 and again in 2028. The bank estimates those companies need about $300 billion in annual AI revenue just to break even.

Markets have begun pricing in the slowdown. Tech stocks slid on October 22 after Meta raised its capital spending guidance and Tesla cut its full-year capex guidance by $5 billion. Kyle Rodda, an analyst at Capital.com, said the concern has flipped from hyperscalers spending too much to investment growth losing momentum.

Financing costs compound the problem. The 10-year Treasury yield climbed in early October to its highest level since 2002, and Federal Reserve Chair Kevin Warsh has acknowledged that hyperscalers borrowing in the bond market are adding to competition for capital. The largest hyperscalers have so far absorbed higher funding costs, but smaller AI players are more exposed.

Hatzius also flagged several downside risks to U.S. growth. Energy prices have climbed back toward or above the highs of March and April on fears of renewed Middle East escalation, and long-term rates across developed markets have risen about 100 basis points since the start of 2026, partly because of large deficits and AI-related borrowing. He estimated that if the pattern persists it would shave 0.6% from U.S. GDP on a cumulative basis. Hyperscaler data-center investment as a share of GDP is expected to peak in the fourth quarter at roughly 1.5% year over year, which signals a fading marginal contribution to growth rather than an outright decline in spending.

Households have benefited from the boom as well. Goldman estimates the wealth effect from AI-linked stocks added 0.4 percentage point to U.S. consumer spending growth over the past year, equal to about 0.25% of GDP. Hatzius expects that support to diminish over the next year, gradually if the bank's relatively upbeat equity outlook holds and more sharply if AI shares keep falling. Consumer sentiment slipped to near a record low in October as rising prices tied to the Middle East conflict darkened views of the economy.

For workers and investors, the bank's research points to adoption rates and entry-level hiring as the early warning signals rather than the unemployment rate, which tends to show damage late. Both are already flashing caution, from American Express trimming staff ahead of its AI push to young workers struggling to land jobs.

Goldman itself is deeply embedded in the boom it is analyzing. The bank co-led a $95 billion equity offering for Meta, served as sole structuring agent on a $40 billion chip financing for SpaceX, and was named lead bookrunner on Anthropic's planned IPO, according to Bloomberg. The three deals total at least $210 billion and could generate tens of millions to hundreds of millions of dollars in fees. Goldman President John Waldron said in May that the firm feels it is "right at the center of this investment supercycle."

Underwriting revenue across the five largest investment banks reached nearly $19 billion in the first nine months of the year, with equity capital markets fees up 48% and debt underwriting up 13%, though the banks do not break out AI-related fees separately.