The AI stock boom may be entering its final stages, even as Anthropic prepares to pursue one of the largest US initial public offerings in recent history.
The artificial intelligence company is expected to go public on Thursday in a deal that could raise about $64.35 billion at the top of its range. That would be almost twice the total raised by all US IPOs combined last quarter, according to Renaissance Capital.
The flotation is being watched as a test of whether investors remain willing to back AI companies at lofty valuations. Joe Maher, a markets economist at Capital Economics, pointed to falling share issuance and “investors’ waning appetite for new shares of AI-firms at lofty valuations”. He described Anthropic as “a key test”.
The timing matters beyond the fortunes of Anthropic’s bankers. AI investment has become a major driver of US economic growth, while employment has expanded only modestly – a combination Goldman Sachs economists have described as “jobless growth”.
The US economy grew at an annual rate of 3.4% in the third quarter, but employers added an average of just 48,000 jobs a month in the 12 months to September. Employment then rose by 165,000 in October. Goldman estimates that AI spending accounted for more than half of business investment growth during the first three quarters of the year.
David Kelly, chief global strategist at J.P. Morgan Asset Management, expects real GDP growth to slow to 1.6% over the next four quarters, with monthly hiring remaining close to 50,000.
He said two years of 2.3% growth should have produced roughly 155,000 net new jobs a month. Instead, the average gain was 48,000, which he largely attributed to “a lack of available workers due to a sharp reversal of immigration policy”.
Anthropic IPO puts AI investment under scrutiny
The threshold for stable employment has fallen as the labour force has grown more slowly. Morgan Stanley’s research digest, summarising chief US economist Michael Gapen’s outlook for 2027, estimated that about 50,000 jobs a month are needed to prevent unemployment from rising.
Federal Reserve economists Seth Murray and Ivan Vidangos believe labour force growth “could be near-zero starting this year”. If that happens, the pace of hiring needed to maintain employment would also be close to zero, leaving productivity growth as the source of any increase in potential GDP.
That places greater pressure on AI investment to deliver lasting improvements in productivity and output. However, much of the spending so far has benefited companies overseas rather than being fully reflected in measured US growth.
Goldman estimates that AI investment could add 0.6 percentage points to “true” GDP growth in 2027, but only 0.2 points to recorded GDP. Chips are undercounted in official data and many of the servers used for AI are imported, meaning their contribution is offset in the calculations.
Morgan Stanley is more optimistic, forecasting that hyperscalers will spend $1.6 trillion on capital expenditure in 2027 and add 0.9 percentage points to growth. Gapen expects the US economy to expand by 2.5% next year, compared with Goldman’s 2.4% forecast and J.P. Morgan Asset Management’s 1.6% projection over the next four quarters.
The two banks broadly agree that hiring will remain subdued. Gapen expects AI disruption and tighter Federal Reserve policy to weigh on employment, while AI investment is largely supporting higher-income households as lower- and middle-income consumers lead the slowdown.
Investor demand for Anthropic remains strong for now. The offering was reported to be about six times oversubscribed, while Nvidia is buying about $10 billion of Anthropic stock in a private placement alongside the flotation.
Workers face a low-hire, low-fire economy
Goldman Sachs has described the current labour market as “low-hire, low-fire”, a phrase previously associated with former Federal Reserve chairman Jerome Powell. The conditions are more favourable for people who already have jobs than for those trying to enter the workforce.
The unemployment rate for Americans aged 20 to 24 was 7.5% in October, compared with 4.1% for the overall population. Research by Stanford’s Digital Economy Lab found that employment among 22- to 25-year-olds in the most AI-exposed jobs was about 19% below the level it would have reached if it had kept pace with less-exposed work.
Pay growth has also been weak in real terms. Average hourly earnings rose by 3.1% in the year to October, but J.P. Morgan Asset Management said real year-on-year wage growth had been negative for seven consecutive months despite a tight labour market.
Productivity in the US non-farm business sector increased at an annual rate of 3.6% in the third quarter. Yet labour’s share of output – the proportion paid to workers as compensation – fell to 53%, its lowest level since records began in 1947.
Aubrey Woessner of Indeed Hiring Lab said AI investment appeared to be contributing to the productivity gains, but added that “increased labor productivity has not yet translated into higher real pay”.
Research by Apollo Global Management suggests that wage pressure has so far been more evident than widespread job losses in occupations most exposed to AI. Economists Sania Edlich and Torsten Slok compared occupations according to how heavily their tasks featured in conversations with Anthropic’s Claude system.
They found that, since 2023, real wage growth in the most exposed occupations had lagged less-exposed jobs by about 6.7 percentage points, without a statistically significant employment effect. Workers in the lowest-paid quarter of those occupations experienced a gap of about 10.7 percentage points.
“The first measurable labor market effect of AI adoption has been wage compression rather than employment displacement,” the economists wrote. They estimated that about 5.8 million workers were employed in the most exposed occupations, while cautioning that the period could have been partly affected by post-pandemic labour market changes.
Other measures offer a different picture. Advertised salaries for the most AI-exposed occupations have risen by 46% since 2021, compared with 25% for the least exposed, according to Indeed data. Those figures measure what employers offer new recruits rather than the pay received by people already in the roles.
Public sentiment remains subdued. The University of Michigan’s preliminary November consumer sentiment reading was 51.0, with survey director Joanne Hsu reporting that about 57% of respondents spontaneously said their living standards had been eroded by high prices.
Capital Economics believes the AI stock boom is nearing its end. Goldman Sachs expects the low-hire, low-fire labour market to continue for about another year, before AI displacement begins to have a larger effect in subsequent years.
