Investors are translating doubts about AI spending into a lower growth forecast for the whole US economy

US economic data still look strong: third-quarter GDP grew at an annualised 3.4%, above the 3.2% consensus. Yet inflation-adjusted Treasury yields fell sharply over the past fortnight, with the 10-year real yield reaching 2.58%, its lowest level since early September. The Financial Times reports that investors are buying government bonds because they increasingly see weaker growth as one possible consequence of the AI trade's reassessment.

The catalyst was not a collapse in model demand. It was the gap between spending and cash generation. Meta, Alphabet, Amazon and Intel all contributed to a 2% weekly fall in the Nasdaq 100; Alphabet dropped 13% after reporting negative free cash flow for only the second quarter since its 2004 listing. Other large technology companies raised capital spending even as some revenue growth slowed.

This converts a familiar company-level question—will AI investment earn an adequate return?—into a macroeconomic one. Data-centre construction, chips, power equipment and supplier finance now contribute enough to business investment that a slowdown would reach contractors, utilities, lenders and local tax bases. Conversely, continued spending with disappointing productivity could keep demand and interest rates high while weakening corporate cash flow.

That is why the market is pricing wider tails rather than a simple recession call. PGIM's chief US economist raised his estimate of recession risk from 10% earlier this year to as much as 25%, while also acknowledging a greater chance of unusually strong growth. The same AI build-out can support rapid expansion or overheating before it becomes a drag.

The bond move should not be mistaken for proof that the boom is ending. Year-on-year growth comparisons have become harder after an exceptional 2025, and current output remains robust. It is evidence that investors no longer treat AI capital expenditure as an unambiguously positive growth input. The financing chain has become large enough that doubts about utilisation, depreciation and margins now alter the price of safety across the economy.