Bank of America: Slower AI capital spending could lower peak chip earnings and extend the industry cycle

Why might capital spending cuts extend the chip cycle?

BofA believes financing pressure could force some AI projects to be delayed or scaled back. If monetizable demand remains strong, slower growth in compute supply would raise equipment utilization, support prices and reduce the risk of overbuilding.
For chipmakers, the trade-off could be a lower earnings peak; the benefits would be a longer earnings cycle and potential support for valuation multiples. BofA therefore remains cautious but positive on the AI semiconductor sector.

Long-term contract backlogs at major cloud providers compiled by BofA exceed 2.5 trillion dollars. Recently, AWS revenue rose 39% year over year, Azure grew 49%, and Google AI revenue rose more than 100% year over year.
Despite compute constraints, AI token usage rose to 10 times its year-earlier level; according to Silicon Data, rental rates for Hopper and Blackwell have risen 33% and 51%, respectively, since the start of the year, indicating continued pricing support for existing compute capacity.
Enterprise paid demand is also expanding: Fortune 500 companies' spending on Amazon Bedrock rose to about 7 times its year-earlier level; the number of Microsoft customers purchasing more than 50 thousand Copilot seats rose to more than 8 times its year-earlier level, and Copilot daily active users increased 50% quarter over quarter.

The market consensus expects the combined free cash flow shortfall of eight major US cloud providers and emerging cloud providers to widen from about 128 billion dollars in 2026 to 335 billion dollars in 2027, rising from about 6% to 13% of revenue.
The same set of forecasts shows the shortfall narrowing to about 77 billion dollars in 2028, or about 2% of revenue; BofA therefore believes 2027 could be the peak year for cash burn.
BofA expects Nvidia, Broadcom and Micron to generate combined free cash flow of about 577 billion dollars in 2027 and more than 800 billion dollars in 2028, potentially providing a buffer for ecosystem financing, though supplier financing also increases the risk of circular financing.

At the time of the report, the Philadelphia Semiconductor Index traded at 16.4 times forward earnings over the next 12 months, a discount of about 10% to the S&P 500; since 2019, its deepest discount to the S&P has been about 20%.
BofA calculates that, if the S&P 500 valuation and earnings forecasts remain unchanged, a further decline of about 11% in the semiconductor index would return it to that historical extreme discount. The bank believes the market has priced in a significant degree of capital spending risk ahead of any weakening in demand data; whether valuations can rise still depends on the resilience of paid demand.