The Bank for International Settlements (BIS) cautioned in its quarterly September report that investors are beginning to reassess the profitability of large-scale AI projects, the durability of high margins, and the rapidly swelling debt burdens of technology firms. The convergence of elevated interest rates and stretched valuations has already undermined the momentum of the AI-fueled global equity rally, yet overall risk appetite remains resilient, and the report stops short of declaring a definitive trend reversal in AI-linked markets.

High rates are now testing AI valuations, with the BIS noting that tech share momentum started to fade after late June. Beyond inflationary concerns stemming from geopolitical tensions and energy prices, the prospect of interest rates staying higher for longer has raised financing costs for AI ventures and depressed the present value of their future earnings. While major cloud computing and data center operators continue to post solid profits, their valuations have already undergone adjustment. During the review period, the price-to-earnings ratios of hyperscalers fell by nearly 15%, while semiconductor manufacturers saw a steeper decline of roughly 40%. In contrast, US small-cap stocks, traditional sectors, and markets outside the US have outperformed, signaling that investors are reducing their concentrated bets on a handful of AI frontrunners. The BIS also noted that risk appetite recovered somewhat in August, keeping overall market volatility at subdued levels, with the more notable shift being a widening dispersion in performance across individual stocks and sectors rather than a systemic selloff.

Tech firms are accelerating debt issuance, as investor focus pivots from AI revenue growth to whether capital expenditures can generate sufficient cash flow. The BIS previously estimated that the five largest hyperscalers would collectively spend over $1 trillion on AI capital outlays between 2025 and 2026. With investment commitments now exceeding corporate earnings and free cash flow, reliance on bond financing has become increasingly critical. The latest quarterly report shows that large tech companies are expanding their issuance of long-term bonds, while their credit default swap spreads have widened, reflecting creditors' demands for higher risk compensation. At the same time, issuance of high-yield bonds and leveraged loans has slowed, suggesting that funding remains available for tech giants with robust balance sheets, but that lenders are turning more cautious toward lower-rated borrowers.

Pricing of private credit risk is drawing scrutiny, according to research published alongside the report. Borrowing by US software and technology firms through private credit has surged from roughly $22 billion in 2010 to over $1 trillion by 2025, lifting their share of total private credit from 22% to 44%. This statistic spans the entire software and tech sector rather than being exclusively tied to AI projects, though AI companies have been a major driver of demand growth since 2020. Private credit arrangements typically rely on recurring revenue and intangible assets as lending criteria, offering a funding channel for tech firms that lack traditional collateral. However, narrowing loan spreads combined with a weakening borrower fundamentals have led the BIS to worry that related risks may not yet be fully priced in.

The BIS's assessment implies that the long-term productivity potential of AI technology must be evaluated separately from the current pricing of related assets. If corporate earnings materialize more slowly than anticipated, high valuations, mounting debt, and concentrated positions could amplify a downturn in tandem. Conversely, should profit and productivity improvements keep pace with capital spending, markets may still absorb the ongoing expansion in financing.