The recent sell-off in US technology stocks has intensified concerns about the financial risks tied to the AI boom. On Friday, the Nasdaq dropped 2.86%, its steepest decline since June 5, as disappointing results from Intel and Amazon fueled worries about the rapid return on massive AI investments. While the primary focus has been on whether the billions spent on data centers, chips, and computing infrastructure will yield profits, another risk is emerging: a complex web of financing arrangements that could create hidden financial exposures.
Credit-rating agencies are particularly watchful. Mark Habib, S&P Global's head of digital infrastructure, noted that the biggest risk to hyperscalers' credit ratings remains the uncertainty over whether AI investments will generate expected profits. S&P estimates that six major US hyperscalers, including Amazon, Microsoft, Alphabet, Meta, Oracle, and SpaceX, will spend over $7 trillion on AI-related infrastructure between 2025 and 2030. These companies are projected to experience negative free cash flow until at least next year. However, Habib also highlighted the possibility of undisclosed AI-linked debt and financial liabilities, stating, "It is safe to assume that there is more (debt) out there that we are not seeing."
The financing structures behind AI infrastructure are growing increasingly intricate. S&P Global Ratings reported that combined capital expenditure by the six hyperscalers could exceed $1.3 trillion by 2027. These investments are often financed through debt, leases, joint ventures, special-purpose vehicles, and residual-value guarantees. For instance, Google and Anthropic are involved in a financing arrangement where Google invests in Anthropic while also providing cloud-computing capacity, creating a complex financial relationship. Such structures, while not inherently problematic, make it difficult to track ultimate financial exposure.
The risks extend beyond technology companies. Private-credit funds, infrastructure investors, and banks are heavily financing AI-related physical infrastructure. A Reuters analysis noted that AI-related transactions accounted for 34% of global private-credit deals in 2025. The Bank of England estimated that around $700 billion of data-centre spending between 2026 and 2028 could rely on private credit. However, there is a potential mismatch between the life of the financing and the assets being financed, as data-centre equipment can become technologically outdated faster than expected, complicating debt servicing.