US lawmakers have expressed concerns regarding the financial risks associated with the rapid expansion of AI infrastructure. They have explicitly stated that there will be no federal bailouts or financial support if the sector faces a downturn.
Projections indicate that the AI infrastructure buildout will require over $10 trillion in investment by 2032, representing approximately 3.6% of the annual US GDP. To achieve projected returns, the industry is expected to need $3.7 trillion in annual revenue by 2032.
Companies including Oracle, Microsoft, Amazon, Meta, and Alphabet are projected to have a combined capital expenditure exceeding $800 billion in 2026, a figure that surpasses their current operating cash flow.
Concerns center on the use of complex debt structures, such as special-purpose vehicles (SPVs), leases, and project finance, which often keep borrowing off corporate balance sheets. For instance, a data center project in New Mexico involved $18 billion in bank loans and $3 billion in equity.
The reliance on off-balance-sheet financing complicates risk assessment for regulators and lenders. Current revenue estimates for OpenAI and Anthropic are approximately $100 billion, highlighting a significant gap between current performance and long-term industry requirements.
