Lawmakers flag financial risks in AI buildout but rule out bailouts

Lawmakers in Washington say the AI infrastructure boom could pose risks to the broader financial system. They also have no plans to increase federal financial support if those risks turn into real problems.

The size of the bet

The AI infrastructure buildout is projected to require more than $10 trillion in US investment through 2032. That works out to approximately 3.6% of annual US GDP.

Sen. Elizabeth Warren is among the lawmakers raising alarms about these rising capital expenditures. An analysis released on September 24, 2026, tied the risks to the complicated debt structures increasingly used to pay for the buildout.

Those structures include leases, project finance and special-purpose vehicles, or SPVs. An SPV is essentially a separate legal entity created to hold an asset and its debt. That setup can keep the borrowing off the parent company’s own balance sheet.

Oracle, Microsoft, Amazon, Meta and Alphabet are expected to post combined capex of over $800 billion in 2026. That figure exceeds what those companies currently generate in operating cash flow.

Where the debt is hiding

The shift toward off-balance-sheet financing changes the calculus. Leverage is rising, and the risk is moving outward to the banks, private credit funds and real estate vehicles now involved in the financing.

One example shows how this looks in practice. A New Mexico data center associated with Oracle and Blue Owl drew $18 billion in bank loans plus $3 billion in equity.

When debt sits inside SPVs and lease arrangements, it becomes harder for regulators, investors and lenders to see the full exposure in one place.

The revenue gap

The industry needs around $3.7 trillion in annual revenue by 2032 to deliver projected returns. Current combined estimates for OpenAI and Anthropic sit at about $100 billion.

Why Washington is paying attention now

The current cycle adds a wider cast of financiers. Banks, private credit funds and real estate vehicles all have exposure, which raises both the total leverage in the system and the difficulty of tracking it.

With the US midterms approaching, questions about who bears the cost if the AI boom stumbles are likely to become more charged. Policymakers have stated their refusal to provide federal backstops or bailouts.

What this means for investors and lenders

The key things to watch are the disclosures. How much debt sits in SPVs and leases, which lenders hold it, and whether regulators push for more transparency will shape how contained any downturn might be.