US AI Corporate Borrowing Hits Scale Where Pandemic-Era Fed Facilities Become Structural Backstop

Corporate capex by US hyperscalers on data centers, GPUs and power infrastructure exceeded $180 billion in 2024, financed largely through investment-grade issuance and leveraged loans. Primary records from SEC filings show Microsoft, Amazon and Google parent Alphabet alone added $95 billion in net debt since 2022 while maintaining dividend and buyback programs. The incentive structure rewards scale because market participants price in implicit Fed support calibrated to 2020 facility terms.

State interest centers on preserving US compute lead over Chinese competitors. Treasury and Commerce Department export controls on advanced chips create a protected domestic build-out, yet the same controls raise input costs and require sustained private leverage. Primary documents from the CHIPS Act and executive orders confirm the policy priority is technological supremacy rather than financial stability metrics.

The two-sided ledger shows gains in strategic positioning offset by concentration risk: top five issuers account for 62 percent of AI-related debt. If utilization rates fall below 70 percent by 2026, refinancing costs rise without new Fed facilities. No public record indicates the Fed has modeled AI-specific stress tests.

Next step is continued issuance through 2025 at current spreads, with any activation of emergency facilities requiring documented deterioration in employment or credit spreads exceeding 400 basis points.