Growing credit risk could pop the AI bubble (www.axios.com)

Big Tech is quietly leaning on off‑balance‑sheet and private credit to fund a massive AI infrastructure buildout, venture capitalist Paul Kedrosky and S&P’s Ruth Yang tell Axios. Firms are routing loans through special‑purpose vehicles and private lenders—moves that can hide true leverage from investors—as they race to build data centers and buy costly chips that account for more than half of buildout costs. Meta alone is reportedly seeking about $29 billion from private credit, while Carlyle projects $1.8 trillion of capital will be needed by 2030 to meet AI demand. Because AI monetization remains immature and chips require repeated upgrade cycles, these financing structures raise questions about who ultimately bears the risk and how the debt will be repaid. The significance for AI/ML is twofold: operational and systemic. Operationally, repeated, expensive hardware upgrades mean ongoing capital needs with unclear revenue offsets, challenging assumptions that free cash flow will cover the rollout. Systemically, the rise of opaque private credit—born from banks’ post‑crisis retreat—could concentrate hidden leverage across tech, lenders and fund investors, increasing the chance of overbuilding and a market correction similar to past debt‑driven busts. Warnings about circular investing (e.g., large firms funding their big customers) and project‑finance complexity suggest we should watch the AI buildout cycle closely: it could either underpin long‑term innovation or become the canary for a popped AI bubble.