The biggest sign of an AI bubble is starting to appear – debt (www.axios.com)

Big Tech is increasingly funding the AI infrastructure boom with debt — and much of it is being routed through private lenders and special-purpose vehicles (SPVs) that keep borrowing off corporate balance sheets. Meta is seeking roughly $29 billion in private capital for data centers, Oracle issued $18 billion of debt for AI expansion, and Nvidia’s reported $100 billion investment in OpenAI highlights circular flows of capital between suppliers and customers. Analysts such as Dario Perkins warn this shift into hidden leverage — and firms’ apparent indifference to near-term returns — looks like a classic late-cycle signal: heavy upfront spending on compute, hardware and buildings funded by borrowing that investors may not fully see. The implication for AI/ML is twofold. Technically, the spending is buying real, durable assets (GPUs, servers, data center capacity) so the capex retains value even if AI monetization lags; strategically, however, leveraging to build infrastructure assumes future revenue streams that may not materialize, concentrating systemic risk. SPVs and private debt can mask borrowing from rating agencies and shareholders, increasing tail risk through off‑balance-sheet leverage and “recycling” of capital. For practitioners and investors, that means watching hidden debt, insider exits and the economics of monetizing large-scale models — and recognizing that diversification remains the safest hedge if this financing-driven froth reverses.