The AI investment boom has caused the off-balance-sheet debts of five U.S. tech giants to surge eightfold in approximately four years, reaching $1.65 trillion and surpassing their on-balance-sheet debts, with Meta and Oracle experiencing particularly significant increases. These off-balance-sheet debts primarily stem from long-term contract procurement and data center leases related to AI R&D. Under current accounting standards, these debts are not included on the balance sheet but are only disclosed in the footnotes of financial reports, potentially leading individual investors to underestimate the risks. The tech giants claim that future revenues will cover these debts, but their investments have already exceeded profitability levels, relying on methods such as issuing bonds and issuing additional shares to raise funds. Overheating signs have emerged in AI infrastructure investments. Relevant reports warn that this 'shadow borrowing' model may trigger market risks, and off-balance-sheet debts will inevitably be included on the balance sheet sooner or later. If AI demand falls short of expectations, significant losses could result. Meanwhile, part of the demand in the AI industry comes from circular investments, further exacerbating the risk of overbuilding data centers.