Generative AI companies are increasingly turning to the corporate bond market to finance data centers and semiconductor purchases, moving beyond funding from big technology companies and venture investors. Investment banks overseeing the potential IPOs of OpenAI and Anthropic are persuading Moody’s, S&P and Fitch to award investment-grade ratings, the Financial Times reported on the 8th. Such ratings could reduce borrowing costs and open access to pension funds and insurers whose mandates restrict them to investment-grade debt. AI-related capital expenditure is expected to exceed $1 trillion this year and reach $5 trillion to $7 trillion cumulatively by 2030. Reuters Breakingviews, citing Goldman Sachs data, said AI-linked debt reached about $500 billion from the start of this year through early August, representing roughly 20% of high-quality U.S. corporate bond issuance, compared with 1% in 2024. Complex funding arrangements involving private loans, SPVs (separate legal entities created for financing), securitization and conditional guarantees are making total AI-related debt and risk difficult to assess. The concern is that interest payments begin before data centers generate profits, while power connections, permits and semiconductor supplies can delay operations. Oracle has drawn particular scrutiny after reporting negative free cash flow alongside a $260 billion long-term data-center lease contract and $129.5 billion in debt; S&P cut its rating to BBB-, one notch above speculative grade, in July. AI corporate bond yields have risen to 7%-8% annually. The European Central Bank has warned that investors shifting into U.S. big-tech bonds could crowd European companies out of the corporate bond market. Brown Advisory fixed-income chief Colby Steel said risk increases when debt repayment depends on money companies have not yet earned, especially when they lack positive free cash flow.