路透社：AI 融资激增，是科技革命还是债务泡沫前夜？

Core Insights - The AI investment race has entered a "burning money" phase, with technology companies exploring new financing channels to meet an expected investment demand of up to $7 trillion by 2030 [1] Group 1: Surge in Investment-Grade Bond Issuance - Major tech companies focused on AI issued $75 billion in investment-grade bonds in September and October 2025, more than double the average annual issuance of $32 billion from 2015 to 2024 [1] - Meta issued $30 billion, while Oracle issued $18 billion, with AI-related companies now accounting for 14% of Morgan Stanley's investment-grade index, surpassing the banking sector [1] Group 2: Stock Prices and Risks - Oracle's stock price surged by 54% in 2025, marking its strongest annual gain since 1999, but the credit default swap (CDS) index also rose significantly, indicating increased concerns over rising debt levels [4][7] Group 3: Emergence of High-Risk Financing Channels - The financing landscape for AI is shifting from traditional investment-grade bonds to higher-risk areas, with companies like TeraWulf and CoreWeave issuing $3.2 billion and $2 billion in high-yield bonds, respectively [10] - Private credit from non-bank institutions has also grown significantly, with UBS estimating that AI-related private credit loans may have doubled in the past year [10] Group 4: Revival of Complex Financial Instruments - Asset-backed securities (ABS) are gaining popularity to meet the funding needs of AI data center construction, with the digital infrastructure segment expanding eightfold in less than five years [13] - Special purpose vehicles (SPVs) are being used again, as seen with Meta's $30 billion debt financing for its Louisiana data center, isolating debt from the company's balance sheet [13]