IMF《全球金融稳定报告》警告：AI行情一旦回调后果将是全球性的，三大近期风险「不容自满」
曼谷——国际货币基金组织（IMF）10月13日在曼谷年会期间发布新一期《全球金融稳定报告》，给正处在高位的全球市场泼了一盆冷水。报告的判断是：市场迄今表现得很有韧性——企业盈利扎实、AI乐观情绪浓厚——但金融稳定风险仍然高企，决策者「没有自满的余地」。这句话是IMF货币与资本市场部副主任瓦姆瓦基季斯与全球市场分析部助理主任吴（Jason Wu）在配套文章与发布会上的共同口径。
报告点名三个可能成真的近期触发点：一是AI行情回调，二是主权债券收益率急升，三是重大网络攻击。IMF认为，单看每一项都未必动摇金融稳定，但如果它们与体系内既有的脆弱性叠加，就可能演变成系统性冲击，并导致金融条件骤然收紧。
AI是报告着墨最多的一条。IMF指出，越来越多证据显示，AI能力进步未必需要市场目前假设的那么大的投入，其经济收益也可能没有预期那么大；一旦AI投资动能放缓，行业内已经堆起来的杠杆会放大冲击——强迫卖盘、波动率飙升都可能跟着来。两个数字最能说明杠杆堆到什么程度：超大规模云厂商（hyperscalers）如今已占全球投资级债券发行量的10%，而2024年时还不到1%；IMF复盘2023年以来14笔AI领域的「循环融资」交易，发现其中约有3,000亿美元股权投资与6,500亿美元债务，对应的是1.4万亿美元的未来收入承诺——一家公司的收入预期落空，会顺着股权与债务链条传导到其他公司。吴在发布会上把话说得很直白：如果AI的变现能力受到质疑，股权投资会缩水、债务可能难以偿付、未来的收入承诺也可能无法兑现。IMF因此呼吁各国决策者约束「过度」的杠杆使用，包括衍生品与保证金贷款。
这份报告与同日发布的《世界经济展望》（GSM已另行供稿）是一组互补的判断：展望报告算的是增长——AI投资托住了被油价冲击的全球经济；稳定报告算的是风险——托住增长的同一股力量，估值与债务也堆得最高。IMF在展望的下行情景中测算过，若AI投资因盈利失望而陡降、引发资产价格大跌，到2028年全球GDP可能被拉低1.1%，美国产出降幅可达1.8%。
IMF Warns an AI Market Correction Could Go Global, Naming AI Selloff, Bond-Yield Spike and Cyberattack as Near-Term Catalysts
Bangkok — The International Monetary Fund released its latest Global Financial Stability Report on October 13, during the annual meetings in Bangkok, throwing cold water on markets riding high. The Fund's judgment: markets have so far shown resilience — solid corporate earnings and thick AI optimism — but financial stability risks remain elevated, and policymakers have 'no room for complacency.' That was the joint message of Athanasios Vamvakidis, deputy director of the IMF's Monetary and Capital Markets Department, and Jason Wu, assistant director for global markets analysis, in an accompanying article and at the launch briefing.
The report names three plausible near-term triggers: an AI market correction, a sharp rise in sovereign bond yields, and a major cyberattack. The IMF said that any one of them, taken alone, might not shake financial stability, but if they stack on top of vulnerabilities already inside the system, they could evolve into a systemic shock and cause financial conditions to tighten abruptly.
AI gets the most ink in the report. The IMF said evidence is growing that advances in AI capability may not require investment on the scale the market currently assumes, and that the economic payoff may be smaller than expected; if the momentum of AI investment slows, leverage already piled up in the sector would amplify the shock — forced selling and volatility spikes could follow. Two figures best show how high the leverage has climbed: hyperscalers now account for 10% of global investment-grade bond issuance, up from under 1% in 2024; and reviewing 14 'circular financing' deals in the AI field since 2023, the IMF found roughly $300 billion in equity investment and $650 billion in debt set against $1.4 trillion in future revenue commitments — if one company's revenue expectations fall short, the hit travels along equity and debt chains to other companies. Wu put it bluntly at the briefing: if AI's ability to monetize is called into question, equity investment will shrink, debt may go unpaid, and future revenue commitments may not be honored. The IMF accordingly called on policymakers to rein in 'excessive' leverage, including derivatives and margin lending.
The report forms a complementary pair with the World Economic Outlook released the same day (covered separately by GSM): the Outlook does the growth math — AI investment is propping up a global economy hit by the oil shock; the stability report does the risk math — the same force propping up growth has also piled valuations and debt the highest. In a downside scenario in the Outlook, if AI investment dropped sharply on earnings disappointment and triggered a major asset-price fall, global GDP could be reduced by 1.1% by 2028, with U.S. output falling by as much as 1.8%.
图片来源：国际货币基金组织华盛顿总部大楼。APK 摄，CC BY 4.0，维基共享资源
