谷歌、Meta、亚马逊和微软等科技巨头自2024年以来已借入约5000亿美元用于资助AI基础设施建设，标志着企业债务出现历史性激增。
Tech giants like Google, Meta, Amazon, and Microsoft have borrowed around US$500 billion since 2024 to fund AI infrastructure, marking a historic surge in corporate debt.
投资者要求向这些公司贷款获得更高回报，导致借贷成本上升，并降低了对美国政府债券的需求，推高了美国利率。
Investors demand higher returns for lending to these firms, causing increased borrowing costs and reducing demand for US government bonds, pushing up US interest rates.
专家警告称，如果AI热潮降温，将面临金融风险，可能引发市场冲击和传染效应，甲骨文等公司面临债务压力和项目延期便是例证。
Experts warn of financial risks if the AI boom slows, with potential market shocks and contagion fears linked to companies like Oracle facing debt and project delays.
AI生成 纽约——全球最富有的公司已无法仅靠庞大的现金储备在人工智能竞赛中保持领先，它们突然开始大规模举债，这种转变正向全球传导影响。
AI generated NEW YORK – The world’s richest companies can no longer rely on their massive cash piles alone to stay in the artificial intelligence race and have suddenly started borrowing massively in a shift that is sending repercussions across the world.
包括作为全球经济锚定资产的美国国债在内的美国利率上升，正在金融界引发震动，一些分析师将AI借贷狂潮指认为罪魁祸首之一。
Rising US interest rates, including on the Treasury bonds that anchor the global economy, are sending tremors through the financial world, and some analysts point to the AI borrowing bonanza as one of the culprits.
自2024年几乎为零起步，科技行业借贷在1月至9月的九个月内已达到约5000亿美元（约6397亿新元），谷歌、Meta、亚马逊、微软等公司争相发债，为芯片、服务器和数据中心等AI算力基础设施融资。高盛预计2027年将进一步攀升至1.2万亿美元。
From next to nothing in 2024, tech sector borrowing has reached around US$500 billion (S$639.7 billion) in the nine months since January, as Google, Meta, Amazon, Microsoft and others raise debt hand over fist to finance the chips, servers and data centres that power AI. Goldman Sachs expects a further ramp-up in 2027, to US$1.2 trillion.
“这是我们前所未见的现象，”融资咨询公司Post Oak Group高级副总裁Chris Della Fave表示，他估计AI目前占所有企业债券发行量的25%，而两年前这一比例仅为4%。
“This is not something that we’ve seen before,” said Chris Della Fave, senior vice-president at fund-raising advisory firm Post Oak Group, who estimates that AI now accounts for 25% of all corporate bond issuance, up from 4% two years ago.
按通胀调整后计算，预计2026年AI行业的借贷规模将超过美国有线电视运营商为铺设整个互联网而借入的资金，也超过19世纪美国铁路繁荣时期铁路公司的借贷规模。
flation-adjusted terms, the AI sector is expected to borrow more in 2026 than US cable operators did to build out the entire internet, or than railroad companies did during the 19th-century US rail boom.
目前，投资者热衷于借钱给这些科技巨头，但他们要求的回报率在不久前对于此类蓝筹股公司来说是不可想象的。
So far, investors have eagerly snapped up the chance to lend to the tech giants, but they have demanded returns that would have been unthinkable for such blue-chip companies not long ago.
即使是 Meta 也不得不提供超过 7% 的年收益率，而风险较高的云数据中心专业公司的收益率甚至已突破 9%。
Even Meta has had to offer more than 7% a year, while riskier cloud data centre specialists have gone above 9%.
这种影响远远超出了构建 AI 的公司范围——它们的债务甚至开始挤占对作为金融体系锚定资产的美国政府债券的需求。
The impact reaches well beyond the companies building AI – their debt is even starting to crowd out demand for the US government bonds that anchor the financial system.
Siebert Financial 首席投资官 Mark Malek 表示，本可能购买美国国债的投资者“可能会决定改买微软”的债券。他解释说，这种转变推高了华盛顿的借贷成本。
An investor who might otherwise buy a US Treasury bond “might decide to buy Microsoft” instead, said Mark Malek, chief investment officer at Siebert Financial, referring to the tech giant’s bonds. That shift pushes up the rates Washington pays to borrow, he explained.
这加剧了推高美国借贷成本的另一股力量：通胀，而通胀是由对伊朗的战争和高能源价格助长的。
This adds to the other force driving up US borrowing costs: inflation, fuelled by the war against Iran and high energy prices.
10 年期美国国债收益率——华尔街的基准，被广泛视为全球金融领域最重要的数字，决定着从房贷到车贷等一切利率的基调——目前已升至 5.3% 以上，为 2002 年以来的最高水平。“更剧烈的回调”加剧波动的是，对冲基金以前所未有的规模涌入美国国债市场，在 2025 年底持有流通国债的 7%，尽管此后这一比例已有所下降。
The interest rate on 10-year US government bonds – Wall Street’s benchmark and widely seen as the most important number in global finance, setting the tone for everything from mortgages to car loans – is now above 5.3%, its highest level since 2002. ‘Sharper correction’Adding to the volatility, hedge funds had piled into US government bonds like never before, holding 7% of all those in circulation at the end of 2025, though that share has since fallen.
对冲基金在市场上进行大额押注，其资金流动速度远快于保险公司和养老基金等更为审慎的投资者。
Hedge funds, which place big bets on markets, move their money far faster than more cautious investors such as insurers and pension funds.
Della Fave 表示，即使战争和石油局势趋于稳定，“老实说，我不指望收益率会大幅下降，因为 AI 债务情况的这种影响”。
Even if the war and the oil situation stabilised, Della Fave said, “I wouldn’t expect the yields to dramatically reduce, to be honest, because of this influence of the AI debt situation”.
除了借贷成本上升之外，一些人正在质疑押注 AI 繁荣的风险，因为它可能像 2000 年的互联网泡沫一样撞上墙壁。
Beyond the rising cost of borrowing, some are questioning the risks of betting on an AI boom that could hit a wall, as the dom bubble did in 2000.
Malek 警告称，即使建设狂潮的步伐仅出现适度放缓、个别项目延期或营收增速不及预期，都可能在金融市场引发冲击。
Even a moderate slowdown in the frenzied pace of construction, delays on certain projects or weaker-than-expected revenue growth could trigger a shock in financial markets, Malek warned.
9月下旬，英国央行金融政策委员会警告称，“大幅调整的风险持续存在”，特别是如果对人工智能发展或采用速度的担忧影响了盈利预期的话。
In late September, the Bank of England’s Financial Policy Committee warned that “the risk of a sharper correction persists”, particularly if concerns about the pace of AI development or adoption hit earnings expectations.
7月份，在一些人对人工智能繁荣的猜测中，以科技股为主的纳斯达克指数下跌了近7%。在此背景下，云专家Oracle有时被视为风向标。
In July, amid some second-guessing about the AI boom, the tech-heavy Nasdaq index fell nearly 7%. Against this backdrop, cloud specialist Oracle is sometimes seen as a bellwether.
由于巨额债务（1250亿美元）、现金储备每季度都在萎缩，以及新墨西哥州一个大型数据中心项目可能推迟，拉里·埃里森的团队正在闪烁几盏警示灯。
With massive debt (US$125 billion), cash reserves that shrink every quarter and a possible delay on a huge data centre project in New Mexico, several warning lights are flashing for Larry Ellison’s group.
“假设甲骨文遇到了问题..他们无法支付某些东西，”马利克说。他补充说，债务问题“可能会引发整个人工智能金融的蔓延”。
“Let’s say Oracle has a problem... They can’t pay for something,” Malek said. Trouble with its debt “could trigger contagion” across AI finance as a whole, he added.