Industry71°

保尔森与李彦宏同日警示 AI 泡沫风险

Why we picked it

当年主导 2008 银行救助的 Paulson 和百度李彦宏同一天说 AI 可能是泡沫。更扎心的数据是:AI 股权融资半年缩水 57%,缺口靠 5700 亿美元债务填。CoreWeave 债券收益率破 16% 这个细节,关心烧钱模式还能撑多久的人都该看看。

曾主持 2008 年美国银行救助的美国前财政部长 Hank Paulson 公开表示,自己从未见过这么多资金追逐一个缺乏盈利和投资回报支撑的机会,警告市场低估了 AI 热潮退却的风险。同一天,百度 CEO Robin Li 在深圳 APEC CEO 峰会上表示,如果企业无法把 AI 采用转化为真实利润,AI 可能变成泡沫,并指出基础设施投入巨大但应用普及不及预期。

Body

资金结构的变化是关键担忧:据 Fortune 引用的数据,全球 AI 融资从 2026 年第一季度的 2391 亿美元降至第二季度的 1531 亿美元,再降至第三季度的 1033 亿美元,六个月内下滑约 57%。缺口由债务填补,Morgan Stanley 预计今年全球 AI 相关债券发行接近 5700 亿美元,是去年的两倍多。风险信号已经出现:CoreWeave 2031 年到期债券的收益率在 10 月突破 16%,通常被视为财务困境信号。

Microsoft、Alphabet、Amazon 和 Meta 等巨头主要靠自有现金流支付开支,Nvidia 单季度利润达 653 亿美元,股市回调未必演变为银行危机。但债务集中在链条末端的弱信用借款人身上,2008 年危机正是从那里开始的。叠加美联储可能加息、10 年期美债收益率超过 5%,AI 建设正从股权融资转向举债融资,而借贷成本恰在上升。

Original · Gary Marcus

“I've never seen so much money chasing one opportunity with not a lot of earnings and ROI backing it up." is *exactly* what i have been saying all along. Ricardo @Ric_RTP The man who ran America's 2008 bank bailout just warned that markets are underestimating the risk of the AI boom unwinding. Hank Paulson was Treasury Secretary when Lehman collapsed. Before that he ran Goldman Sachs. And he just said: "I've never seen so much money chasing one opportunity with not a lot of earnings and ROI backing it up." He wasn't alone that day. Baidu CEO Robin Li told the APEC CEO Summit in Shenzhen that AI could turn into a bubble if companies can't turn adoption into real profits. He said the spread of AI applications has fallen short of expectations despite all the infrastructure spending. So the former head of the US Treasury and the boss of one of China's biggest AI companies said basically the same thing on the same day. But the thing that worries me most: The money funding AI is changing shape. Global AI funding fell from $239.1 billion in the first quarter of 2026 to $153.1 billion in the second quarter, then to $103.3 billion in the third, according to figures cited in a Fortune op-ed. That's a drop of about 57% in 6 months. And the gap is being filled with DEBT. Morgan Stanley projects nearly $570 billion of global AI-related debt issuance this year, more than double last year. Equity funding is shrinking, and lenders are stepping in to fill the hole. That matters because debt is how losses spread. A missed payment hits banks, insurers and pension funds that never chose to bet on AI. The cracks are already visible: In October, the yield on CoreWeave's 2031 bonds blew past 16%, a level that usually signals distress. And the price of borrowing is going UP. Fed Governor Christopher Waller said this week the Fed may need to shock people with a half-point hike in December. Markets are pricing a 100% chance of a hike next month. The 10-year Treasury yield is still above 5%. So the AI boom is switching from equity money to borrowed money at the exact moment borrowing gets more expensive. And this rhymes with 2008. Back then, the problem was loans made against the assumption that house prices would keep rising. In 2026, it's loans made against the assumption that AI demand keeps rising fast enough to pay for the data centers. And the stock market is starting to feel it. It's true that the biggest spenders, Microsoft, Alphabet, Amazon and Meta, mostly pay for this out of enormous cash flows. Nvidia just reported $65.3 billion of profit in a single quarter. Most of this debt sits with strong companies, and a correction in AI stocks wouldn't automatically become a banking crisis. But that's only true for the giants. CoreWeave's bond yields blowing past 16% tell you the risk sits at the bottom of the chain with the weakest borrowers, and in 2008 that's where the trouble started. Paulson also said governments need to work much more closely with business to cut the odds of a catastrophe or a financial crisis. When a former Treasury Secretary says that out loud in Silicon Valley, I think people should listen.

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Industry" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to US-China foundation model benchmarks and commercialization comparison and AI regulation, compliance frameworks, and sovereign compute policy. Key entities and technical topics under continuous tracking include AI bubble, Hank Paulson, Baidu, CoreWeave.

Source attribution: curated from Gary Marcus (@GaryMarcus).