Are US Treasury bonds the "ultimate killer" of the AI bubble? Institutions warn: a 10-year yield exceeding 5% could be the trigger.

The main contributor to the current US debt surplus is the government, not corporations. This structural difference renders the traditional narrative of "corporate deleveraging" ineffective and pushes the 10-year US Treasury yield to a crucial position that will determine the fate of AI's prosperity.

Ruchir Sharma, chairman of Rockefeller International, wrote in the Financial Times that the US federal debt has exceeded $40 trillion and the yield on 10-year US Treasury bonds has risen to about 4.8%; once it "decisively breaks through 5%"—the upper limit of the range since the dot-com bubble era—the AI bubble may burst.

The crowding-out effect of high-interest government bond issuance is driving up the debt and equity financing costs for AI companies. The estimated annual revenue from AI applications is around $200 billion, a mere fraction of the over $1 trillion spent on infrastructure such as data centers. This shortfall is heavily reliant on external financing, and soaring long-term interest rates will directly impact this pace.

The current market pricing implies the assumption that "AI investment is not constrained by macro interest rates," but every major bubble in the past 300 years has ended with a significant increase in the borrowing costs of core enterprises. In this round, the source of risk has shifted from enterprises to the government, and once the 5% threshold is triggered, it may cause a narrative reversal.

Excessive debt falls on the government: the bubble structure is changing.

Looking back 300 years, every major bubble burst began with a significant increase in borrowing costs for core enterprises—the 19th-century railroad bubble and all the bubbles of the modern central bank era are no exception. Typically, during market euphoria, companies borrow heavily to bet on hot themes, causing inflation and interest rates to rise and eventually burst the bubble, after which the government intervenes to take over the debt and stimulate the economy.

The difference this time is that the government continued to stimulate the economy even as it was performing well. Since the 2020s, the US fiscal deficit has consistently accounted for about 6% of GDP, more than double the average level of previous decades. During the same period, households and businesses did not significantly increase leverage until the past year, when tech giants, after depleting their cash surpluses, began to borrow heavily for massive AI infrastructure projects—but their leverage levels remained manageable for these large companies.

Sharma argues that the substantial over-borrowing in this cycle has been piling up on government books so far, and the problem begins at the very depths of this over-borrowing.

Crowding-out effect emerges: 5% becomes the lifeline for financing.

The expansion of government debt is having a crowding-out effect on the bond market. The outstanding balance of U.S. public debt has risen to $40.05 trillion, breaking the $40 trillion mark for the first time, with interest payments reaching $1.17 trillion this fiscal year; the yield on 30-year U.S. Treasury bonds has climbed to 5.32%, a new high since 2007. Public debt interest payments have more than doubled in the past five years, exceeding 3% of GDP, setting a U.S. record and representing the fastest growth and highest level among major developed economies.

Concerns about fiscal policy, coupled with rising energy prices, have pushed up global bond yields. High risk-free interest rates, in turn, have increased the financing costs for AI companies—which constitute the largest portion of newly issued corporate bonds. Sharma has identified the 10-year US Treasury yield as a key indicator: a "decisive breakthrough of 5%" would signal the beginning of a tighter monetary policy, making financing for mega-AI projects even more difficult. When large tech companies are forced to compete for funding with government bonds offering yields exceeding 5% and corresponding inflation-adjusted returns of over 2.5%, many companies will be squeezed out of the bond market.

Expectation Gap and Valuation Threat: Constraints Not Yet Priced in by the Market

AI applications currently generate an estimated $200 billion in annual revenue, a mere fraction of the over $1 trillion spent on infrastructure such as data centers. AI companies are increasingly relying on new bond and equity financing to fill the gap, and a 10-year yield below 5% will simultaneously slow down both of these channels—a level that will also exceed the earnings yield of US stocks, historically a headwind for the stock market.

If the 10-year yield breaks through 5% before November, it would mean an increase of over 75 basis points within six months. Historically, such a surge has ended bull markets. Some analysts believe that breaking 5% is simply a return to the 1990s—when the 10-year yield remained above 5% for an entire period and the US stock market was strong. However, the US was far less reliant on debt then than it is today: that decade ended with a fiscal surplus, and now public debt is approaching 100% of GDP, with debt servicing costs far higher than then. The rise in public borrowing costs will squeeze other borrowers more quickly and severely impact the already overvalued AI market.

We need to closely monitor how quickly the 10-year US Treasury yield breaks through 5%, and the threat it poses to AI capital expenditure and US stock valuations.

Risk Warning and DisclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.