The other bond borrower: how AI debt competes with US Treasuries

In 2026, the investment wave driven by artificial intelligence (AI) is not being financed with equity alone; it also runs through the bond market. Data centres and the power needed to run them are expensive, and part of that cost is borrowed. At the end of July, the bond debt of US companies stood at USD 11.5 trillion, up 7% in a year. In the same period, Treasury debt rose by 6.6%, to USD 19.1 trn; so, for every dollar of marketable government debt, there are now 60 cents of corporate debt. Both borrowers are moving at the same pace, and both are addressing the same investors.

Indeed, FOMC chair Kevin Warsh addressed the issue directly at the September Federal Reserve meeting when he said: “The so-called hyperscalers are out in the market raising funding. And so the competition for capital is real. And I think it partly explains the increase in yields.”

This issue of Simply put asks: is there a direct causal relationship between corporate bond issuance and government bond yields?

Read also: Why are long-term bond yields rising?

What is crowding out?

At any point in time, the savings seeking a home in bonds are not unlimited. Two large borrowers show up: the government and the corporate sector. If one of them asks for much more money than usual, it has to offer a more attractive yield to attract investors. That is what economists call ‘crowding out’.

The usual reading of crowding out runs one way only: the government borrows heavily, taking up additional space in the market, and as a result companies are forced to also pay more to sell their bonds. That reading is not wrong, but it is incomplete. Crowding out is a matter of competition, and competition can work in either direction. If corporate investment absorbs a growing share of available savings, it is potentially the government's funding that can become more expensive.

The question is: can corporate debt to fund AI create a reverse crowding-out effect?

What the research tells us

Friedman (1978)¹ framed the question in these terms almost fifty years ago: do public deficits push private investment out, or do they pull it in? The work that followed tried to quantify the first possibility. Engen and Hubbard (2004)² found that one extra point of government debt, measured against the size of the economy, adds around three basis points to long-term rates. Laubach (2009)³ found a similar figure for the level of debt, close to four basis points, and a much larger one when working from the expected deficit rather than accumulated debt – in the region of 25 bps per point of GDP.

Greenwood and Vayanos (2014)⁴ pin down what really matters: it is not the amount borrowed but the maturity of the bonds issued. An investor buying a ten-year bond carries interest-rate risk for ten years and asks to be paid for that service. The more long-dated bonds the market has to absorb, the higher that compensation, which is known as the term premium. Crowding out therefore works less through total debt than through the quantity of interest-rate risk introduced to the market.

The reverse direction has been studied less, but it is documented. Greenwood, Hanson and Stein (2010)⁵ show that companies adjust their issuance to the Treasury's; when the government steps back from part of the market, firms move in to fill the space. The two borrowers watch each other and respond to each other. And if companies know how to fill a gap, they also know how to crowd the room. What can analysis from the past few decades teach us about this?

When do buyers demand a higher yield to lend to the US government?

The rise in US long-term rates over recent years has come mostly through real rates. The real rate is the cost of capital: it weighs on valuations as much as on corporate and household investment. At Treasury auctions, buyers still turn up; the question is therefore not whether the government finds takers but under what circumstances those buyers demand a higher yield to absorb its paper.

Answering that question starts with separating the routine from the unexpected. The market roughly knows how much the Treasury will borrow from one quarter to the next, and that expected amount is already in prices. An auction in line with expectations does not move yields, only the excess over that habit can. The measure used to capture what the government borrows beyond its usual pace follows from this: namely, the growth of the Treasury's debt stock over three months minus its average growth over the past twelve months. We will call that gap the ‘issuance surprise’.

Read also: Higher real yields and the reaction function of central banks

Figure 1 measures how much the government yield moves when that surprise rises by one point. The calculation is run four times, with the quantity companies are borrowing at the same moment increasing each time. The result is clear. When companies issue little, a Treasury issuance surprise comes with a 14 bps fall in the yield. When firms issue the most, the same surprise comes with a four-bps rise. Between the two situations, the gap reaches 18 bps for an identical amount of government supply.

Two factors explain these differences in the change in yield. When companies are not borrowing, the economy is usually doing poorly – and that is precisely when the government borrows. Investors looking for safety move into Treasuries, with high demand resulting in falling yields despite the flow of new paper. In contrast, when companies borrow heavily, the government has to compete for available savings instead – and pay a higher yield to attract them. The same issuance therefore produces opposite effects, depending on when it happens.

FIG 1. Beta of US government yield to unexpected changes in the government debt stock, by corporate debt issuance quartile (2000-2026)⁶

The pace of borrowing

This leaves one question open: does the sensitivity really depend on the two borrowers meeting, or only on what companies are doing? Figure 2 answers this question by crossing both sides of the market. Four separate situations are defined based on whether the government and companies are borrowing faster or slower than usual.

Only one situation clearly pushes yields up: when both borrowers come to the market at the same time there is a rise of 2.7 basis points per point of surprise. When the government borrows alone, the effect turns negative again, at -9.8 basis points. The conclusion is therefore sharper than the previous one: what sets the price is not the amount the government borrows, but the presence, or the absence, of another large borrower across the table.

FIG 2. Beta of the government yield to unexpected changes in the government debt stock, by issuance regime (Low/High Gov × Low/High Corp, 2000-2026)⁷

Read also: Is central bank complacency weakening the yen?

We've been here before

This situation where both borrowers come to the market at the same time has a precedent. The Treasury's debt stock shrank by 14% between 1997 and 2000, as the budget surpluses of the time allowed the government to repay more than it borrowed. Over the same period, corporate bond debt grew by more than 40%, driven by investment in telecommunications and technology. Yet the US 10-year yield rose by more than 160 bps in 1999 alone. The government was stepping out of the market, and the cost of capital rose all the same.

A final element explains why the phenomenon is becoming visible again today. The yield gap between large-company debt and government debt has narrowed with the ratings gap between them. In the eyes of a bond investor, the two securities have grown closer – and therefore more substitutable. ‘Reverse’ crowding out does not require companies to replace the government; it is enough that they resemble it closely enough to compete for the same savings.

For an investor, the consequence is direct. Tracking the Treasury's issuance programme is not enough to anticipate long-term rates; what matters is who else shows up to borrow at the same time. As long as companies fund their investment through the bond market, government supply stands a greater chance of weighing on yields.

Simply put, corporate issuance that has become comparable in size to government issuance explains part of the rising cost of US debt.

Macro/nowcasting corner

The most recent evolution of our proprietary nowcasting indicators for global growth, global inflation surprises and global monetary policy surprises is designed to track the recent progression of macroeconomic factors driving the markets.

Our nowcasting indicators currently show:

• The global growth signal increased, driven by improvements in the US, while the eurozone and China remained broadly stable
• Our global inflation indicator edged higher, supported by increases in the US and the eurozone, while the Chinese indicator declined
• Our monetary policy nowcaster was broadly unchanged across regions, remaining in a low but gradually rising regime.

World growth nowcaster: long-term (left) and recent evolution (right)

Reading note: LOIM’s nowcasting indicator gather economic indicators in a point-in-time manner in order to measure the likelihood of a given macro risk – growth, inflation surprises and monetary policy surprises. The nowcaster varies between 0% (low growth, low inflation surprises and dovish monetary policy) and 100% (the high growth, high inflation surprises and hawkish monetary policy).