AI debt vs Treasuries

Unlock the Editor’s Digest for free

Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.

Good morning. Happy Fed day to all who observe, and good luck to Kevin Warsh. As Rob wrote the other day, we think a US rate increase makes sense, even leaving questions of central bank credibility. The big question is whether the Fed can, with rates or rhetoric, help pull Treasury yields down. The 10-year yield closed at a 19-year high yesterday. The latest fund manager survey from Bank of America, released yesterday, asked respondents to name the biggest tail risk for the market. The most popular response: a disorderly rise in bond yields, just a bit ahead of a bursting AI bubble. But why pick one when you could have both? Send us happier thoughts: [email protected]

Crowded house

Are the hyperscalers eating Scott Bessent’s lunch? More prosaically: are Big Tech companies issuing so much debt that they are leaving investors with less room in their portfolios for Treasuries? The idea often features prominently on the laundry list of reasons why Treasury yields are flying higher.

There’s something to this; bond investors and bankers have told us so. The question is the size of the effect. Lotfi Karoui of Pimco thinks it’s small:

If you go back to the last 12 months and you look at the largest hyperscaler deals, and look at the performance of Treasuries around those deal announcements, you don’t really see any abnormal returns around those days . . . Of course, that may change a year from now as you bring more debt, more variety. But for now, the market narrative has kind of got ahead of itself a little bit. 

If hyperscaler debt was really crowding out Treasuries then we would have seen a big rise in the term premium for sovereign debt, Karoui argues. And when you look at the data, there has been a bit of a pick-up in the term premium since late June — but a more noticeable uptick in the inflation expectations:

The hyperscaler crowding-out argument relies on the notion that tech bonds are a substitute for Treasuries. Are they? Meghan Robson at BNP Paribas thinks the substitution is highly imperfect, for the simple reason that the tech bonds contain spread risk and Treasuries do not. And as hyperscalers issue more debt, the risk of widening credit spreads becomes more acute:

A lot of the debt issuance we’ve seen this year and that we forecast has been from the hyperscalers issuing long-duration debt. Ten years and beyond represents over two-thirds of the debt they’ve sold. It is in the same wheelhouse of maturity or duration to some of the Treasury yields in terms of competition. 

But looking forward, our forecast for corporate supply for the first time and looking into 2027 actually exceeds Treasury supply; that’s excluding T-bills on the Treasury side, just looking at coupon [ie longer duration] issuance. Our forecast for hyperscaler issuance this year is $250bn. Next year we see a number closer to $400bn, and we don’t see demand necessarily keeping pace with that growth. So we see the floor on credit spreads as a bit higher. 

Karoui at Pimco also points out that corporate debt in general, and hyperscaler debt in particular, is not as easy to trade as Treasuries. Even with improvements in secondary market corporate liquidity over the past decade, and loads of recent hyperscaler issuance, the corporate market is much shallower than the world’s largest debt market. “You could take the most liquid investment-grade corporate bond and calculate the turnover on that, and it will be a fraction of what you can get in Treasuries,” he says.

Ed Al-Hussainy of Columbia Threadneedle argues that crowding-out should show up in the relative performance of Treasuries and interest rate swaps, but it doesn’t:

If crowding out and fiscal factors were significant, we would see cash long-end Treasury bonds (which have a term premium) perform poorly relative to interest rate swaps (which just price the Fed). But this hasn’t happened. So I’m left with more evidence for a Fed repricing driving rates more than anything else.

There are enough big, unconstrained bond managers with a mandate to buy across sovereign and corporate bonds that there is probably some crowding-out of Treasuries at the margin. But there are much bigger forces at work on yields.

One good read

Conspicuous silence.

FT Unhedged podcast

Can’t get enough of Unhedged? Keep following Rob Armstrong and Katie Martin’s incisive and unfiltered analysis on the Unhedged podcast, a twice-weekly, 15-minute lively dive into the latest markets news and financial headlines. New episodes every Tuesday and Thursday.