Two Campaigns, One Balance Sheet

On September 15, the Treasury ran a routine sale of 20-year bonds and got back a number that maturity had never printed. The indirect bid — the bucket where foreign central banks and funds buy — came in at 52.5%, the weakest on record, and the bonds cleared at 5.42% with the largest tail since 2024, the widest gap since then between the yield traders expected and the yield it took to clear the sale. Days earlier, Treasury Secretary Scott Bessent had told Congress the auctions were succeeding ^[1]. That number is the question here: who still shows up for this paper, and at what price, when one Treasury is carrying two campaigns at once. The first is the war. Harvard's Linda Bilmes is flat about where this one ends.

I am certain we will spend $1 trillion for the Iran war. — Linda Bilmes

Current spending runs around $2 billion a day, and the Pentagon has asked the White House for another $200 billion on top ^[2]. The second campaign sits in a different drawer but on the same account. Through the Commerce Department, Washington now holds 10% of Intel, 15% of MP Materials, a golden share with veto power in U.S. Steel, and an agreement to own up to 20% of Westinghouse ^[3]. Senator Ted Cruz objects on principle.

I don't think that this is going to be sort of a socialist takeover. — Kevin Hassett

For this piece the stakes are entries, not a verdict — a second ledger of obligations that has to be financed, not just built. The fusion shows first where the money is. Analysts describe government debt and what they call "hyper stellar AI issuance" now competing for the same global capital pools ^[4]. You can see it in the bond ladder itself: Chase index data shows TSMC's bonds trading at lower yields than Meta's, and SK Hynix only nine basis points above Amazon, investors pricing some U.S. tech debt as riskier than emerging-market corporate debt ^[5]. The Gulf sovereign funds that spent decades recycling oil money now hold $5 trillion, and AI firms are among the targets it is pointed at ^[6]. The old lane, repurposed. The second place is the inflation reading. Fed Governor Michael Barr names the two pressures in one sentence.

we’re going to have to grow our way out of this. — Scott Bessent

Chair Kevin Warsh says the same thing in plainer terms: the buildout is hitting the demand side of the economy before it hits the supply side. That is why the Fed's answer to both campaigns has been the same lever — the September hike to 3.75–4.00%, the first in three years, with CPI at 3.7% ^[7]. The third place is the lender base. Net interest has crossed $1.05 trillion a year ^[8], and the 10-year has touched its highest level since 2002 ^[9]. Treasury is managing the strain with the tools in hand: shifting into short-term bills, floating a revival of Operation Twist or outright yield-curve control, while Warsh has proposed a formal Treasury–Fed accord to coordinate what gets issued ^[8]. There is one dissent buried in all this. Treasury's own career analysts drafted a July report warning the AI market is forming a bubble with systemic risks, worse-entangled than the dotcom era. The administration dismissed it as unvetted, and Bessent's answer was the China risk — that the only real danger is China getting there first ^[10]. None of this is a rupture, not yet. The dollar strengthened in early October, precisely because war-driven oil prices and rising yields sent money to the haven ^[11]. Goldman's credit trader Jeffrey Papai is buying the dip, but he gives the calm a shelf life of about two months — about when the paused AI borrowing cycle switches back on ^[12]. So the strain reads as a slope, priced in bid and yield, not a break. The next 20-year sale is the place to watch: if the indirect share snaps back toward its norm, or the Fed pauses while both campaigns keep running, the fusion hasn't bitten. Until then, the auction room is where the fusion gets its regular mark-to-market.