Minus Thirty Basis Points

MEMORANDUM
TO: Investment Committee
FROM: Office of the CIO
DATE: Tuesday, 6 October
RE: Monday's record, and why I will not be signing the ticket

---

The Nasdaq Composite closed at 27,477.31 on Monday, up 1.05%, a record. Nvidia added 2.1% to touch roughly $5.76 trillion in market value. The S&P 500 finished at 7,773.95, a rounding error away from its own high. Several of you have asked whether we are being precious about all this. Here is the arithmetic.

In the same session the 10-year Treasury touched 5.349%, the highest print since 2002, and settled near 5.31%. The 30-year closed around 5.67% after brushing 5.70%. The 20-year hit 5.717%, a fresh 52-week high. Equities spent the day celebrating Friday's payrolls number, a miserable +29,000, which gutted the case for an October hike. CME futures now price about a 20% chance, down from roughly 70% a week ago. The bond market read the identical number and sold. The 2-year dipped to 4.693% on the release, then closed Friday at 4.837%, five basis points higher on the day.

I have seen this film, though with the reels in the wrong order. Between June 2004 and June 2006 the Fed hiked seventeen times and the long end sat on its hands, a behavior Greenspan filed under "conundrum." This Fed has delivered a single hike, on 16 September, to 3.75–4%. Since then the 10-year has climbed from about 5.04% to 5.31%. In 2005 the puzzle was a long end that refused to rise. In 2026 it is a long end that refuses to fall, and the weak-jobs trade that was supposed to fix it has now failed twice in a week.

The premium, overdrawn

On Friday the 25th, the WSJ's Birinyi-sourced table had the S&P 500 at 19.94 times forward earnings, an earnings yield of 5.02%, against a 10-year at 5.167%. Hana Securities, using slightly different inputs, put the gap at three basis points, the thinnest since 2004. Since that Friday the S&P has gained about 0.4% and the 10-year has gained about 14 basis points. Hold forward estimates flat, which is generous to the bulls given that nobody has been revising anything up for a month, and the multiple is now about 20x, the earnings yield just under 5.0%, and the premium for owning equity over a government bond is minus thirty basis points. Call it arithmetic done on a napkin. The napkin does not look good.

It gets worse down the cap structure. The Nasdaq-100 sat at 24.15x, an earnings yield of 4.14%. The Russell 2000 sat at 28.39x, or 3.52%, which means the index everyone buys as a rate-cut proxy yields about 180 basis points less than a risk-free bond that is no longer in a hurry to fall.

For the historians on the committee: the S&P made its pre-crisis peak in October 2007 with the 10-year near 4.5%, and equity still paid a positive premium over it. Today's market offers a thinner cushion and a taller hurdle. I will resist drawing the obvious line from there to 2008, because a thin premium is a statement about expected returns over years. It says nothing about next month. The gap in 2004 was thin too, and the market kept going for three more years.

The best case against me

Fairness requires it. The last time the 10-year went above 5%, in October 2023, the S&P fell 10.3% and the Nasdaq 12.3%. This time, as of 30 September, the drawdowns were about 3.2% and 4.1%, and the Nasdaq is now at a record. There is no recession scare in the tape. Hyperscaler capex is forecast near $1.1 trillion next year, up 31%, and a rate shock cannot easily reach spending that is funded on that scale. Earnings growth, in other words, has to do the work the premium no longer does. It might.

Here is the part of the bull case I have not seen anyone price. The same hyperscalers that underwrite the earnings are now close to 10% of US investment-grade issuance, raising money at 6–7% and sustaining capex well beyond free cash flow. The cushion under the equity market and the weight on the bond market are one set of balance sheets. If duration keeps cheapening because they keep borrowing, the discount rate applied to their own cash flows keeps rising. Equity desks and credit desks are each holding half of a single loop and calling it a diversified view.

What oil did not do

Brent slipped 1.9% on Monday to about $100 as Middle East exports picked up and the G7 pledged more supply. In the old regime that headline buys you ten basis points on the long end. On Monday it bought nothing. Cheaper oil failed to rally bonds, a soft labor print failed to rally bonds, and a Fed sitting still failed to rally bonds. What remains is the thing no central banker can talk down: the price of absorbing the paper, whatever the inflation print says.

The calendar

The September FOMC minutes arrive Wednesday. A December hike is largely priced, and sixteen of eighteen participants penciled in at least one more increase last month, so the interesting question is how many of them still mean it after a +29K payroll. Bank earnings begin next week. Delta and PepsiCo report before then. None of them will move the 10-year, but each will tell us whether the earnings number carrying the whole structure is still intact.

Recommendation

I am not asking anyone to short a record, and I would be unwise to try. I am asking for something duller. Before the next data release, the committee should write down the forward earnings figure that makes a 5.3% risk-free rate and a 20x multiple coexist, and how many quarters of 30%-margin growth it takes to get there. If we can state it, we own an informed bet on growth. If we cannot, we own an option we never priced, and the bond market has just told us its strike.

Office of the CIO