Investment Grade Now Costs More Than the Government's Own Duration Risk. It Took 3 Weeks.

Investment-grade OAS printed 0.94% on 30 October against an ACM 10yr term premium of 0.9089% - +3.11 basis points, the first positive print since 2026-07-28 and the end of a 65-session run in which corporate risk was priced cheaper than sovereign duration risk. Over the same 3 weeks the 10yr fell 34.0 basis points, 32.0 of it real yield and 2.0 of it breakeven, while IG OAS widened 12.0 and high yield 35.0. The all-in cost of the riskiest borrower is +1.0 basis points. The all-in cost of the safest is -22.0.

There are two different prices for waiting, and they are quoted in the same units, which is why they get confused. The first is a credit spread: the extra yield a company pays over the government because it might not pay at all. The second is a term premium: the extra yield the government pays for borrowing long instead of rolling short, because a holder of a 10yr bond takes interest-rate risk for a decade. One prices a balance sheet. The other prices time. For most of 2026 the corporate price was the higher of the two - rationally, since a bondholder's loss given default is worse than a sovereign's. On 30 October the ordering flipped: the investment-grade index demanded 0.94% to lend to a corporation while the New York Fed's estimate of the 10yr term premium fell to 0.9089%, so the corporate premium sat 3.11 basis points above the duration premium. That inversion is the subject here, and the reason it matters is political: a sweep changes the fiscal path, the fiscal path changes how much duration the Treasury has to sell, and that is the only pipe through which an election reaches a corporate spread.

The two premia crossed, and the crossing is 3 weeks old

On 30 October the investment-grade option-adjusted spread closed at 0.94% (FRED BAMLC0A0CM) and the ACM 10yr term premium at 0.9089% (FRED THREEFYTP10, the New York Fed's estimate of the compensation for holding a 10yr note rather than rolling 3-month bills). The difference is +3.11 basis points. It is a small number and it is the first positive one since 2026-07-28, when it was +1.02. Between those 2 dates sit 65 consecutive sessions in which the market charged less for a corporation's default risk than for the government's interest-rate risk.

That run began from an unusual place. At the start of 2026 the gap averaged +20.56 basis points and it peaked at +38.73 on 2026-02-27: through the first 4 months the corporate premium exceeded the duration premium by 16 to 39 basis points, which is the ordinary ordering. May broke it. The gap crossed zero, averaged -0.2 in May, -0.7 in June, -1.1 in July, -4.9 in August and -16.0 in September, then bottomed at -26.63 on 2026-10-07. Thirty-nine of the 65 negative sessions came after 1 September.

Read the size of the 30 October move and the argument gets sharper. The gap widened +7.31 basis points in one session against a median absolute daily change of 1.48 across the 208 sessions in this file - the 4rd largest single-session move of the year. Over 16 sessions it gained +29.74, which is the largest 16-session move of the 209 windows the 2026 sample holds and the 24th largest of 729 windows going back to November 2023. This is not drift. Something repriced.

Where the 3 weeks went: duration collapsed, credit did not

Split the window from 2026-10-07 to 2026-10-30 into its parts and the mechanism is arithmetic rather than narrative. The 10yr par yield fell from 5.28% to 4.94%, -34.0 basis points. Of that, the real yield fell -32.0 basis points - 2.92% to 2.60% - and inflation compensation fell 2.0, so almost all of the rally was a real-yield move rather than a change in what the market expects inflation to do. The ACM term premium did the rest: 1.0863% to 0.9089%, -17.7 basis points. More than half the entire bond rally was the price of duration risk falling.

Credit went the other way in the same 3 weeks. Investment grade widened 12.0 basis points, from 0.82% to 0.94%, and high yield 35.0, from 3.09% to 3.44%. Excess spread - high yield less investment grade, the part that is about the weakest borrowers rather than the average one - widened 23.0 to 250.0 basis points.

Put the two halves together and the crossover stops being a curiosity. A spread that widens 12 basis points into a 34-basis-point rally is not pricing default. Default risk does not rise because the risk-free rate fell. What it is pricing is supply: the same buyers who were being handed a 35-basis-point gain on their Treasury duration were being asked to absorb corporate paper at a fixed spread, and they asked for more of it. That is the credit version of an auction tail, and it lands hardest where duration is longest and the buyer base is thinnest.

The political fact reaches this through one pipe: the funding mix

The mechanism, stated plainly. A sweep changes the fiscal path: one party holding the presidency and both chambers can pass a budget resolution that divided government could not. A larger deficit has to be funded. The Treasury then chooses how to fund it - bills, which carry no duration, or coupons, which do. That choice, not the election, is what reaches a bond price, and it reaches credit second-hand because a corporate spread is quoted over the curve the Treasury just supplied.

The choice the Treasury is making is visible in the auction data. Across the 90 days to 2026-11-02 gross settlements totalled $7,931.8bn against $7,936.1bn in the prior 90 days - a difference of -4.3bn, which is to say no change at all. Underneath that flat total the mix rotated hard: bills went from $6,703.8bn to $6,960.3bn, +256.5bn, and coupons from $1,232.3bn to $971.5bn, -260.8bn. The bill share of gross offering rose +3.3 percentage points, from 84.5% to 87.8%.

That is a quarter of a trillion dollars moved out of coupon supply and into bills while total issuance stood still, and it is the reason a sweep did not lift the term premium. There was no extra duration to absorb. The political fact arrived in a market whose long-end supply was, if anything, being withdrawn - so the price of duration risk fell with the real-yield rally instead of rising against it, and the corporate premium crossed above it almost by default.

This is also where the capital side of the agenda bites. If the Treasury is not adding duration, then the duration that institutional buyers are being asked to take is corporate duration - long-dated investment-grade paper sold to fund capital spending, and the data-centre complex is the largest single source of it. A fixed pool of long-end buyers facing unchanged Treasury coupon supply and rising corporate supply clears at a wider spread. The 12 basis points is that clearing, and it is why the widening showed up in investment grade, the part of the market that actually sells 10yr and 30yr paper, rather than only in high yield.

The all-in number a treasurer actually sees

Spreads are a market price. All-in cost is a decision, and the two diverge when the base moves. On 2026-10-07 the 10yr was 5.28% and high-yield OAS 3.09%, so the marginal high-yield borrower's reference cost was 8.37%. On 2026-10-30 the same arithmetic gives 4.94% plus 3.44%, or 8.38% - +1.0 basis points. The 10yr fell 34.0 and the borrower got nothing.

Investment grade got the rally. 5.28% plus 0.82% was 6.10%; 4.94% plus 0.94% is 5.88%, -22.0 basis points. Against the 3 November pair - 10yr at 4.95%, IG OAS 0.93%, HY OAS 3.40% - the all-in costs are 5.88% and 8.35%.

That asymmetry is the crossover expressed in the only unit that matters to a corporate treasurer. A rally in the risk-free rate that is not matched by a tightening in the spread is a transfer from risky borrowers to safe ones. The 3 weeks delivered a 35-basis-point gift to anyone holding duration and handed the weakest borrowers a flat cost of funds. This is what a late-cycle credit market looks like from the liability side: the sovereign's own risk gets cheaper while the corporate's does not.

What makes this a midterm piece and not a credit note

The temptation is to read the crossover as a risk signal - corporates deteriorating ahead of a growth slowdown. The data refuses that reading. High-yield OAS at 3.44% is 84.0 basis points off its 2026 low, and investment grade at 0.94% is tied with 03-16 for the highest print of the year, which is a level consistent with a tight market that has given back some ground, not with stress. Initial claims and the labour series in this file are calm. Nothing in the credit data says a default cycle has started.

What the data does say is that the composition of sovereign risk changed and the credit market repriced against it. The term premium is the price of the government's duration; at 0.9089% it sits at the 83.3th percentile of its 209 prints this year and the 58.4th percentile of the 9194 prints since 1990, inside a 2026 range of 0.4627% to 1.0965%. It is expensive for this year and ordinary for 3 decades. A market that thought a sweep had opened a fiscal channel would not have let it fall 17.7 basis points in 3 weeks while the 10yr fell 34.0.

So the political read is narrower and more useful than the headline one. The sweep did not reach the long end, because the funding mix kept duration supply flat and the coupon channel was closed. It reached credit, because credit is where the marginal duration now has to be sold. The instrument that carries the political information is not the 10yr yield. It is the gap between what a corporation pays and what the government pays for the same ten years of waiting.

Where this is wrong

First falsifier, on the crossover itself. Investment-grade OAS back at or below 0.89% with the ACM 10yr term premium at or above 0.9389%. That puts the gap back under -4.89 basis points and says the 30 October print was a month-end artefact: a single session's spread marking against a term-premium estimate that lags the market by a day. Both inputs are published daily and the condition is checkable within a week.

Second falsifier, on the mechanism. A term premium above 1.0965% - the 2026 high, set on 5 October - with the gap still positive. That combination would mean the fiscal path did reach duration after all, the term premium rose rather than fell, and the crossover was a spread event rather than a supply event. The mechanism would be wrong, not early.

Third falsifier, on the funding read. A bill share below 84.5% in the next 90-day window, with coupons back above $1,232bn. That would say the rotation into bills was a seasonal artefact of the September quarter rather than a funding choice, and the coupon channel is open. The fiscal feed publishes this quarterly and the desk recomputes it.

What confirms it. The gap holding positive through the 10 November 10yr auction while the indirect bidder share stays under its 2026 mean, and the bill share staying above 86.8% in the next window. That would say the corporate premium is now structurally above the duration premium, which is a statement about who is being asked to buy duration and at what price.

Notes

1. Investment-grade OAS is FRED BAMLC0A0CM and high-yield OAS is FRED BAMLH0A0HYM2, both in percent, both published with a lag: the newest print in this file is 2026-11-03, and the 30 October comparison dates are the last session where the term premium is also published. The crossover therefore describes 30 October, not today, and is dated as such throughout.

2. The term premium is FRED THREEFYTP10, the ACM 10yr estimate maintained by the New York Fed, in percent. It is a model estimate rather than a traded price, is revised, and is published with a lag; it is not comparable to a quoted spread in the sense of being an executable level. The comparison is between two estimates of the same quantity - the extra yield demanded for waiting - and the argument depends on their relative move, not on the precision of either level.

3. The gap is computed as IG OAS in percent minus term premium in percent, multiplied by 100, on 209 matched sessions in 2026 and 745 sessions back to 2023-11-06. Percentiles are computed over the sample named in the sentence, not over a fixed window.

4. Treasury levels are the 2026-10-30 par close from data/treasury_par.csv: 2yr 4.61%, 3yr 4.68%, 5yr 4.74%, 7yr 4.83%, 10yr 4.95%, 20yr 5.35%, 30yr 5.32%. The Levels line in the call block quotes the 4 November close, the newest row in the ledger: 2yr 4.58%, 10yr 4.90%, 30yr 5.28%, 2s10s 32.0 basis points. Par yields are quoted to 2 decimals because that is the precision Treasury publishes.

5. All-in cost is the 10yr par yield plus the index OAS on the same date. It is a reference cost for the index, not an offer for any borrower; a single-name deal prices off its own spread and tenor. The 3 November pair mixes a 3 November spread with a 3 November par close and is labelled as a pair.

6. Issuance figures are gross auction settlements from the Treasury Fiscal Data API, windows 2026-08-04 to 2026-11-02 and 2026-05-06 to 2026-08-03. Gross settlements count reopening and new issues alike and are not net coupon supply; no net figure is quoted here because the desk holds no source for it.

7. All 4 charts are generated by script from the series named above. Chart 1 plots the gap with the 65-session negative run shaded; chart 2 the all-in costs; chart 3 the decomposition; chart 4 the issuance mix. None is a forecast.