Is the relentless bid under long yields the debt? Forty-seven years of auction records, 237 years of debt history and 36 years of the daily curve point that way, but not through the channel almost everyone names.
Something structural is happening in the long end, and it is not news-driven. But it is not inflation either. Since the end of 2022 the 10-year has risen 80bp, and 83bp of that is real yield. Ten-year inflation expectations fell 3bp over the same stretch. The market is not demanding protection from inflation. It is demanding to be paid more to own duration at all.
The 30-year real yield closed at 3.00% on 14 Aug, tied for the second-highest reading in the series Treasury has published since February 2010. The record is 3.03%, set two weeks earlier on 31 July 2026. All 42 sessions above 2.75% have occurred since 27 March 2026.
That is the whole report in two numbers. What follows is the evidence chain, and what it means for pricing loans.
A note on what this argument is. The data below establish that high real yields, record issuance, a short maturity profile and a changed buyer base all coexist. They do not contain a counterfactual, so they cannot prove debt supply caused the yield level. Read the title as the hypothesis this report documents and tests, not as a verdict the evidence closes.
TLT printing an all-time low is not the same statement as "yields are at all-time highs." It is a statement about what Treasury issued in 2020.
Take the all-time-low premise as given. It comes from the fund's price history, which sits outside this report's Treasury datasets. The yield does not explain it. The 30-year closed at 5.25%, tied for the 17th-highest of the 5,133 available closes since TLT launched in July 2002; the record is 5.35%, set on 12 June 2007. (A further 889 dated sessions in that window have no 30-year value at all, principally the 2002–2006 suspension of the bond.) On yield alone, TLT should be near its 2007 low, not through it.
It is through it because of the collateral. There is $2.67T of Treasury paper with 20+ years remaining, and 43.8% of it carries a coupon under 3%, with $0.43T under 2%. Treasury sold that paper into the 2020–21 rate floor, and a near-zero coupon at a long maturity is the highest-duration instrument the market produces.
| Security | Outstanding | Yrs left | Price @ 2.00% | Price @ 5.25% | Change |
|---|---|---|---|---|---|
| 1.250% of May 2050 | $73.6B | 23.8 | $85.87 | $46.06 | −46.4% |
| 1.375% of Aug 2050 | $89.0B | 24.0 | $88.13 | $47.47 | −46.1% |
| 1.625% of Nov 2050 | $85.5B | 24.3 | $92.82 | $50.60 | −45.5% |
| 1.875% of Nov 2051 | $90.2B | 25.3 | $97.53 | $53.08 | −45.6% |
| 2.000% of Aug 2051 | $90.0B | 25.0 | $100.00 | $55.04 | −45.0% |
Those five bonds alone are $428B of face value trading around half of par. Nothing about the yield level is unprecedented, the coupon was. That is why the fund, and not the yield, is the thing printing a record low.
TLT is a bad thermometer for "how high are rates," because its price mixes today's yield with 2020's coupon decisions. For reading the actual pressure on mortgage pricing, watch the 30-year real yield and the 10-year directly. Both are tracked on the dashboard.
Every nominal Treasury yield is the sum of a real yield and an inflation expectation. Treasury publishes both. Split the move and the argument ends.
If the bond market feared inflation, breakevens would be widening. They are not. They have been effectively motionless for four years while the real yield has done all of the work, a 346bp climb off the 2020 floor.
Monthly, Jan 2003 – Aug 2026. Real yield is the 10-year TIPS constant-maturity rate; breakeven is nominal minus real.
| Date | Nominal | Real | Breakeven | 30Y nominal | 30Y real |
|---|---|---|---|---|---|
| 4 Aug 2020 (the low) | 0.52% | −1.05% | 1.57% | 1.19% | −0.46% |
| 31 Dec 2021 | 1.52% | −1.04% | 2.56% | 1.90% | −0.44% |
| 30 Dec 2022 | 3.88% | 1.58% | 2.30% | 3.97% | 1.67% |
| 19 Oct 2023 | 4.98% | 2.49% | 2.49% | 5.11% | 2.55% |
| 18 Sep 2024 (first cut) | 3.70% | 1.58% | 2.12% | 4.03% | 1.90% |
| 31 Dec 2025 | 4.18% | 1.93% | 2.25% | 4.84% | 2.62% |
| 14 Aug 2026 | 4.68% | 2.41% | 2.27% | 5.25% | 3.00% |
Since 4 Aug 2020: +416bp nominal = +346bp real and +70bp breakeven. Since 30 Dec 2022: +80bp nominal = +83bp real and −3bp breakeven.
Monthly, Feb 2010 – Aug 2026. TIPS 30-year constant maturity.
Because the news moves the two things that are not moving, and cannot reach the one thing that is.
Economic data reaches long rates mainly through two channels: it changes the expected path of Fed policy, and it changes expected inflation. Breakevens show the second channel is inert. The first is worse than inert. It has been running backwards.
Since the Fed's first cut, announced 18 September 2024, the 3-month bill has fallen 98bp. Over the same window the 30-year rose 122bp. That is a 220 basis point divergence between the rate the Fed sets and the rate that prices a mortgage.
Monthly, Jan 2024 – Aug 2026. Dashed marker is the first cut of the cycle, announced 18 Sep 2024.
| Tenor | At the first cut | Now | Change |
|---|---|---|---|
| 3-month | 4.84% | 3.86% | −98bp |
| 2-year | 3.61% | 4.17% | +56bp |
| 5-year | 3.47% | 4.36% | +89bp |
| 10-year | 3.70% | 4.68% | +98bp |
| 30-year | 4.03% | 5.25% | +122bp |
The 3-month-to-30-year spread went from −81bp (inverted) to +139bp. The curve did not un-invert because the front end normalized. It un-inverted because the back end broke away.
Treasury's own daily ledger, back to 1790.
Total public debt outstanding stands at $39.93T, of which $32.20T is held by the public. That second figure is the one the bond market has to absorb. It has grown by $15.03T since the end of 2019, an average of $2.27T per year.
The debt added since 31 December 2019, $16.73T, is larger than the entire national debt accumulated from the founding of the Republic through 15 March 2013. Two hundred and twenty-three years of borrowing, matched in six and a half.
Annual, log scale, the only scale on which two centuries and $39T fit in one frame. Each gridline is a 10× increase.
The deficit is what Treasury borrows. Gross issuance is what Treasury actually has to sell, and the gap between them is enormous.
In calendar 2025 Treasury ran 444 auctions and sold $30.75T of securities. In 2019 that figure was $12.00T; in 2007, $4.49T. Most of that is not new borrowing. It is the same debt being sold over and over, because 84.3% of gross issuance is Treasury bills that mature in a year or less and must immediately be sold again.
Calendar years 1998 – 2026 (2026 through completed auctions on 17 Aug). Total accepted at auction, all securities. Stacked: bills vs. notes, bonds and TIPS. Reopenings count as gross issuance; excluding cash-management bills lowers the 2025 bill share to 83.9%.
| Matures in | Amount | Share | Avg coupon (ex-bills) | Market rate today | Repricing gap |
|---|---|---|---|---|---|
| Under 1 year | $10.48T | 33.3% | 2.77% | 3.98% | +121bp |
| 1 – 2 years | $3.67T | 11.7% | 2.90% | 4.17% | +127bp |
| 2 – 3 years | $2.93T | 9.3% | 3.14% | 4.24% | +110bp |
| 3 – 5 years | $4.29T | 13.6% | 3.02% | 4.36% | +134bp |
| 5 – 10 years | $4.33T | 13.8% | 3.43% | 4.68% | +125bp |
| 10 – 30 years | $5.75T | 18.3% | 3.15% | 5.25% | +210bp |
One third of the entire marketable debt, $10.48T, 33.3%, including $353B of floating-rate notes, turns over inside twelve months. On a weighted-average basis every bucket reprices upward, though roughly $1.1T across 23 individual securities already carries a rate above the 3.98% one-year replacement rate and would reprice down.
Weighted average maturity is 70.0 months, statistically unchanged from 69.4 months in 2015 and 69.9 in 2021. Treasury has held the maturity profile flat while the stock of debt went from $12.80T to $31.45T. Holding the duration constant while the principal grows 2.5× means the absolute volume of paper hitting the market for refinancing has grown 2.5× too. That is the supply the long end is choking on.
This is the part that makes the rest structural rather than cyclical.
The average interest rate Treasury pays across all outstanding debt bottomed at 1.556% in January 2022, the lowest in the history of the series. It is 3.447% today. It is still far below what Treasury pays on new money, which means the average keeps climbing mechanically as old paper rolls off, even if the market never moves again.
Monthly, Jan 2001 – Aug 2026. Average rate across all interest-bearing debt vs. the 10-year constant maturity.
FY2010 – FY2025 actual; FY2026 annualized from ten months of Treasury's fiscal-year-to-date figures. Gross interest on all public issues plus government account series.
| FY | Receipts | Outlays | Deficit | Gross interest | Interest / receipts |
|---|---|---|---|---|---|
| 2015 | $3.25T | $3.69T | $0.44T | $402B | 12.4% |
| 2019 | $3.46T | $4.45T | $0.98T | $575B | 16.6% |
| 2022 | $4.90T | $6.27T | $1.38T | $719B | 14.7% |
| 2023 | $4.44T | $6.14T | $1.70T | $883B | 19.9% |
| 2024 | $4.92T | $6.74T | $1.82T | $1,133B | 23.0% |
| 2025 | $5.23T | $7.01T | $1.78T | $1,220B | 23.3% |
| 2026 (annualized) | $5.38T | , | , | $1,404B | 26.1% |
Through ten months, FY2026's deficit of $1.799T has already exceeded the full-year FY2025 deficit of $1.775T, with two months still to run.
Take all $31.45T of marketable debt. Reprice every security at the yield its own maturity commands on today's curve, nominal debt on the nominal curve, TIPS on the real curve, floating-rate notes at the 13-week bill index plus their spread. Assume rates never move again and Treasury never borrows another dollar. The interest bill goes from $1,017B to $1,306B.
That is +$289B a year, a 28% increase, the arithmetic of paper issued at 2020 coupons rolling into 2026 yields. The short end of the real curve is unpublished below five years; extending it two different ways moves the answer only between +$287B and +$290B. Every additional 100bp across the stock adds a further $315B.
This is a constant-curve scenario, not an obligation. Future curves, issuance mix, buybacks and refinancing choices all move it. It is the answer to one narrow question, what does today's stock cost at today's prices, and nothing more.
The auction record does not support a buyers' strike. It supports something more durable and, for anyone pricing a mortgage, more important.
The popular version of this story says foreign buyers are walking away and auctions are on the verge of failing. Treasury's own auction results say otherwise, and the distinction matters enormously for planning.
Across all 2026 30-year auctions including reopenings, the bond averaged a 2.42 bid-to-cover. Indirect bidders, the category that captures foreign central banks and overseas accounts, took 57.0% of 30-year issuance versus 50.7% in 2019. And primary dealers, the firms legally obliged to bid, whose share rises when nobody else shows up, took just 9.2%, the lowest annual share in the bidder-allotment record, which begins in 2009. The 10-year tells the same story: 61.1% indirect, 8.6% dealer.
Share of total accepted, by calendar year, 2009 – 2026. Nominal 30-year auctions including reopenings, keyed on original security term. A rising dealer share is the classic distress signal.
This is a repricing, not a strike. Nobody is refusing to fund the United States. Buyers are pricing duration risk honestly for the first time in fifteen years. That is worse news than a failed auction, because a failed auction gets fixed in a week and a re-rated term premium does not.
Underneath the healthy headline, though, the composition of the buyer base has shifted in a way that matters. Foreign holdings hit an all-time dollar high of $9.27T in December 2025, and yet the foreign share of publicly held debt has collapsed from 49.3% in April 2008 to 30.1%. Foreigners kept buying; they simply could not keep pace.
Annual, 2000 – 2025. Treasury International Capital data over Debt to the Penny.
| Holder | Dec 2025 | Change vs 2020 | Character |
|---|---|---|---|
| Foreign official (central banks) | $3.878T | −7% | Price-insensitive |
| China, Mainland | $0.684T | −36% | Price-insensitive |
| Japan | $1.185T | −5% | Price-insensitive |
| United Kingdom | $0.863T | +96% | Fund / custodial |
| Cayman Islands | $0.422T | +89% | Fund / custodial |
| Belgium | $0.477T | +88% | Fund / custodial |
| France | $0.369T | +231% | Fund / custodial |
| Canada | $0.468T | +291% | Fund / custodial |
Foreign official holdings shrank 7% in nominal dollars while debt held by the public grew roughly 43%. Every dollar of growth came from the fund and custodial domiciles, the Caymans, Luxembourg, London, Dublin, which is where leveraged relative-value money books its Treasury positions.
Central banks buy Treasuries because they need reserves; they are largely price-insensitive. Funds buy Treasuries because a trade is profitable, and they demand compensation to show up. The custody data are consistent with the marginal holder shifting from the first kind of buyer toward the second, and that shift is a plausible contributor to a higher required yield and to sharper liquidity-driven selloffs of the kind seen in March 2020.
Stated as a hypothesis on purpose. TIC reports the location of the custodian, not the beneficial owner, and carries nothing about who is leveraged or why they bought. The country totals above are solid; the story about what kind of investor they represent is an inference that would need separate flow and ownership data to confirm.
The uncomfortable part: the spread trade the market has been waiting on has already happened.
The 30-year conforming rate stands at 6.647% against a 10-year at 4.68%, a spread of 197bp. Take 170bp as the reference "normal" spread: that figure is a widely used rule of thumb, not a number measured in this report, there is no historical mortgage series in the data behind it, so treat it as an illustrative anchor rather than a computed long-run mean.
Against that anchor the well is nearly dry. Full normalization to 170bp from here is worth 27bp, about $72 a month on a $400,000 loan. Which means the mortgage rate is now very close to a pure function of the 10-year, and the 10-year is a function of the real yield, and the real yield is a function of everything in sections 4 through 7.
| 10-year | @ 150bp | @ 170bp (ref) | @ 197bp (today) | @ 220bp |
|---|---|---|---|---|
| 4.00% | 5.50% | 5.70% | 5.97% | 6.20% |
| 4.25% | 5.75% | 5.95% | 6.22% | 6.45% |
| 4.50% | 6.00% | 6.20% | 6.47% | 6.70% |
| 4.68% (today) | 6.18% | 6.38% | 6.65% | 6.88% |
| 5.00% | 6.50% | 6.70% | 6.97% | 7.20% |
| 5.25% | 6.75% | 6.95% | 7.22% | 7.45% |
| 5.50% | 7.00% | 7.20% | 7.47% | 7.70% |
On a $400,000 loan, principal and interest runs $2,398/mo at 6.00%, $2,567 at 6.647%, and $2,661 at 7.00%. Each 25bp is roughly $67/month.
Three things follow from the data, and none of them require a forecast:
Falsifiable triggers, all of them computable from data the dashboard already pulls daily. If none of these fire, nothing has changed regardless of the headlines.
Monthly CPI prints. Ten-year breakevens have moved a net −3bp in three and a half years, through every inflation scare in that window. Trading the long end off CPI has been a losing read since 2022, and the data says it still is.
The government spends more than it collects, so it borrows the difference by selling IOUs called Treasury bonds. Back in 2020 it sold an enormous pile of them promising to pay only about 1% a year, because that was all anyone expected to earn back then. Those IOUs are still out there, and they still only pay 1%, which is why anybody holding one now can only sell it for about half of what they paid, and why a fund full of them keeps hitting new lows.
Here is the part that matters for your mortgage. The government now has to borrow a mountain of new money every year and replace all the old cheap IOUs as they come due, about a third of the entire pile gets replaced every twelve months. Buyers will absolutely take all of it, but only if they get paid a lot more than 1%. That higher price is set in the same market that decides mortgage rates, so when the government has to pay more to borrow, so does everyone buying a house. And notice what is not happening: people are not worried about prices rising fast, that expectation has not budged in four years. They just want more money to lend for thirty years to a borrower this size. That is why good economic news keeps arriving and mortgage rates keep not falling.