On Tuesday, 18 August 2026, the US federal debt crossed $40 trillion for the first time. The following morning, the US Treasury announced it was doubling the size of its bond buybacks at the long end of the curve.
Those two events happened one day apart, and there is no way to read them separately.
Errors or inaccuracies are possible. Spotted something that looks wrong? Write to me and I will correct it.
What Treasury Actually Announced
The release came out on 19 August and its title is deliberately dry: Treasury Announces Increased Sizes of Nominal Long-End Liquidity Support Buybacks Beginning September 9.
| What changes | The cap on each buyback operation |
| From | $2 billion |
| To | at least $4 billion |
| The sectors | 10-20 years and 20-30 years |
| When | 9 September to 4 November 2026 |
| What comes after | To be set at the Quarterly Refunding on 4 November |
And the official rationale, in Treasury's own words, is that the increase "reflects Treasury's desire to provide greater liquidity support in longer-dated nominal sectors where there is consistent strong sponsorship from market participants, as evidenced by the significant volume of high-quality offers Treasury routinely receives in longer-dated buyback operations."
Read that sentence again, because it is written with unusual care.
Treasury does not say yields are too high. It does not say it is intervening. It says there is strong demand to sell it paper, and that it is expanding its capacity to buy.
The justification is technical - secondary-market liquidity - not macroeconomic.
And this is not merely legal cosmetics. The US Treasury has no mandate to set interest rates; that mandate belongs to the Fed. The moment Treasury concedes that it is acting to lower yields, it steps into territory that is not its own. So the announcement is framed as an infrastructure measure, and the market read it as a policy measure.
The gap between those two readings is the whole story.
What the Market Did - and the Number That Gives It Away
These are the official closes of the US yield curve across the three relevant trading days:
| 17 Aug | 18 Aug | 19 Aug | Change on 19 Aug | |
|---|---|---|---|---|
| 3 months | 3.87% | 3.86% | 3.86% | unchanged |
| 2 years | 4.19% | 4.19% | 4.19% | unchanged |
| 5 years | 4.38% | 4.37% | 4.35% | -2 bp |
| 10 years | 4.72% | 4.71% | 4.65% | -6 bp |
| 20 years | 5.30% | 5.28% | 5.17% | -11 bp |
| 30 years | 5.31% | 5.28% | 5.19% | -9 bp |
And the next day, 20 August, the 30-year closed at 5.23% and the 10-year at 4.69% - four basis points higher in both. The two-year stayed at 4.19%, a fourth consecutive session.
The second row of that table is the proof.
The two-year yield did not move by a single basis point. 4.19% before the announcement, 4.19% after it.
And why that is decisive: the short end of the curve prices the Fed's policy rate - what the market thinks the base rate will be over the next two years. The long end prices that, plus a risk premium for inflation, for supply and for duration - the term premium.
When the short end does not move and the long end falls 11 basis points, there is no way to read that as a change in rate expectations.
What changed is the pricing of supply. The market understood there would be an additional and larger buyer for long paper - and charged less premium for holding it.
Which is precisely what Treasury says it is not doing.
And Then Thursday Arrived
Thursday, 20 August, closed with the 30-year at 5.23% - four basis points above Wednesday's close, that is, almost half the fall erased within a single day.
And during that day's session it had already touched 5.251%.
Krishna Guha of Evercore ISI put it this way: Bessent is "again showing his tactical skill as an activist Treasury secretary - hitting bond shorts with a surprise announcement of an increased buyback program on an August day with thin liquidity." And immediately after, he added that he is "skeptical that this operation will have a material impact over any more extended period," and that it "changes almost nothing in terms of the fundamentals."
The day after gave him his answer.
And Why It Did Not Hold - JPMorgan's Argument
The next day JPMorgan's strategy team, led by Jay Barry, published the sharpest reading of the move I have seen. And it is not about the size of the buybacks but about credibility.
Their wording: "Absent real fiscal consolidation, we fear the markets will view this action as lacking credibility. This could contribute to higher term premium and yields over time should Treasury become more opportunistic in its approach to debt management and move further away from its 'regular and predictable' tenet."
And the argument is powerful because it reverses the direction.
The intuitive reading: Treasury buys more long paper, net supply falls, yields decline.
Their reading: Treasury is signalling it is willing to depart from a predictable issuance schedule in order to influence prices. And anyone holding a 30-year bond prices that uncertainty too - and demands compensation for it.
That is, the very same action can lower yields for one day and raise them over time.
And the root, according to that team, is not in the secondary market at all: a deficit of 6% of GDP in an economy close to full employment. On their estimate, the funding gap over the coming fiscal years exceeds $3.5 trillion - which will require more long-dated supply, not less.
What a Treasury Buyback Actually Is
It is not money printing, and this is the point people get wrong most often.
The difference between a Treasury buyback and Fed quantitative easing is substantive, not semantic.
The Fed buys bonds with money it creates - new bank reserves. The monetary base grows. It expands the system's balance sheet.
Treasury cannot create money. It can only tax or borrow. So every dollar it spends buying back a long bond has to come from somewhere else - and in practice it comes from issuing short-dated bills.
The result is a duration swap, not an injection. Treasury buys a 30-year bond and issues a three-month bill to fund it. The quantity of debt does not change; its maturity shortens.
The closest historical parallel is not QE but Operation Twist - except that this time the institution running it is the Treasury rather than the central bank.
And the programme itself is not new. Treasury ran a buyback programme in 2000-2002, when the United States was in budget surplus and retiring debt. It was revived in May 2024 under the heading "liquidity support," with its role defined as giving primary dealers a way to sell back older, less liquid securities.
The difference between 2000 and 2026 is the context. Then it was a tool for retiring debt out of a surplus. Today it is a market-management tool, running against a $2.1 trillion deficit.
The Numbers Underneath All of This
This is the debt as at 18 August 2026, from Treasury's own daily statement:
| Total federal debt | $40,047,425,768,420 |
| Held by the public | $32,265,798,542,898 |
| Intragovernmental | $7,781,627,225,522 |
| A year earlier (29 Aug 2025) | $37,274,265,917,907 |
| Growth over the year | $2.77 trillion - 7.4% |
And this is the pace at which we got here:
| The mark | First crossed | Time from the previous mark |
|---|---|---|
| $30 trillion | 31 January 2022 | |
| $35 trillion | 26 July 2024 | two and a half years |
| $40 trillion | 18 August 2026 | two years and three weeks |
Ten trillion dollars were added in under five years - and each five trillion arrived faster than the one before it.
And on the cost side:
| Average rate on marketable debt (31 Jul) | 3.443% |
| of which: bills | 3.758% |
| of which: notes | 3.309% |
| of which: bonds | 3.442% |
| The 30-year yield in the market | 5.19% |
| Interest on the public debt, ten months | $900 billion |
| Including intragovernmental interest | $1,170 billion |
And the gap between those two bolded lines is the entire problem.
The existing debt carries 3.443% on average. New debt at the long end is issued at 5.19%.
That is, every dollar of old debt that rolls today makes itself more expensive - and it happens automatically, without anyone deciding anything, every time a security matures and is reissued.
This is not a scenario. It is arithmetic already running.
The Maturity Wall - the Figure That Explains the Urgency
I calculated this directly from the 1,087 marketable securities listed in Treasury's monthly statement of the public debt, as at 31 July 2026.
| Time to maturity | Amount | Share of marketable debt |
|---|---|---|
| Within one year | $10,482 billion | 33.3% |
| One to five years | $10,894 billion | 34.6% |
| Five to ten | $4,326 billion | 13.8% |
| Ten to twenty | $2,874 billion | 9.1% |
| Twenty to thirty | $2,877 billion | 9.1% |
| Total | $31,451 billion |
A third of the marketable debt of the United States - $10.48 trillion - matures within twelve months.
Of that, $6.99 trillion is bills, which roll every few weeks in any case. The interesting slice is the rest: $3.49 trillion of notes and bonds maturing over the coming year, carrying an average coupon of 2.49%.
They will be replaced by paper yielding between 3.86% and 4.19%.
And this is what makes the debate about "will rates come down" less relevant than it seems.
Even if the Fed cuts tomorrow, the debt rolling over the coming year will still price substantially above the coupon it carries today. The gap between 2.49% and 4% was built up over a decade, and it closes every time a security matures.
And on $3.49 trillion, every percentage point is worth about $35 billion a year.
Who Stopped Buying
On 17 August - two days before the announcement - Treasury published its international capital flow data for June.
| May | June | ||
|---|---|---|---|
| Total foreign holdings | $9.371tn | $9.299tn | -$72.1bn |
| Japan | $1.143tn | $1.116tn | -2.3% |
| United Kingdom | - | $939.9bn | -1% |
| China | $659.3bn | $633.4bn | -4% |
| Net inflows into Treasuries | $56.6bn | $6.8bn | -88% |
China's holdings fell to their lowest level since September 2008 - the month Lehman Brothers collapsed.
And net inflows all but evaporated: from $56.6 billion in a month to $6.8 billion.
On that same day the 30-year yield touched 5.31%, its highest since 2007. The next day it traded intraday above 5.33%. And the day after that, Treasury announced the buybacks.
That sequence of dates is not proof of causation - June data is known with a two-month lag and is not news in itself. But it describes well the conditions the announcement walked into.
The Part That Worries Me More Than the Yields
To fund the long-end buybacks, Treasury issues short bills. And that enlarges a slice that is already large.
| Structure of marketable debt, 31 Jul 2026 | ||
|---|---|---|
| Notes | $16,172bn | 51.4% |
| Bills | $6,989bn | 22.2% |
| Bonds | $5,489bn | 17.5% |
| Inflation-protected | $2,150bn | 6.8% |
| Floating rate | $652bn | 2.1% |
The Treasury Borrowing Advisory Committee (TBAC) recommended in 2020 that the bill share be kept in a range of 15% to 20% of outstanding debt, with flexibility to move modestly beyond it.
Today it stands at 22.2%. And the buybacks will be funded by enlarging it further.
And why a high bill share is a risk rather than a saving.
The advantage is clear and immediate: a three-month bill yields 3.86%, and a ten-year note yields 4.65%. Funding short is cheaper, and it reduces reported interest expense this year.
The drawback is that the saving is not locked in. A ten-year note fixes the cost for a decade. A three-month bill fixes it for three months - and then the government returns to the market and pays whatever it is asked.
So the larger the bill share grows, the more sensitive the federal budget becomes to short-term rates - and, indirectly, the more dependent it becomes on the Fed's decisions.
And that is the junction where a debt-management question turns into a central-bank-independence question.
And the Irony You Cannot Skip Past
In 2024, when he was still running a hedge fund, Scott Bessent was one of the loudest voices attacking then-Treasury Secretary Janet Yellen for exactly this.
He adopted and amplified the analysis by Stephen Miran and Nouriel Roubini, which coined the term "Activist Treasury Issuance" - the argument that Yellen was flooding the market with short bills to hold long yields down and ease financial conditions.
Bessent said at the time that Yellen had "taken control of monetary policy" through Treasury issuance.
Today he sits in her chair, and is expanding the tool.
I raise this not to catch him in a contradiction - a Treasury secretary who arrives in office and finds a bond market under stress faces decisions the analyst he used to be never faced. I raise it because the critique he himself formulated in 2024 is the most precise critique available of what was done on 19 August 2026 - and it did not become less true because the person making it changed seats.
And on That Exact Day, the Fed Published Its Minutes
On 19 August the Federal Reserve released the minutes of the FOMC meeting of 28-29 July. And it is an unusual document.
| The decision | Hold at 3.50%-3.75% |
| The vote | 9 to 3 |
| The dissenters | Hammack (Cleveland), Kashkari (Minneapolis), Logan (Dallas) |
| What they wanted | A 25 basis point increase |
Three dissenters, and all of them in the same direction - upward.
This is the first time since September 2016 that three committee members have dissented with a unified view on the direction of rates.
And it is only the second meeting under Chair Kevin Warsh, who has removed forward guidance from the committee's statements.
What the minutes say on inflation: it "remained elevated" relative to the 2% objective, and price increases were "broad based, spanning various categories of goods and services."
And what they say on yields: "Nominal Treasury yields rose 25 to 30 basis points, driven by corresponding increases in real interest rates."
Real rates, not inflation expectations. That is, the market demanded more genuine compensation to hold the paper.
And here is the situation that defines this moment:
The Fed is holding rates high, and three of its members want them higher still. Treasury, on the same day, acts to bring long yields down.
The two arms of US economic policy are pulling in opposite directions - in public.
And the Question Underneath It All: Does This End in a Crisis
Here it matters to separate two entirely different risks, because they get mixed together constantly.
The first is a slow repricing. A higher discount rate compresses multiples, particularly in long-duration assets. That is erosion, not a crisis, and it is already happening: the 30-year has traded above 5% on 27 days in 2026, 12 of them consecutively - the most since 2007. And unlike 2007, the Fed's policy rate is 150 basis points lower today - meaning the compensation demanded to hold long paper is greater than it was at the onset of the subprime crisis.
The second is an accident in the machinery, and it is quantifiable.
| From a Fed note, 22 June 2026 - data through September 2025 | |
|---|---|
| Hedge fund long Treasury positions | $2.4 trillion |
| of which the basis trade | $830 billion - 35% |
| of which swap spread arbitrage | $305 billion - 13% |
| Repo borrowing funding those positions | $3.0 trillion |
And the Fed's own wording: "the combination of large scale, high concentration, and elevated leverage creates the potential for systemic stress if multiple strategies face simultaneous pressure."
And it has already happened. The same note describes how in April 2025, around the tariff announcements, about $60 billion - 20% of swap spread positions unwound within days, in conditions the Fed calls "strained."
The mechanism is a loop: the bond falls, margin requirements rise, selling is forced, yields jump, margin rises again.
And what makes this an equity-market matter and not only a bond-market one: in such a scenario the government bond stops functioning as a hedge and becomes the source of the shock. A portfolio holding both stocks and bonds rests on the assumption that when one falls the other rises - and in 2022, and on days in March 2020, both fell together.
And on the other side, three things that restrain that scenario:
The United States borrows in its own currency. There is no default event here in the sense that exists for an emerging market; the adjustment mechanism is inflation and the currency. The Fed has tools it has already used - a standing repo facility, and purchases in extremis. And at 5%, buyers appear: pension funds and insurers need that yield to match long-dated liabilities.
Those two sides define the range. On one, large, concentrated, repo-funded leverage; on the other, a currency of its own, a central bank with tools, and demand that returns at a price. And what will decide whether this is erosion or an accident is not the level of the debt - but whether the leverage sitting on top of it unwinds gradually or all at once.
What Could Happen From Here
I prefer to describe a range of states rather than guess at one. These are the branches that look significant to me.
The first path: the buybacks work, and calm holds.
Demand for long paper returns, the term premium compresses, and the 30-year settles below 5%. The condition for that is that the deficit begins to narrow - because otherwise this is buying time.
What would signal it: strong demand at the 20-year and 30-year auctions, and foreign flows returning to May's levels.
The second path: the buybacks keep growing, and become permanent.
The 19 August announcement runs only to 4 November. If the cap is raised again at the coming Quarterly Refunding, the tool will have moved from a liquidity instrument to a pricing instrument.
BMO's head of Treasury trading warned that continuing to scale up buybacks risks giving Treasury an "overarching presence" in the market and undermining primary dealers' ability to function in periods of stress. That is a real risk: when the largest buyer is also the issuer, price discovery weakens.
The third path: the bill share keeps swelling, and the budget becomes Fed-dependent.
This is the path I pay the most attention to, because it is the only one already happening and requiring no new event. The shorter the funding, the faster each Fed decision feeds into the federal budget - and from there it is a short step to open political pressure on the Fed to cut rates for budgetary reasons.
The term for this is fiscal dominance, and it is not theoretical. What would signal it approaching: a bill share continuing to climb above 22%, alongside political statements explicitly linking rates to the cost of the debt.
The fourth path: foreign demand does not return.
China at 2008 lows, Japan trimming, and net inflows down 88% in a month. If the trend persists, someone has to fill the gap - and it will be pension funds, US banks and households. They will demand a yield for it.
What would signal it: the July and August TIC data, due in mid-September and mid-October.
What I Am Watching
| 4 November | The Quarterly Refunding - where it is decided whether the cap holds, rises or reverts |
| 9 September | The first operation at the new size - and how much paper is actually offered into it |
| 27-29 August | Jackson Hole - Warsh's first address as Chair, next week |
| Repo market strains | The one leading indicator of an accident in the machinery |
| The bill share | 22.2% today; above 25% is a different story |
| The 30-year against the two-year | 100 basis points on 19 August |
| The 20 and 30-year auctions | That is where real demand is measured, not in the secondary market |
| TIC data | Whether foreign flows come back |
הזווית שלי
דעה אישית של אילן אברמוב - לא ייעוץ ולא המלצה
What troubles me in this story is not the level of yields, but what it took to move them for one day.
The US Treasury doubled a tool, surprised the market on a thin-liquidity day, and got nine basis points. By the next morning six of them were back. That says something about the relative scale of the forces: $4 billion per operation against $31.45 trillion of marketable debt is 0.013% of it.
And the central insight I take is from the row that did not move. The two-year stayed at exactly 4.19%, before and after. That is, the market did not change its mind about the Fed - it changed its mind about supply. Which means what is being priced here is not monetary policy but the quantity of paper that has to find a buyer.
And what I try to hold onto is the distinction between a symptom and a disease. A 5.19% 30-year is a symptom. The disease is that a $2.1 trillion annual deficit produces paper faster than existing demand absorbs it - and any tool that addresses the supply side of the secondary market, rather than the issuance side, is treating the symptom.
And here I want to be fair to the other side of the argument. There is an entirely reasonable reading that says Treasury is doing exactly its job: a dysfunctional government bond market is a systemic risk, targeted liquidity intervention is legitimate and practised everywhere, and doing nothing would have been more dangerous. That argument is strong, and I do not dismiss it.
The real disagreement is not whether the action is justified, but how you know when it has become something else. And that line runs through exactly one place: whether the tool is deployed according to liquidity conditions, or according to the level of yields. The first is infrastructure. The second is interest-rate policy through the back door, by a body with no mandate for it.
Which is why 4 November interests me far more than 9 September. If the cap returns to $2 billion after yields have calmed - it was a liquidity tool. If it rises again because yields have not calmed - it is no longer one.
And a last word on orders of magnitude. The debt crossed $40 trillion on 18 August. At the pace of the last year - $2.77 trillion in twelve months - the $45 trillion mark is less than two years away. I do not know what yields will be then, and nobody does. But I do know that the average rate on that debt today is 3.443%, and that every security that rolls moves it closer to today's yields.
That should not frighten anyone. It should be the starting point of every discussion about long-dated government bonds.
(It is important to stress: this is my personal opinion only, and nothing herein constitutes a recommendation to take any action.)
