Near ten trillion of U.S. debt comes due and must be refinanced
Ten trillion to roll, one trillion to buy
Ten trillion to roll is the ordinary refinance. One trillion to buy is a checking account.
The United States government almost never pays its debt down the way a family pays off a house. When a Treasury security comes due, the Treasury sells a new one and uses that money to repay the old lender. That replacement is called a refinance. Over the next year a very large amount of existing government debt comes due and will have to be replaced this way. The Government Accountability Office puts the amount near $9.7 trillion for fiscal year 2026, which is about a third of all the debt the government currently owes. Round that to ten trillion and you have the number now being treated as a coming collision.
Exhibit · Roll vs cash
To roll / refinance
$10T
debt coming due · FY2026
−1/3 of all debt
TGA cash / checking
~$1T
cash on hand · Aug 27
not a buying program
Quiet note: 1pp rise ≈ 0.3% of GDP on rolls (vs 1.25% whole-stock thought experiment).
Refinance is huge. The checking account is not a buying program.
How to read this: Left is debt coming due this year. Right is cash on hand — not a purchase program.
The ten trillion is mostly short-term loans, not a pile of long-term bonds
Most of that ten trillion is not long-term bonds whose interest rate is about to jump. Most of it is Treasury bills, which are short-term loans, often for a few weeks or months. The Treasury already replaces those bills every week as they come due. That weekly replacement is ordinary operations. It is expensive when rates are high, but it is not a sudden reset of the country's entire debt.
The scare version of the same number treats it as a coupon wall. A coupon is the stated interest payment on a note or bond that lasts for years. A coupon wall would mean a mass of that longer debt coming due together, so that the government would have to reborrow for years at whatever rate lenders are charging today. That is not what the ten trillion is. If you name it a coupon wall, you will look for a rescue that could buy a wall. The actual job is replacing short-term bills, which the Treasury already knows how to do.
Exhibit · Bills vs scare
Ordinary
Mostly short-term bills
Weekly roll · weeks/months
Most of the $10T
Scare story
Not a long-bond coupon wall
Not a mass of long debt coming due together
Most of the ten trillion is weekly bill rolls, not a wall of long bonds.
How to read this: Ordinary weekly refinance vs the scare story.
The one trillion is cash in a checking account, not a buying program
The rescue being offered against that ten trillion is about one trillion dollars the Treasury could use to buy its own bonds. That trillion is not a license to run a giant purchase program. It is cash sitting in the Treasury General Account, which is the government's checking account at the Federal Reserve. On August 27 that account held about $950 billion.
A buyback, which is a different thing, is when the Treasury offers to repurchase older bonds from investors. That can help those bonds trade more easily. Cash in a checking account can pay the bills that come due this week. It is not a standing order to bid for thirty-year bonds, and it cannot absorb ten trillion dollars of new borrowing.
The Treasury does buy some of its own longer-term bonds, and those purchases have gotten larger. On August 19 it said it would at least double them in two parts of the long-term market, covering bonds with roughly ten to thirty years left until they mature. Each purchase will be at least $4 billion, from September 9 through November 4. Four billion dollars at a time is a tool for keeping those particular bonds easier to trade. It is not one trillion, and it is not ten. It cannot hold down the interest rate on thirty-year debt if that rate is rising because the government is still borrowing more than it takes in.
Exhibit · Cash / buyback
TGA
~$950B
checking account
Aug 27
Buyback
≥$4B
each op · trading tool
not a $10T sponge
Buybacks are a trading tool. They do not absorb ten trillion.
How to read this: Cash balance vs small buyback ops — not the same size as the roll.
A one-point rise in rates does not cost 1.25 percent of the economy this year
A second claim says that if interest rates rise by one percentage point, the extra cost is immediately 1.25 percent of the country's annual economic output. That would be true only if every dollar of government debt were refinanced overnight at the new rate, and only if the debt were about 125 percent of annual output. The Congressional Budget Office puts debt held by the public near 101 percent of output this year.
The cash the Treasury actually has to pay in the first year, on the debt that actually comes due and is replaced, is closer to 0.3 percent of output. The 1.25 percent figure is a thought experiment about the whole stock of debt. Mixing the two is how a year of replacing bills gets talked about as if the country had already been charged the new rate on every dollar it owes.
The danger is not this year's bill auctions. It is a 6 percent thirty-year rate that lasts
The interest rate that matters for long-term borrowing is the yield on the 30-year Treasury, which is what investors demand to lend to the government for thirty years. On Friday, August 28, that cash yield was about 5.21 percent. On the morning of August 31 it was about 5.23 percent, then about 5.26 percent by late morning. Six percent is still a step away. From Monday morning the gap is about three-quarters of a percentage point. From Friday it is a little more, about 79 hundredths. That is a large move if it happens. It is not a move that arrives automatically with the next few bill auctions.
In this series, a tail is a bad outcome that is not the usual case. The usual case is the government replacing short-term bills at today's rates, which are already high. The tail is the 30-year rate jumping to about 6 percent and staying there.
That 6 percent would not set home-loan rates one-for-one. A mortgage is priced a bit above the Treasury rate. The extra is the spread, which covers the risk that a homeowner defaults or pays the loan off early, and the cost of turning a government bond into a home loan. But the 30-year Treasury does set the terms when the government next sells long-term bonds to replace maturing ones. If that rate sits at 6 percent, new long-term public debt is borrowed at 6 percent, and mortgages rise with it after the spread. Four-billion-dollar purchases cannot stop that if the reason the rate is rising is the budget itself.
What this year is, and what it is not
Here is the distinction the rest of the piece is about.
This year, the government will replace about ten trillion dollars of old debt with new debt. Most of that replacement is short-term bills, the same weekly operation it already runs. Interest costs are high. The Treasury has about $950 billion in cash, which can pay what comes due, and it is buying some longer-term bonds at about $4 billion per operation. That is a costly year. It is not a crash, and the cash balance cannot buy the long-term interest rate down by itself.
The year that would actually break things is different. It would be the thirty-year interest rate going to about 6 percent and remaining there. Then the government would have to borrow its next long-term money at that higher rate, and home loans, which cost a little more than the government rate, would get more expensive too. Higher long-term rates would make the next round of borrowing cost more, which is the spiral. We are not in that year. We are still about three-quarters of a percentage point below it.
Ten trillion to roll is the ordinary refinance. One trillion to buy is a checking account. Six percent is the danger, and it has not happened.