Two numbers that cannot both be true
Everyone knows the Federal Reserve sets interest rates. It's the one piece of economics that makes the evening news. Rates too high? The Fed cuts. Borrowing gets cheaper. Simple.
Except look at what actually happened over the past two years.
The Fed pushed its rate down 1.7 percentage points — 170 A basis point is one hundredth of a percentage point. Rates move in amounts too small for whole percentages to describe, so the market counts in these instead: 100 bp = 1%. It's the unit on the tape at the top of this page., in the market's own shorthand. The 30-year Treasury went up. These two lines are supposed to move together. Watch them come apart:
View as table
In early 2024 the Fed's rate sat above the 30-year — an inverted curve, the classic recession warning. Today it sits far below it. The Fed let go of the rope and the long end floated up on its own.
If the Fed controls interest rates, why did the loan you actually care about get more expensive?
The answer is that the Fed doesn't control interest rates. It controls one rate — the overnight one, what banks charge each other to borrow until tomorrow morning — and hopes the rest follow. Right now they aren't following.
Everything else is set somewhere the Fed doesn't sit: an auction room, and the enormous market that trades around it. To understand what's happening to your mortgage you have to go into that room. It takes about fifteen minutes. Meet Peter.
Anatomy of a thirty-year promise
Peter has $1,000 and doesn't want to gamble it. So he lends it to the United States government.
In exchange he gets a contract with three numbers on it, and that contract is the bond:
$1,000
What the government owes him back at the end. Also called par. Printed on the contract; never changes.
4% — $40 a year
The annual interest, calculated on the face value, every year, no matter what happens to anything else. Fixed at birth.
30 years
When he gets the $1,000 back. Until then he collects $40 a year — $1,200 in total.
That's the whole product. A bond is a loan you can resell.
Two markets, not one
There are two different places a bond can change hands, and the difference between them is where all the drama lives.
The primary market
Where Peter got his bondThe government auction. New bonds, straight from the Treasury, sold roughly weekly. This is the government actually borrowing money.
Peter paid $1,000 and received a brand-new contract paying 4%. Because he paid exactly face value, his coupon rate and his return are the same number: 4%.
The secondary market
Everyone elseInvestors trading bonds that already exist. Peter isn't obliged to hold for 30 years — he can sell to another investor tomorrow.
Around $32 trillion of US government debt sits out here, changing hands continuously. The government gets none of this money; it already got its $1,000.
Peter bought a $1,000 bond with a 4% coupon, so he is owed $40 a year, frozen, for thirty years. That obligation is written into the contract and cannot change. But the price another investor will pay him for that contract moves every second of every day. Everything that follows falls out of that single asymmetry: the payment is fixed, the price is not.
So where does the 4% come from in the first place? Nobody at the Treasury picks it. It comes out of the auction — and the auction is the part almost nobody explains.
The room where your mortgage is decided
Roughly every month the Treasury announces it needs to sell, say, $25 billion of 30-year bonds. It does not announce an interest rate. It announces a quantity, and then it holds an auction and lets buyers tell it the price.
Who is actually in the room
These are the buyers everyone means when they say "the bond market":
Primary dealers
About 25 giant banks and broker-dealers. They have a standing obligation to bid at every single auction — the backstop that guarantees the government can always sell. They mostly resell within days.
Indirect bidders
Foreign central banks, sovereign wealth funds and overseas institutions, bidding through the dealers. Japan alone holds about $1.2 trillion of US government debt.
Direct bidders
Large domestic institutions bidding for themselves: pension funds, insurers, mutual funds, hedge funds.
Non-competitive
People like Peter, buying small amounts through TreasuryDirect. They don't name a price — they accept whatever the auction decides. Capped at $10m each; a rounding error.
How the bidding works
Everybody in groups 1–3 submits a sealed bid saying, in effect: "I will lend you $2 billion, but only if you pay me at least 5.10%." A bidder who demands a lower yield is offering the government a better deal.
The Treasury then does something that surprises most people the first time they see it:
-
Sort every bid from the cheapest yield upward
The most generous lenders — the ones willing to accept the least interest — go to the front of the queue.
-
Fill the queue until the $25 billion is sold
Work down the list, accepting bids, until the entire amount is placed. Everyone past that point gets nothing.
-
Charge everyone the worst accepted yield
The yield of the last bid needed to finish the sale becomes the rate paid to all winners — even the bidder who said they'd accept 4.88%. This is a single-price, or Dutch, auction, and it's the whole ball game.
Because everyone pays the worst accepted price, the government's borrowing cost is decided entirely by how far down the queue it has to go. Sell a little, and you only need the eager lenders at the front. Sell a lot, and you keep going until you reach people who will only lend at much higher rates — and then pay that rate to everybody. This is the mechanical link between how much the government borrows and what it costs. Not a theory — an arithmetic consequence of the auction rules.
View the bid book as a table
The two numbers traders stare at
When an auction finishes, the result is published within minutes and two numbers decide whether the market treats it as good news or bad.
Bid-to-cover is total bids divided by the amount offered. August's auction drew about $60 billion of bids for $25 billion of bonds — a ratio of 2.39. Above roughly 2.5 is healthy appetite; below 2.0 is a warning. It answers: how many people wanted in?
The tail is subtler and matters more. Before the auction, traders already buy and sell the not-yet-issued bond in a grey market called when-issued trading, which gives everyone a live guess at where the auction should clear. If the auction actually clears at a higher yield than that guess, the auction "tailed" — the government had to pay more than the market expected to get its money away.
August's auction tailed, and primary dealers were left holding 11.5% of the issue. Dealers absorbing an unusually large share is the tell: the real end-buyers didn't show up in the size expected, and the obligated backstop had to eat the difference.
How the auction becomes everyone's price
Here's the link that ties the two markets together, and it's the piece most explanations skip.
The moment that auction clears at 5.22%, a brand-new, government-guaranteed, 30-year bond paying 5.22% exists and is available. That is now the benchmark. Every other 30-year claim on the US government in the world — including Peter's — must instantly reprice to compete with it, because no rational buyer pays more for an inferior version of the same thing.
The causality runs in a loop, continuously:
A rising yield does not mean more people want the bond. It means the opposite. Yields rise when demand is weak — when the government has to reach further down the queue, offering more interest, to find enough lenders. Strong demand pushes yields down. Whenever you read "yields surged," read it as "buyers had to be bribed harder."
Why every holder just lost money without selling anything
Three years pass. Peter still holds his $1,000 bond paying him $40 a year. Then a Treasury auction clears at 5%, and Peter's world quietly changes.
A fresh $1,000 bond now pays $50 a year. Peter's pays $40. Suppose he needs his money early and takes his contract to the secondary market to sell it.
Why would anyone pay Peter $1,000 for $40 a year, when the Treasury will sell them $50 a year for the same $1,000?
They wouldn't. Peter has exactly one lever: cut the price. He has to discount his bond until the fixed $40 a year, measured against the smaller amount the buyer actually pays, works out to the same 5% return available everywhere else. That price is about $846. There, the buyer is indifferent, and only there.
Peter's income never changed — still $40 a year, exactly as the contract says. The price fell so that the yield could rise to meet the market. They are two ends of one seesaw. "Bond prices are falling" and "bond yields are rising" are the same event, described from opposite ends.
So when you read that yields hit a 19-year high, you are also reading — in the same breath — that everyone already holding those bonds just took a serious loss.
So does Peter actually sell?
This is the question the charts never answer, and it's the one that decides whether any of this hurts him. He has two choices, and they are genuinely different.
If he holds to maturity
He loses nothing in cashThe contract still pays $40 a year, and in year 30 the Treasury still hands back the full $1,000. The $846 was never a real loss — it was a quote. Nobody can force him to accept it.
What he does lose is invisible: he's locked into 4% while the world now pays 5%. That's roughly $10 a year of forgone income, every year, until 2056. Not a loss on a statement — a loss of the better life he could have had with that money.
If he sells today
He realises the $154He takes $846 and the loss becomes real, printed, permanent. But here's what almost nobody points out: he can immediately put that $846 into new bonds paying 5%.
$846 at 5% earns him $42 a year — more than the $40 he was getting. He has less capital and more income, and the two exactly offset. That is what "fairly priced" means.
Both paths leave Peter in the same place. The market moved him from one to the other without making him better or worse off, which is precisely why the price is $846 and not some other number. The $154 isn't money being taken from him — it's the market's measurement of how much worse a 4% promise is than a 5% one, over thirty years. The loss is real, but it was caused by rates rising, not by selling. Selling only decides when he acknowledges it.
Which makes the dangerous case obvious: it's the holder who must sell. Someone who planned to hold to maturity but needs cash early — a saver, a fund facing withdrawals, or a bank whose depositors all want their money at once — is forced to accept the $846. For them the paper loss becomes a real one at the worst possible moment. Remember that; it kills a bank in Act 5.
Why the thirty is the dangerous number, not the four
Peter lost 15% of his bond's value from a one-point move in rates. That feels too big, and it's worth understanding where it comes from, because it explains why the headline is about the 30-year specifically and not about bonds in general.
The size of the loss has almost nothing to do with the 4% coupon. It comes from the 30. Peter's problem isn't that he's earning $10 a year too little — it's that he's earning $10 a year too little thirty times over. The buyer discounts every one of those thirty shortfalls, and the price has to fall far enough to compensate for all of them at once.
Shorten the maturity and the same rate move barely registers, because there are only a handful of shortfalls left to compensate for:
View as table
A 2-year bond shrugs the move off — it loses under 2%, because it only has two years of being slightly wrong before it repays. The 30-year loses more than eight times as much.
That asymmetry is why the 30-year is the number that makes headlines. It's the most sensitive instrument in the market: the place where a small change in what the world believes about the next three decades produces the largest, most visible change in price. It is the bond market's amplifier.
This is why "bonds are the safe asset" broke. In 2022 long-dated Treasuries had their worst year in modern history — not because America came close to defaulting, but because rates rose fast and duration did the rest. Safe from default is not the same as safe from price. Every one of those bonds will still pay back Par is the face value printed on the bond — the amount the government repays at maturity, regardless of what the bond traded for in between. Peter's was $1,000.. Anyone who had to sell first did not get to find out.
Why the long end broke loose
We now have the machinery. Yields rise when the Treasury has to reach further down the bid queue. So the real question is: why is the queue getting harder to fill? Three forces, and none of them is the Fed.
Force 1 — There is simply too much of it
The US government now owes just over $40 trillion, a line it crossed this month. To keep running it has to sell more than $2 trillion of new bonds a year, on top of refinancing everything that matures.
Go back to the auction. Selling $25 billion means stopping partway down the queue. Selling $45 billion means pushing past the eager lenders into people who'll only lend at 5.36% — and then paying that to everyone. Rising supply mechanically walks the government into worse and worse bids. That's not sentiment. That's the auction rules.
Force 2 — Thirty years is a long time to trust a currency
Inflation is running at 3.4%, still meaningfully above the Fed's 2% target. If you lend at 5% for thirty years and inflation averages 4%, you earned 1% a year for three decades of risk.
Inflation is the bond investor's only true enemy. A share in a company can raise its prices; a bond's $40 coupon cannot. It is a fixed number being slowly eaten. Lenders know this, so when inflation looks sticky they refuse to bid at low yields — which, in the auction, is exactly the behaviour that pushes the clearing rate up.
Force 3 — The term premium came back
Here is the direct answer to the question from Act 0. A 30-year yield is really two things added together:
- What the market expects short-term rates to average over the next 30 years. The Fed heavily influences this part.
- The term premium — extra compensation for locking money away for three decades when the future is uncertain. The Fed does not control this part.
For fifteen years the term premium was crushed near zero: quantitative easing had the Fed buying bonds at any price, foreign central banks were hoovering up Treasuries, and the world was convinced inflation was dead forever. All three have reversed. The term premium is back, and it is doing work the Fed cannot undo by cutting.
If 5.27% is so attractive, why isn't everyone buying?
This is the obvious objection, and it's the right one. A guaranteed 5.27% from the United States government is the best risk-free deal in twenty years. If it's really that good, buyers should pile in — and buying pushes prices up, which pushes yields back down. The problem should fix itself.
It hasn't. There are five reasons, and together they're the honest answer.
Supply outruns them
Buyers are showing up — August's auction still drew $60 billion of bids. But the Treasury keeps arriving with more paper, week after week. Demand isn't absent; it's being outpaced. You can be thirsty and still drown.
Waiting is the better trade
If you think the 30-year is heading to 5.6%, buying today at 5.27% is a mistake — duration maths says you'd lose about 5% of your capital on the way. So buyers stand back. That hesitation is itself a withdrawal of demand, which pushes yields higher, which confirms the thesis.
The income can vanish
5.27% a year sounds generous until you remember a further half-point rise costs about 8% of capital. That's eighteen months of income wiped out in a few weeks. For anyone who might need to sell, the yield isn't compensation enough.
A serious new competitor
The AI buildout turned big tech into enormous borrowers. Hyperscalers and related firms issued about $225 billion of bonds in the first half of 2026 — a tenfold jump — on pace for $400 billion. Same pool of savings, and they pay more than the government.
Yields don't fall just because bonds look cheap. They fall when demand finally exceeds supply at the going price — and that requires either the government to borrow less, or buyers to become convinced that inflation is beaten and rates have peaked. Until one of those happens, "it's attractive" and "it keeps getting cheaper" can both be true at the same time. Markets can stay mispriced for exactly as long as the supply keeps coming.
Who is actually holding all of this
It's worth knowing who is on the other side, because "the bond market" sounds abstract until you see that it's pension funds, foreign governments and the Fed itself.
View as table
How it reaches your bank account
You will probably never buy a 30-year Treasury. It reaches you anyway, because it is the reference price for long-term money everywhere in the economy. Nobody lends to you for less than the government pays. The government's rate is the floor; everything else is that floor plus a markup for the risk that you, unlike the Treasury, might not pay it back.
Your mortgage is a Treasury bond wearing a hat
A 30-year mortgage is long-term lending — the same product the Treasury sells. So lenders start from the Treasury yield and add roughly 1.9 points for the risk that you default or refinance early. (They use the 10-year, not the 30-year, because most mortgages are repaid or refinanced within about a decade.)
Nothing about you changed. Your salary, your credit score, the house — all identical. The bond market repriced, and a six-figure sum moved from your pocket to your lender's over the life of the loan.
It doesn't stop at housing. Car loans, credit-card APRs, student debt, business loans and commercial real estate all reprice off the same curve, each with its own markup on top.
Savers stopped bleeding
In pure interest, the change is staggering: $100,000 earned about $70 a year in 2021 and earns roughly $4,030 now. That's an increase of about 58 times, and it's the number you'll see quoted everywhere.
But it overstates the case badly, because it ignores inflation. In 2021 you earned 0.07% while prices rose 4.7% — you were losing about 4.6% of your purchasing power every year. Today you earn 4.03% while prices rise 3.4%, so you're finally ahead, by about 0.6%. The truthful summary isn't "savers got 58× richer." It's savers stopped bleeding. After fifteen years of being quietly taxed to subsidise borrowers, cash roughly breaks even again.
That's a smaller claim than the headline number, but it's a real one, and for some people it matters enormously. A pension fund that must fund a payment in 2050 can now lock in over 5% a year, guaranteed, for doing nothing risky at all. In 2021 that was mathematically impossible — such funds were forced into stocks and property to have any hope of meeting their obligations.
The catch, and it's the same one throughout: this is good news only if you're buying bonds now. If you already owned long-dated bonds, Act 3 already happened to you.
Why a share price falls when a bond yield rises
A share is a claim on profits that arrive in the future. To decide what that's worth today, you ask: how much would I need to invest right now, risk-free, to end up with that same amount? That risk-free rate is the Treasury yield. When it rises, you need less money today to reach the same future sum — so the future profit is worth less now.
Concretely: suppose a company will earn $100 per share at some point in the future. Drag the slider to see what that $100 is worth today, at a 4% discount rate versus today's 5.27%.
Slide it to 3 years and the damage is trivial — under 4%. Slide it to 25 and it's over a quarter of the value, gone, with nothing having changed about the company at all.
That's the whole story of which stocks get hurt. A utility or a bank earning steady cash today barely flinches. A tech company priced on profits in 2035 gets hit hard, because its value is almost entirely made of distant promises — it has, in exactly the sense from Act 3, high Duration is how much a price moves when rates move. The longer you wait to be paid, the more a change in rates costs you — which is why a 30-year bond lost 15% where a 2-year lost under 2%.. A growth stock is a long-dated bond that happens to be traded on an equity exchange.
For fifteen years the argument for owning stocks was There Is No Alternative — bonds paid nothing, so where else would you go? At 5.27% risk-free for thirty years, there is now a very obvious alternative. Every pension fund and allocator on earth is rerunning that comparison, and some are choosing the bond. A fund that needs 6% a year to meet its obligations used to have to own equities to get there. Now it doesn't.
And there's a third channel, the one that killed Silicon Valley Bank. Banks park deposits in Treasuries. When yields spiked in 2022, those holdings lost value exactly as Peter's did. Normally that's a paper loss — hold to maturity and you get par. But SVB's depositors all asked for their money at once, which forced it to sell into the loss. It's the forced-seller case from Act 3, at institutional scale.
The hurdle rate went up
Companies fund expansion by borrowing, and corporate borrowing is priced off the Treasury curve plus a credit spread. When the base rate rises, every project has to clear a higher bar to be worth doing at all.
A factory expansion returning 7% was clearly worth building when money cost 4%. At 6.5% it isn't — you'd be taking real risk for half a point. It doesn't get cancelled dramatically; it just quietly never starts. Multiply that across an economy and you get fewer new plants, fewer expansions, slower hiring.
This is the intended mechanism, incidentally. It's how higher rates are supposed to cool an overheating economy — the discomfort is the policy working. What's unusual now is that it's happening through the long end, which the Fed didn't choose and can't easily reverse.
Most exposed: construction, real estate, and any business whose customers borrow to buy the product — cars, homes, capital equipment. Most insulated: businesses with no debt and customers who pay cash.
The AI buildout complicates this in an interesting way. Hyperscalers are spending around $725 billion on capital expenditure in 2026, up 77% in a year, much of it debt-funded. That's an enormous investment boom happening despite high rates, because the expected returns clear the hurdle. It props up employment in data centres, chips and power — while the borrowing that funds it competes with the Treasury and helps keep rates high for everyone else.
Interest now costs more than the military
For the first sustained stretch since the Second World War, the United States spends more servicing its debt than defending itself.
Here is what that actually means in plain terms. The government collects tax and spends it on things: roads, the military, Medicare, salaries. Interest is now one of the biggest items on that list — but unlike every other item, nobody votes on it. It's contractual. It gets paid first, automatically, before Congress debates anything else.
So a rising interest bill doesn't announce itself as a cut. It just means that out of every tax dollar collected, a bigger slice is already spoken for — sent straight to whoever holds the bonds — and a smaller slice is left for everything the government might otherwise choose to do. The alternatives are to raise taxes or to borrow the difference, and borrowing the difference means issuing more bonds, which brings us back to Force 1.
There's a distributional edge worth naming plainly. Interest payments flow to whoever owns the bonds — pension funds, wealthy households, and foreign governments, roughly 30% of it overseas. The taxes that fund those payments are collected broadly. Mechanically, rising yields are a transfer from taxpayers to bondholders.
The part that keeps professionals awake
Everything so far has been mechanical: rates go up, borrowing costs more. Here's the piece that makes bond investors uneasy, and it's a feedback loop.
Higher yields raise the interest bill. A bigger interest bill means the government spends more than it collects by an even wider margin. That gap has to be borrowed, so it issues more bonds. More bonds means the auction reaches further down the bid queue — and the clearing yield rises. Around again.
The ratchet that's already locked in
This isn't a slow-moving hypothetical, and here is the single most important chart on this page.
The government doesn't pay today's rate on all its debt. It pays a blended average of every rate it ever borrowed at — including trillions issued in 2020 and 2021 at 1% or less.
That average is currently 3.45%. New 30-year debt is being issued at 5.27%.
View as table
Even if the 30-year froze exactly where it is today, the interest bill would keep growing for years — as 1% debt rolls into 5% debt. The pain isn't a forecast. It's already contracted.
Why the obvious fixes are hard
Loops like this do end, by one of four routes. It's worth being honest about how plausible each one currently looks.
Most likely it's a muddled combination over years, with option 3 doing quiet work and option 4 held in reserve. But now you know what you're watching for, and why "they'll just cut spending" isn't an answer anyone in the market takes seriously.
Three things to watch
- Auction results. Every few weeks the Treasury sells long bonds. Watch the Total bids divided by the amount offered. Above roughly 2.5 means healthy appetite; below 2.0 is a warning. August's 30-year drew 2.39. and whether it An auction "tails" when it clears at a higher yield than pre-auction trading expected — the government had to pay more than the market thought to place its debt. A sign real buyers stayed away.. That's the market's live vote, in real money — and now you know exactly how to read it.
- Inflation prints. Monthly CPI. Sustained progress toward 2% would let the long end relax; renewed acceleration would not.
- The 10-year specifically. It's the one that prices your mortgage. Watch it more closely than the headline-grabbing 30-year.
This is an explainer about a mechanism, not advice about what to do with your money. Whether higher yields are good or bad for you depends entirely on whether you're a borrower or a lender, and on a personal situation no web page knows. What's worth taking away isn't a position — it's the ability to read the next bond-market headline and know exactly what it means for your own balance sheet.
The Fed sets the rate for borrowing overnight. An auction room sets the rate for borrowing for thirty years. Only one of those is the price of a house.