Neel Kashkari did not sound like a man describing an emergency. Speaking on Wednesday, the president of the Minneapolis Federal Reserve said what central bankers have been saying in one form or another since the spring: "Inflation is still too high." Then he added the clause that mattered more. "It's been elevated now for more than five years."

The next morning, the market supplied its own arithmetic. In the early hours of 1 October, two Chinese refiners, the state-owned PetroChina and the private Zhejiang Petrochemical, were reported to have suspended fuel exports for October, and PetroChina cancelled cargoes it had already arranged. China's commercial stocks of gasoil and diesel were sitting roughly twenty million barrels below the level Beijing requires before it lets fuel leave the country. Brent crude, which had slipped below $100 in the previous days, climbed back above it. And on the screens where government debt is traded, the yield on the ten-year United States Treasury touched 5.34 percent, its highest level since April 2002. The thirty-year bond reached about 5.67 percent, also a level last seen in 2002. In London the thirty-year gilt crossed 6 percent for the first time since 1998.

That is the story most readers met on Thursday: oil up, bonds down, a twenty-four-year record. It is true, and it is incomplete, because a yield on a screen is the price of the next dollar a government borrows. It says nothing directly about the price of the dollars it has already borrowed. For those, the United States Treasury publishes a different number every month, and on the same Thursday that number was just under 3.5 percent.

Between 3.5 and 5.34 lies most of what this piece is about. The screen shows where the price of money is going, while the 3.5 shows where the federal budget still is. The distance between them does not close on the day of a headline. It closes on a schedule, maturity by maturity, and in fiscal year 2026 alone the Treasury had to refinance about $9.7 trillion of maturing securities. Oil did not write that schedule, though at the moment it is one of the forces deciding at what price the schedule gets filled.

5.34 and 3.5: The Price of the Next Dollar and the Price of All the Others

Every large borrower carries two interest rates at once, and most public discussion only ever looks at one of them.

The first is the marginal rate: what the market demands today for a new loan of a given length. For the United States government, the ten-year Treasury yield is the best-known version of it, and it is the number that moves on screens, appears in headlines, and on Thursday reached 5.34 percent. The second is the average rate: what the borrower is actually paying, today, across everything it has outstanding. That number moves slowly, because most of the debt was issued in the past, at whatever the market charged then, and it keeps that price until it matures. The Government Accountability Office put the average interest rate on marketable Treasury securities at 3.4 percent at the end of fiscal year 2025. A Cato Institute analysis published on 1 October put it at "just under 3.5 percent." In 2014, it was 2 percent.

Of the two, only the second shows up in the federal budget. A household understands this without being told: the mortgage advertised in a bank's window this week is not the rate on the mortgage you took out in 2021. Your payment changes only when your fixed period ends and the old loan has to be replaced by a new one. Governments work the same way, at a scale that turns a slow mechanism into a large one.

Three facts set the speed. According to the GAO, about 33 percent of federal debt was scheduled to mature within twelve months as of September 2025, up from 24 percent in 2014. Treasury bills, the shortest instruments, made up about 22 percent of marketable debt in fiscal 2025, against 13 percent a decade earlier. And the weighted average maturity of the debt stood at 71 months, close to its highest level in a quarter of a century. Put together, this is a borrower that refinances a third of what it owes every year, keeps a large share of it short, and holds the rest for about six years on average. The shift toward the short end happened during the years when short money was almost free, and it was a reasonable choice then. It also means that a larger share of the stock now meets the new price every year than a decade ago.

That combination has a consequence that the headline number hides. If the marginal rate stays well above the average rate, the average does not need a crisis to rise. It only needs the calendar to keep running. Each month, some debt issued at 1 or 2 or 3 percent matures and is replaced by debt issued at whatever the market asks that week. Bills roll at short-term rates, which the Federal Reserve now sets between 3.75 and 4 percent. Notes and bonds roll at rates closer to Thursday's screens. The old price drains out of the stock, the new price drains in, and the average climbs toward the market at a pace set by the maturity schedule rather than by anyone's decision.

A government does not pay the yield it sees on Thursday. It pays, year by year, the yields it sees on the days its debt comes due.

The Congressional Budget Office built its February 2026 projections on a ten-year yield averaging 4.1 percent this year, rising to 4.3 percent in 2027 and holding near 4.3 to 4.4 percent after that. On those assumptions, net interest on the federal debt was projected at $1,039 billion in 2026, or 3.3 percent of GDP, rising to $2,144 billion, or 4.6 percent of GDP, by 2036. Debt held by the public was projected at 101 percent of GDP this year and about 120 percent a decade from now. Thursday's market sat more than a full percentage point above the yield those numbers assumed.

PetroChina, Zhejiang and the Diesel That Never Leaves Port

To see how a barrel of oil reaches the price of money, it helps to start somewhere other than the barrel.

The war that began on 28 February 2026 closed the Strait of Hormuz in stages. A June memorandum eased the blockade for only a few days, and the wider ceasefire collapsed on 8 July after attacks on commercial shipping. Since then, crude supply has partly found other routes. Saudi Arabia brought its East-West pipeline back into service and tanker departures from Yanbu on the Red Sea picked up, which is one reason Brent was able to fall back toward and briefly below $100 in late September. The crude shortage, in other words, had begun to look manageable.

The refined-product shortage had not. American diesel prices reached a record of roughly $6.50 a gallon by one count, about 70 percent above their level before the war. By the time the Federal Reserve met in September, crude had crossed $100 and diesel had pushed to new records above $6 a gallon. And on Thursday morning, China, which has spent the past two years exporting refined fuel into a tight world market, decided to keep its fuel at home. The reported reason had nothing to do with Iran. It came down to an inventory threshold: Beijing does not permit exports when domestic stocks fall below a set level, and they had fallen about twenty million barrels below it for gasoil and diesel, and about nine million barrels below it for gasoline.

That decision travelled to Brent within hours. It travelled to the bond market by a longer road, through a number that most investors check before they check the oil price: inflation expectations. Thursday's Institute for Supply Management survey showed American manufacturing still expanding, with the headline index at 54.5. Inside the same survey, the prices-paid component jumped 6.8 points in a month, to 77.9. The factories were still busy; what they were reporting was that the things they buy cost more, and were getting more expensive faster than in August.

A fuel cargo that never leaves a Chinese port does not need to reach Ohio, Manchester or Lyon to change the price of a mortgage there. It needs only to change the expected path of inflation, and with it the expected path of the central bank that answers for inflation, and with it the yield that lenders demand to hold a government's paper for ten years. There is nothing exotic about that chain; it is how an energy shock leaves the energy market.

What makes this autumn different is the condition the chain arrives in. The bond market received Thursday's oil news after eight months of war, a summer of record diesel, and a central bank that had already concluded, two weeks earlier, that the inflation it was seeing was not going away by itself.

Washington, 16 September: The First Hike Since 2023

The Federal Open Market Committee does not meet in a trading room, and its decisions do not look like much from outside: a statement released at two in the afternoon, a press conference half an hour later, a chair at a podium answering questions about single words. On 16 September the statement contained a number nobody in the room had written into a statement since July 2023. The federal funds range went up, by a quarter point, to 3.75 to 4 percent, and all twelve voting members agreed.

Kevin Warsh, who chairs the committee, gave the reason without much ornament. "The plain fact is that inflation is too high, and has been for too long. This summer's inflation readings do not tell me that underlying trends have meaningfully improved." Consumer prices had risen 3.4 percent over the previous year, and 0.4 percent in August alone, the largest monthly increase in four months. The projections released with the decision were blunter than the chair: the large majority of participants expected another increase before the end of the year.

This matters for the bond market in a specific way. A central bank that is cutting, or that is expected to cut, gives long-term lenders a reason to accept lower yields today, because they expect cheaper money tomorrow. A central bank that has just started raising takes that reason away. And Thursday's economic data took away another one. Initial jobless claims fell for a fourth straight week, to 197,000. Manufacturing kept expanding.

In an ordinary year, those would be reassuring numbers. On Thursday, they were not, because each one made it harder to believe that the economy would weaken enough, soon enough, to persuade the Fed to reverse course. A strong labour market and rising factory input prices, with oil back above $100, describe an economy in which inflation is being fed from two sides at once. The market's reply was to demand more for holding long-term government paper. That is how good economic news becomes bad financial news. Investors were not hoping for a recession; they had simply stopped expecting one to arrive in time to bring rates down.

The Fed sets today's price of short money. The ten-year yield prices ten years of guesses about what comes next, plus a premium for being wrong.

That premium, the extra return investors demand for committing money for a long time to a borrower whose future deficits are large and whose inflation outlook is uncertain, is where the oil shock and the fiscal position meet. Jefferies economist Mohit Kumar summarised the mood across markets on Thursday in one line: "Inflation, deficit and issuance concerns continue to weigh on the bond market." Each of those three words points to a different institution. Inflation belongs to the Fed. Deficits belong to Congress. Issuance belongs to the Treasury. The yield belongs to all of them, and is controlled by none.

$9.7 Trillion a Year: Where the Gap Is Closed

Here the two numbers from the opening come back, because this is where the gap between them stops being an abstraction.

In fiscal year 2026, according to the GAO, the Treasury had to refinance $9.7 trillion of maturing securities. That is the volume of old debt that met the new market this year, on top of the roughly $1.9 trillion of new borrowing the CBO projected for the deficit. Some of it was issued years ago at rates near zero. Some was issued last year at rates near today's. The average of everything outstanding, after all of that rolling, still sits just under 3.5 percent, because the debt that matures is replaced gradually and because a large share of the stock was locked in at lower prices during the years when money was close to free.

That average will not stay where it is if the market stays where it is. The CBO said as much on 24 September, a week before Thursday's records, in a set of alternative projections. In one scenario, the average interest rate on federal debt rises about five basis points a year above the baseline until it is a full percentage point higher. Under that path, the CBO projected debt held by the public reaching 222 percent of GDP in 2056, against 175 percent in its baseline, with interest costs some $1.4 trillion higher over the coming decade alone, according to the Committee for a Responsible Federal Budget's summary. A separate rule-of-thumb estimate from the Cato Institute, applying the CBO's sensitivity figures to a ten-year yield averaging 5.3 percent instead of 4.3 percent, put the cumulative additional borrowing over 2027 to 2036 at about $3.8 trillion. The two numbers differ because they model different paths. They agree on the direction, and on the fact that the path is decided mostly by how long the market stays above the assumption.

One detail in the CBO's baseline is easy to miss and more revealing than either headline figure. As the Committee for a Responsible Federal Budget reads the baseline, the average interest rate on federal debt does not rise above the current ten-year yield until 2047. This is no forecasting error but the arithmetic of a slow-moving average chasing a faster market, and it shows how much of the adjustment is still ahead rather than behind. A government whose average rate stays below the market rate for two decades is a government that has not yet paid for the rates it is now being quoted.

The loop that follows is simple enough to fit in a paragraph, and that is precisely why it is easy to overlook. Higher yields raise the cost of each refinancing. Each refinancing raises the average rate a little. A higher average rate raises net interest. Higher net interest widens the deficit. A wider deficit raises the amount of new debt that must be sold. More debt for sale is one of the reasons investors give for demanding higher yields in the first place. Nothing in that sequence requires a failed auction, a downgrade, or a panic. It requires time, a maturity schedule, and a market that stays above the assumption for long enough.

I do not know whether 5.34 percent was a peak or a waypoint, and nothing in this argument depends on knowing. The ten-year yield fell back to about 5.24 percent by Thursday afternoon. It could be at 4.9 by Christmas or 5.6. What does not depend on the next week's trading is the structure underneath: a stock of debt priced at roughly 3.5 percent, a market pricing new debt above it across most maturities, and a calendar that will put some $10 trillion of the old stock in front of the new price every year. The daily number decides how fast the gap closes. The calendar decides that it does.

Every refinancing also has a second side, and it belongs on the same map. The higher price the Treasury pays is income for whoever buys the new paper: money market funds, banks, insurers, pension funds, foreign central banks, and households holding Treasury bills directly or through their savings. A saver rolling a bill at around 4 percent, or a pension fund buying a thirty-year bond near 5.67 percent, is on the receiving end of the same gap that the budget is on the paying end of. That is part of why the arrangement meets so little resistance. The interest that costs the taxpayer pays the saver, and many households are both, which spreads the cost thinly enough that it is rarely felt as one.

The headline rate is news. The average rate is the budget. The distance between them is a schedule that nobody votes on.

1,179 Basis Points: The Weakest Borrowers Meet the Price First

If the federal government is the slowest borrower to feel a change in the price of money, the fastest are the companies that can least afford it, and they have already started to.

On 30 September, the extra yield investors demand to hold American corporate bonds rated CCC, the lowest rung of the junk market, stood at 1,179 basis points over Treasuries, according to the ICE BofA index published by the Federal Reserve Bank of St. Louis. A month earlier, on 1 September, it had been 1,049. That is a widening of 130 basis points in thirty days, and the highest reading in the three years of the series the St. Louis Fed makes public. Bloomberg, using its own index, reported the same day that CCC debt had crossed the 1,000-basis-point line that the market conventionally treats as distressed.

What happened one rung at a time up the ladder is as informative as what happened at the bottom. Over the same month, the spread on single-B bonds widened by about 41 basis points, to 316. The spread on BB bonds, the top of the junk market, widened by about 42, to 194. And the spread on BBB bonds, the lowest tier of investment grade, where much of corporate America borrows, moved by 4 basis points, from 99 to 103. The weakest borrowers' spreads widened by more than thirty times as many basis points as the investment-grade middle. Torsten Slok, chief economist at Apollo, noted that borrowing in the CCC market now costs roughly double the broader high-yield average. Barclays analysts warned that many private credit funds have become heavily concentrated in a few sectors, software above all, where business models were built on the assumption of cheap capital and are now also exposed to disruption from artificial intelligence.

This is the pattern by which expensive money usually travels: not evenly, and not from the top. The strongest borrowers barely notice at first. The middle of the market is repriced in an orderly way. The bottom, the companies whose balance sheets only work if they can roll their debt at something close to the old price, meet the new price first, because their lenders have the least reason to be patient. None of it is sudden: it is the slow compression of refinancing economics, and that is a fair description of the whole mechanism this piece is tracing. The federal government sits at the top of the same ladder whose bottom rung is now distressed. It refinances at the lowest price in the world, and it will keep doing so. But it is subject to the same arithmetic as the CCC issuer: a stock of old debt at an old price, a schedule of maturities, and a market that has moved. The difference is speed and scale, not kind.

Nearly $500 Billion of AI Debt in One Year

The finding of this section is simple, and it is worth stating before the numbers. The Treasury and the builders of AI infrastructure are asking the same market for very large amounts of long-dated money in the same years, and the AI buildout is also one of the things keeping the economy strong enough that the Fed sees little room to ease.

The scale is not in doubt. Goldman Sachs Research reported in August that nearly $500 billion of AI-related debt had been issued in 2026 by that point, and that the largest technology companies alone had sold about $200 billion of corporate bonds, close to double their total for 2025. The money goes into data centres, chips, networking and power supply, long-lived capital financed over many years. Most of it is borrowed by companies with enormous cash flows, which will not be the first casualties of higher yields. But the market has begun to tell them apart. Oracle, which has borrowed heavily for cloud and AI capacity, saw the cost of insuring its debt over five years rise by about 70 basis points, to roughly 215, the largest increase among the large technology names in the period covered by those reports.

The link to Thursday's yields runs in two directions. The buildout supports construction, electricity demand, equipment orders and jobs, and that strength is part of why the Fed stopped expecting to cut. Every borrowed dollar of it also competes for the pool of long-term savings the Treasury must tap to refinance its $9.7 trillion. My reading, and it is an interpretation rather than a measured fact, is that the boom presses on long-term yields from both ends. Nobody has isolated how much of 5.34 percent belongs to that pressure, and I would distrust anyone who claimed a precise figure.

A shortage of oil raises the expected cost of goods, and a surplus of demand for capital raises the cost of money; this autumn both are arriving at the same desk.

7.03 Percent: The Rate on the Kitchen Table

For most households, none of this appears as a Treasury yield. It appears as a quote.

On 24 September, Freddie Mac's weekly survey put the average thirty-year fixed mortgage rate in the United States at 7.03 percent, up from 6.95 percent a week earlier. Mortgage rates track the ten-year Treasury loosely rather than precisely, because lenders add a spread for the risk that homeowners will refinance or default, and that spread changes over time. But they do not exist independently of the yield beneath them. When the ten-year rises by more than a full point above what forecasters assumed for the year, the quote that a couple reads off a lender's screen rises with it.

The household version of the refinancing calendar is visible in every street. Owners who fixed their mortgage when rates were low hold a loan whose price was set years ago; they are, for now, on the old side of the gap. A buyer this autumn is on the new side. A homeowner whose fixed period ends next year will cross from one to the other on a known date, regardless of what happens to oil or to the Fed in between. Nobody receives a letter announcing that the price of money has changed; the only notice is a renewal offer.

Property owners face the same calendar with less protection. A commercial building has a rent, a valuation and, usually, a loan. When long-term yields rise, buyers demand a higher return on the building, which tends to lower its value; when the loan matures, it must be refinanced at the new rate. The owner can be squeezed from both sides at once: an asset worth less, and debt that costs more. None of this has to break on any given day. A loan matures, a lease expires, an appraisal is redone, and a project that worked at one price of money is tested against another.

That is why financial stress so often looks sudden. The repricing has usually happened already, in the market, months earlier. What arrives suddenly is the date on which a particular borrower has to accept it.

London at 6 Percent, Tokyo Near a Thirty-Year High

The same calendar runs in every country that borrowed heavily while money was cheap, and on Thursday it was visible in several of them at once.

Britain's thirty-year gilt yield reached an intraday high of about 6.03 percent, the first time it had crossed 6 percent since 1998. The Bank of England had held its rate at 3.75 percent, but three of the nine members of its Monetary Policy Committee had voted for an immediate increase to 4 percent. The government plans to sell roughly £250 billion of gilts in the 2026-27 financial year, ahead of a budget due on 28 October. And one of the traditional buyers of long-dated British debt, the defined-benefit pension scheme, is maturing and shifting away from long gilts, which leaves more of that supply to be absorbed by investors who are free to demand a higher price. "There is carnage in the bond market which is hitting stocks hard," said Neil Wilson of Saxo UK. In early trading London's main share index fell 1.7 percent, and Germany's DAX and France's CAC 40 each lost more than 1 percent.

In Japan, the ten-year government bond yield moved back toward the thirty-year high it had set the month before. In France, the government was preparing a draft budget against a backdrop of record bond sales and a rising debt ratio, while inflation in France, Germany, Italy and Spain had come in above what analysts expected. The specific pressures differ from country to country: a pension structure in London, a central bank exiting decades of suppression in Tokyo, a budget fight in Paris. But the shape is the same everywhere. Large stocks of debt issued at old prices, a market quoting new prices well above them, and schedules that bring the old debt to the new price one auction at a time.

It would be tempting to read Thursday as a single global event with a single cause, and to name oil as that cause. It would also be wrong. Oil is the most visible of the forces pushing on long-term yields this autumn, and it is the one that moved on the day. Underneath it sit deficits, issuance, wage growth, central banks that have resumed raising, and in Japan's case a structural shift that began before the war. What oil did on Thursday was to arrive at a moment when all of those were already pressing in the same direction, and add to them.

The Strongest Objection

The strongest counterargument to this reading does not dispute any of the numbers in it. It accepts that the average rate on federal debt is around 3.5 percent, that the market is well above it, and that roughly a third of the debt matures each year. It then points out, correctly, that this has happened before and that the United States came through it. In the 1980s and 1990s, Treasury yields sat far above 5.34 percent for years. The average interest rate on the debt was higher than it is now. The dollar remained the world's reserve currency, and demand for Treasuries never failed. A weighted average maturity of about six years means the stock adjusts gradually, which gives policymakers time. And a single day's intraday high is a poor guide to anything: the ten-year fell back to about 5.24 percent within hours, and a few months of easier inflation could take it much lower.

This objection is structurally serious, and it explains a great deal. It explains why Thursday did not look like a crisis, why auctions are still clearing, and why the most likely path from here is gradual rather than dramatic. The reading offered here does not claim that 5.34 percent is a breaking point, that the United States faces a funding crisis, or that oil is the main driver of long-term yields. It claims something narrower: that the damage from a higher price of money is delivered through a refinancing schedule rather than through the headline, that this schedule now meets a debt stock about one hundred percent of GDP rather than the far smaller ratios of the 1980s, and that the gap between the market and the average is large enough that it will keep closing for years even if yields stop rising today.

Both things can be true at once: the system is not breaking, and the bill is still arriving. What separates this decade from the 1990s is less the level of rates than the size of the stock they are applied to, roughly twice as large relative to the economy as in the mid-1990s, so that each percentage point the gap closes costs more than it did then.

What the Calendar Already Knows

Go back to the screen on Thursday morning, the one that showed 5.34 percent. It is the most-watched number in finance, and it measures one thing well: the price at which the market will lend to the United States government today for ten years. What it does not show is how much of the government's existing debt has already met that price, how much is still waiting, and on which dates. That information exists. The Treasury publishes it in the Monthly Statement of the Public Debt. The GAO summarises it. The CBO models it. It does not appear on the screen, and it does not move in a way that makes news.

The arrangement has an odd property. The Federal Reserve sets the short-term rate and answers for inflation. Congress sets the deficit. The Treasury decides what to issue and when. The CBO projects the result. Each institution is responsible for its own part of the loop, and each is behaving rationally within its own mandate. No institution owns the gap itself, the distance between what the government pays on average and what the market now asks. Its speed is steered in part by a decision that is rarely discussed as one: at each Quarterly Refunding, the Treasury chooses how much of its new borrowing to sell as short bills and how much as long notes and bonds, and with that mix it can accelerate or slow the rate at which the old price drains out of the stock. The size of the gap has no owner, and its speed has an owner who is seldom asked about it.

The oil shock is the part of the story people can see, a price at a pump, a Chinese export ban, a tanker turned around, while the gap stays out of sight and will still be closing long after Brent has found another level and the war has found another phase. A fuel cargo that stayed in a Chinese port on Thursday will be forgotten within weeks. The Treasury note that matures in March 2027 has known its date for years. The market can change the price of the date. It cannot change the date of the price.

In the accounts that describe this system, every institution has a line. The households that renew a mortgage, the companies that roll a loan, and the taxpayers whose revenue increasingly goes to interest rather than to anything else appear only once, at the end, on the day the calendar reaches them.

Evidence Map

Facts, interpretations, forecasts, and disconfirming signals.

Core claim. A higher price of money reaches a heavily indebted government through its refinancing calendar, not through the headline yield. With the ten-year at 5.34 percent and the average rate on marketable US debt just under 3.5 percent, the gap will keep closing for years even if yields stop rising; oil is one of several forces setting the speed.

Evidence level. Facts, high: market levels on 1 October, the September rate decision, and the debt figures from the CBO baseline, the CBO alternative scenarios, the GAO debt-management report, Treasury's debt statement and Freddie Mac. Credit spreads from the ICE BofA series on FRED. Interpretation, medium and marked: that AI borrowing presses on long yields from both the growth and the capital side. Forecast, speculative: the average rate keeps rising.

What would confirm this. The average interest rate on marketable Treasury debt reaches at least 3.6 percent by the end of March 2027; CCC spreads stay above 1,050 basis points while BBB spreads remain near 100.

What would disprove this. The ten-year falls and stays below 4.5 percent through the first quarter of 2027 and the average rate stops rising; or CCC spreads fall back below 1,000 basis points while the ten-year holds above 5.2 percent.

Watchlist. Review on 31 March 2027: Treasury's monthly average rate, net interest in the first-half fiscal 2027 data, the maturity mix announced at Treasury's Quarterly Refunding in early November, the Fed's December and January decisions, CCC versus BBB spreads, and the UK budget of 28 October.

Sources: Congressional Budget Office, The Budget and Economic Outlook: 2026 to 2036 (February 2026) and Projections of Deficits and Debt Under Alternative Scenarios (24 September 2026); Committee for a Responsible Federal Budget (24 September 2026); GAO-26-107529; Cato Institute (1 October 2026); Federal Reserve FOMC decision of 17 September 2026; Institute for Supply Management and Department of Labor data of 1 October 2026; Freddie Mac Primary Mortgage Market Survey (24 September 2026); market reporting by CNBC, CNN, TheStreet, Yahoo Finance, AP, The Guardian, Semafor and Quartz (1 October 2026); ICE BofA option-adjusted spreads via FRED (1 to 30 September 2026); Bloomberg (30 September 2026); Goldman Sachs Research via Business Today (August 2026).

Jerry van der Laan writes The Manifest Archive, Forensic Narrative Intelligence Writing on the systems beneath power, money, and history. He traces the structures beneath them.