On the afternoon of August 19, the US Treasury changed a number on its official website.
The single-time limit for long-term Treasury bond repurchase was raised from $20 billion to $40 billion.
No new repurchase transactions took place that day. The Treasury did not spend an extra dollar. Yet, the 30-year Treasury bond yield quickly dropped by around 9 basis points.
Here, a basis point is the second decimal place in an interest rate. 9 basis points equal 0.09 percentage points. The bond yield is the interest rate the US offers to borrow money, so an increasing yield indicates a higher cost of borrowing.
This price tag was already stretched tight. On August 18, the 30-year Treasury bond yield reached a 19-year high. The US government is the largest borrower in this market, holding about $30 trillion in tradable Treasury bonds. If all these bonds were refinanced simultaneously at a rate just 1 basis point higher, the annual interest would increase by about $30 billion. While this won't happen all at once, as old bonds mature, the costs will gradually come in. But the direction is certain.
It's not just quoting for the US Treasury. Interest rates on American mortgages, loans for businesses, and the compensation investors demand when other countries issue bonds often take cues from this curve. The long end refers to the segment of borrowing that lasts longer, including the 10-year and 30-year periods. If the long end suddenly becomes more expensive, the market won't treat it as just the headache of bond traders.
This time, the Treasury adjusted the amount for repurchasing long-term Treasury bonds.
Repurchase may sound like reclaiming IOUs and reducing debt. It's not that clean. The Treasury usually uses cash raised from new bond issuances to buy back older, less liquid securities on the market, then continues to issue new Treasury bonds. The total debt does not disappear into thin air. What changes is which batch of bonds the market holds more of and which batch is harder to sell. For traders holding a certain segment of long bonds, this difference is already significant.
As soon as the announcement was made, long-term yields fell. The money hadn't gone out yet, but the market had already priced in the money that might go out in the future.
Gold rose by about 4% on the same day, the US dollar index fell to its lowest level since mid-May, and the S&P 500 closed up by 0.34%. Superficially, this is a familiar trade. Yields go down, and risk assets breathe a sigh of relief.
But if you zoom in, things are not so neat.
The Treasury Really Bought $20 Billion the Day Before
In the regular repurchase agreement on August 18, dealers wanted to sell about $200 billion in Treasury bonds to the Treasury, but the Treasury only bought $20 billion.
This money was not wasted. It provided a buyer for some old securities and also signaled to the market which securities the Treasury was willing to touch. However, after the operation, long-term yields still rose.
The next day, the Treasury did not buy any bonds but only increased the limit on the schedule, yet prices moved.
This is also the real point of interest in this matter. The market was trading not only today's $20 billion, but also a schedule for the coming weeks. The schedule informed everyone of the maximum amount the Treasury was willing to take if long-dated bonds continued to be difficult to sell. No one needed to wait for it to act; positions could be adjusted in that direction in advance.
Yellen has always known that the role he plays here is not quite like the traditional Treasury Secretary. He has referred to himself multiple times as America's "Chief Bond Salesman" and has openly stated that pushing the 10-year Treasury yield below 4% is a goal.
The most uncomfortable moment for a salesman is not when you don't have any inventory but when everyone is asking the same question: How much of a discount should be applied to your batch of goods?
On August 18, the Treasury did not change the answer with $20 billion. On August 19, it doubled the potential buying amount, and the market began to reassess this discount.
Excess $14 Billion from Seven Operations Goes into New Bonds
According to the Treasury's preliminary schedule, from September 9 to November 4, a total of seven repurchase agreements were arranged for the long end. The cap for each operation was raised from $20 billion to $40 billion, increasing the total amount bought back in the seven operations from $140 billion to $280 billion.
What was truly in excess was the middle $14 billion.
In the same quarter, the Treasury plans to issue over $230 billion in new bonds from the 10-year to the 30-year range. Comparing the $14 billion, it only corresponds to about 5.9%. And this is even a favorable comparison for the repurchase. The repurchase targets existing securities, while the new bonds are a separate set of securities. The market is truly facing the entire yield curve and the continuously rolling stock.
From an interest rate risk perspective, this amount does indeed seem larger than face value. The longer the borrowing, the more sensitive bond prices are to interest rates. The bond market calls this level of sensitivity duration. If the Treasury bought securities near the 30-year maturity, the $14 billion would take away a duration equivalent to about 30% of a 30-year auction.
But a 30% auction is not a 30% market.
Using the duration estimate commonly seen in quantitative easing research to mechanically extrapolate, the direct impact of such an increment on the yield is less than 1 basis point. This algorithm cannot serve as a verdict. Quantitative easing is the central bank's continuous purchase, while the Treasury repurchase is a limited number of security management; the buyers, expectations, and funding sources are different. It at least indicates one thing: it is difficult to explain the fluctuation of 9 to 11 basis points on the long end on that day solely based on a $14 billion spot flow.
The remaining part, no one can precisely break it down. The market will not specify at every basis point which point belongs to actual supply, which belongs to traders closing out positions early, and which comes from speculation on the next step of the Treasury.
Some people are therefore reminded of the 1961 "Operation Twist." Back then, the Federal Reserve and the Treasury tried to independently lower long-term interest rates by selling short-term debt and buying long-term debt. The two actions are not the same. The actions back then involved a genuine maturity swap. This time it's about the upper limit. However, they both encountered the same problem: if the government does not want the price of long-term borrowing to continue to rise, how much debt can they take out of the market.
On August 19, the market answered a part of that for the Treasury.
The Bond Market Only Presses the Long End, Inflation Remains Unchanged
Looking at August 18 and 19 together, the 10-year nominal yield dropped by 6 basis points. The nominal yield is the market return corresponding to the bond's face value.
At the same time, the real yield on 10-year Treasury Inflation-Protected Securities (TIPS) also dropped by 6 basis points. It subtracts the market's inflation compensation, getting closer to what investors can actually obtain in purchasing power.
Both lines moving down, and the 10-year breakeven inflation rate stuck in the middle did not change. The same goes for the 30-year.
The breakeven inflation rate is not a crystal ball; it also includes liquidity and risk premium. However, with both sides moving down and the difference unchanged, it at least shows that the bond market did not take this news as an opportunity to raise bets on long-term inflation. The 2-year and 3-month terms hardly reacted, indicating that traders did not interpret it as the Fed changing its rate hike or cut path because of this.
The minutes of the July Federal Reserve meeting released that evening had a somewhat hawkish tone. Three members at the meeting advocated for an immediate 25-basis-point rate hike. Normally, such documents would nudge the rates in the other direction. Having read it, the short end did not react. After reading the Treasury’s schedule, the long end made the first move.
Two documents on the same day. One discussing how expensive money should be borrowed, and the other discussing how many IOUs promising long-term payment from the market will be in excess.
A bond trader temporarily treats it as the latter.
Gold Doesn't Believe This Is Just a Tweaking of the Bond Variety
Gold and the dollar read something more.
Gold rose by about 4% that day, with silver seeing an even larger increase. The US Dollar Index fell by 0.86%, dropping below the 200-day moving average to its lowest level since mid-May. The 200-day moving average is simply the average price over the past 200 trading days, but many algorithms and funds treat it as a significant line. When the price crosses this line, some positions will not question the reason anymore and will start selling.
During the Asian and European sessions on August 20th, the gold price dropped below $4,500, with most of the previous day's gains still intact.
Deutsche Bank's Head of Foreign Exchange Research George Saravelos referred to this interpretation as "soft financial repression." This term is not mysterious. The government is unwilling to let interest rates rise to the level that the market is willing to accept, so it pushes down the price of borrowing long-term through debt issuance arrangements, bond buybacks, and policy signals. No one is required to lend at a low interest rate, so it is considered soft. However, the losing end is still clear, as those receiving interest will receive slightly less, and borrowers will pay slightly less.
Saravelos places this within the context of "distortionary policies." Gennadiy Goldberg of TD Securities used a shorter phrase, saying that the Treasury Department is engaging in "verbal intervention." Wil Stith of Wilmington Trust sees another layer, with the Federal Reserve and the Treasury Department pushing in not entirely the same direction.
The bond market can only see a change in the supply of long-term bonds. Gold cannot forget that those issuing the debt are also the rule-makers.
If long-term interest rates are being held down by a schedule, and the minutes of the Federal Reserve meetings still indicate a tightening bias, adjustments must be found elsewhere. The forex market is most likely to pick up on this unease first.
S&P Rose Only 0.34%, But Changes are Afoot Below the Index
At the close of the US stock market, the easiest sentence to write was that falling yields boosted risk sentiment, and the S&P 500 rose by 0.34%.
However, on that day, there were far more rising stocks than falling ones. The New York Stock Exchange's advance-decline ratio was about 1.56 to 1, while Nasdaq's was about 2.16 to 1. For every falling stock, there were more than one rising stock standing beside it.
Despite this, the Nasdaq 100 closed lower, and the Philadelphia Semiconductor Index fell by nearly 2%.
Driving the index down were the tech stocks with the highest weighting. Supporting the index was the healthcare sector, which rose over 3% that day, hitting a record high. Moderna surged about 177% in a single day as its personalized mRNA cancer vaccine received positive Phase 3 data. Phase 3 clinical trials are one of the largest hurdles before a drug is approved. When a company succeeds at this stage, its stock price can react as if it suddenly adopts a new valuation system.
So, it wasn't a day of "all stocks rising." Funds were simultaneously pouring into healthcare and small-cap stocks while exiting the most expensive and future cash flow-dependent tech stocks.
This kind of divergence has been happening for a while this year. In market-cap-weighted indices, the big companies carry the most weight. Equal-weight indices give each company an equal say. The equal-weighted index of the seven tech giants only saw a single-digit increase year-to-date. Excluding these seven, the performance of the remaining 493 companies in the S&P 500 was actually better. The Russell 2000 also had its best year in 23 years.
Those investing in indices see a rising line. Those invested in the most famous stocks, however, receive a different report card.
Three Trillion Dollars Yet to Be Accounted for, But Interest Rates Are Already Knocking
The tech selloff does not mean a sudden disappearance in AI demand.
What the market is more concerned about is the money these companies have promised to pay in the future.
Long-term contracts for renting data centers. Commitments to buy computing power. Power, server, and supply agreements. Some of these items may be listed as liabilities, while others are only disclosed in financial report footnotes and cannot all be labeled as debt. The common thread is that even though the payment due date has not arrived, the cash flow has already been allocated for the next few years.
According to a recent analysis by The Wall Street Journal, nine large tech companies have approximately $3 trillion in such future commitments. Their disclosed annual capital expenditures total around $600 billion.
Only when the numbers are put together does the outline of the problem become clear. The money companies are spending each year is one layer, while the money they have promised but not yet paid is a much larger layer.
As long-term interest rates rise, this larger layer becomes less appealing. When refinancing is needed, money becomes more expensive. When explaining today's investments with profits far into the future, the discount rate is higher. The reported revenue growth of Anthropic and OpenAI was recently revealed to be lower than expected, prompting the market to naturally question whether the same batch of future cash flows is sufficient to cover those commitments that have already been made.
On August 18, investors were not selling the idea of AI. What they were reevaluating was a lengthy payment schedule.
The Korean market took a more direct approach to the matter. On August 18, the Korea Composite Stock Price Index fell by 5.8%, with SK Hynix dropping nearly 10%. SK Hynix is a major player in the memory chip industry and a supplier to NVIDIA. At the opening on August 20, SK Hynix announced a $28.6 billion buyback plan, leading to a 6% rebound in the index.
The buyback here is not the same as that of the U.S. Treasury. When a company buys back its own stock, it returns money to its shareholders, signaling to the market that the company is willing to support its stock price. The Treasury, on the other hand, buys back old bonds, managing how the IOUs stack up between maturity dates and bond types.
Same word, different books.
On August 19, the U.S. Treasury's move was merely adjusting the upper limit on the buyback schedule.
The bond market interpreted this as a potential reduction in the long bonds held by the market in the coming weeks. Gold interpreted it as borrowers attempting to restrain the price of borrowing. The stock market initially breathed a sigh of relief for lower yields, but then turned around to calculate how much those distant payment promises were actually worth.
On that day, the Treasury did not spend a dollar on this adjustment.
Yet in a market where a loan lasts 30 years, an unrealized promise already has a price.
Original Article Link