Coincamps

Previously, we spent a lot of time discussing the logic behind the rise in long-term bond yields in Europe and the United States, as well as the various methods central banks and ministries of finance have come up with to address this. Finally, last night, we saw the U.S. Department of the Treasury directly using repurchase agreements to suppress long-term government bond yields. The Treasury Secretary stated, "The current maximum size of $2 billion per operation will be at least $4 billion per operation." I think there are a few points I would like to share: 1. Although this is not in the traditional sense what is known as Yield Curve Control (YCC), in terms of definition and the operational entity, you can find many differences, but such nitpicking discussions are meaningless; this is the government directly intervening in its own funding costs. And I also think there is no need to debate whether this practice will have a long-term effect; the core issue is how far the government is willing to go and what price it is willing to pay. 2. A few days later, remarks from the Fed or Warsh at Jackson Hole became very subtle. It is known that he often communicates with Benson, and the market has always been saying that the Fed needs to raise rates and increase communication to reduce policy uncertainty. However, you can see that Warsh is not as hawkish; in the latest minutes, he tries to continue to reduce communication frequency. This difference is very clear. I think ultimately this is a short-term and long-term issue.
The long-term concern about U.S. government bonds is the deficit rate, which has two parts: economic growth rate and expenditure level. The expenditure level is unlikely to decrease, and may even continue to rise significantly in the autumn and January of next year. So, to solve the deficit rate, it can only rely on the economy. The economy currently consists of two parts: traditional industries and technology. Currently, it seems like a pipe dream for the U.S. to reduce the deficit rate through traditional industries. Everyone understands that if the U.S. wants to achieve a low deficit rate again, it needs higher economic growth, and everyone's hope is on technology. Of course, in the past six months, technology has been in the GAI, and GAI has shown some decline. I believe there will definitely be a new narrative coming up soon because, other than that, I really don't believe that the U.S.'s traditional industries can bring the U.S. back to its heyday.
This is the narrative that the White House, Benson, and even Warsh believe in more. For the reshaping of the industry chain, for the development of technology, more financing, credit, and economic growth are needed, not to mention populist issues. At this time, suppressing demand may be short-term monetary discipline, but long-term it may be a policy error. Powell has always said, "Without price stability, we cannot achieve anything," but in reality, the U.S. needs some economic highlights to reduce the deficit rate. "Without price stability, we cannot achieve anything" is true, but stability in price alone is useless. So, I very much believe what the market says: raising rates can curb long-term rates, and increasing communication can reduce term spreads. I have no doubt that these measures can reduce short-term risks, but they are actually of no help in solving the real problem.
A few days later, we will be able to see the Fed's attitude. If Warsh is willing to share, his choice may be a watershed moment, impacting how the Fed views the issue in the short term. However, I believe that in the medium to long term, they are both facing a strategic issue. How to make a massive investment in the United States while the deficit is high, attempting to develop technology, reshape the supply chain, and at the same time minimize the pain felt by the American people. This is a very difficult issue. Money has three prices: interest rates, exchange rates, and inflation. You can see that over time, when the inflation issue is hard to solve and the market does not buy into the interest rate issue, Bernett, who has always supported a strong dollar, is now starting to ignore the dollar's exchange rate and trying to buy some time. So, the core issue in the short term is to see if the Fed will cooperate with the Treasury Department's actions.
This kind of intervention in long-term interest rates has had several comparable times in U.S. history. In terms of magnitude, they have successively increased. Firstly, the Treasury Department bought back government bonds from 2000 to 2002. This behavior is most similar, but the underlying logic is completely different. At that time, the U.S. deficit rate was very low, and the Treasury Department's issuance of bonds was not to lower the financing rate. According to their own words, it was more about liquidity management rather than rate management. Secondly, twist operations in 2011-2012 and the 1960s. These operations were carried out by the Fed, not the Treasury Department. The scale was larger, but the effect depends on whether the Treasury Department cooperates. Thirdly, the real wartime Yield Curve Control (YCC) during World War II directly pegged the interest rates of short-term and long-term debt. Currently, we are at most at the 1% or 1.5% level. If the Fed joins, it could be 2%. In the short term, I think the Fed may not need to participate in this process. I speculate that Bernett's idea is still to achieve big results with a small budget. During the most tense months in the long-term bond market liquidity, not to cause too much disturbance, to prevent the market from constantly creating a steep yield curve. However, we do not know what Warsh really wants to do and how much influence Trump has on him. This is a practice that fits his personality very well and is consistent with his experience, but the final effect will depend on subsequent fiscal and economic policies.
The more long-term issue is the U.S. economy itself. These liquidity operations or the actions of technical officials are essentially treating the symptoms rather than the root cause. The two ends of the K-shaped U.S. economy are still mired in the mire.
The absolute data of the real estate market looks good simply because prices have risen. Recently, the narrative of the upward part of the K-shaped recovery has loosened, revealing a bleak overall economic outlook. I understand Bessant's perspective very well. Previously, when everyone saw weakening economic data, they thought about lower interest rates. Now, with weakening economic data, the focus might be on increasing the deficit. In addition, the Fed's communication has caused the term spread to widen rapidly in the past month, prompting some administrative intervention. It makes sense to eliminate some of the market's bets.
As for gold, in the short term, it may need to see the Fed's stance. If the Fed's approach is that raising rates is the only way to curb long-term rates, which is currently the view of many market participants, then gold may face fluctuations but not a major issue. If there is no actual improvement in productivity and rates are raised for various reasons other than that, they will eventually come back down. If the Fed's approach is that only loose policy can stimulate more supply to enhance U.S. competitiveness and reduce inflation, then gold may have already broken through.
Lastly, I would like to briefly discuss some abstract topics. Many of America's current issues, whether getting stuck in the Middle East or making huge investments in AI without seeing a reasonable short-term return, cannot be solved by fiscal or monetary means. Technical bureaucratic maneuvers do not address the core issues. (Moreover, I don't think Bessant and Wash are outstanding technical bureaucrats given the current situation. I fail to understand why many people praise Bessant. "China is a crumbling house that will collapse with a kick, the dollar must remain strong for the U.S. economy, tariffs can generate sufficient revenue, 3% economic growth, 3% deficit, 3 million barrels of oil." From 2024 until today, you can do the opposite of what he said to make money.) As a country reaches the mid or late stage, the necessity and demand for reform increase day by day, while the feasibility and drive for reform decrease day by day. Successful reforms exist, but there are more failures. The nature of this matter itself is extremely challenging. Successful reforms are often subtle and gentle, while grand reforms driven by ideology often face too much opposition and therefore fail.
For example, in Chinese history, many have heard of Zhang Juzheng and Wang Anshi, but few have sought to understand the Two-Tax System reform, while the reform of the Two-Tax System during the Tang Dynasty was successful and far-reaching. In a way, I think the views of many old-money financiers on Wall Street are more sophisticated than Bessant's. For example, Jamie Dimon and Ray Dalio both expressed their views on the Iran issue, stating that if war is declared, it should be fought to win, not half-heartedly. They also believe that raising interest rates may alleviate long-term debt pressure. These are all correct but challenging actions. Sometimes people have to do the right thing, even if it's difficult. A skilled politician is one who can persuade society to do the right and difficult things.
If you have always wanted to take a shortcut to find the so-called optimal solution. It feels a lot like many Indians I have encountered. They always feel like they can see things others don't know and easily come out on top. The biggest lesson I have learned from history is that many things cannot be easily manipulated. If you haven't gone through life and death experiences with your brothers, if there is no one to save you when you are in a life and death situation, if you don't have decades of accumulation and trust, then you cannot have a core team to accompany you in doing things that take decades to bear fruit. Our era is full of fast narratives, technology has diminished the importance of humans, but I believe the underlying logic has not changed. Many things, whether done with robots or with humans, require time.
Today, the operation of the U.S. Treasury Department gives me the feeling of encountering Indians at work, who always think they can achieve great results with little money and can see things others can't. I have never believed in this. I think that anyone in front of you is not a fool, assuming you are not a fool yourself. So if this is just the Treasury Department's short-term maneuver to deal with seasonal and geopolitical disturbances, I think it's fine. If the Federal Reserve also joins, then I think the underlying logic will change dramatically.
I rarely discuss the decline of the U.S. dollar, not because I think the U.S. has no problems, but because I always feel that these grand narratives take a lot of time and must witness enough catalysts to have a valuable discussion. I think if the Fed also joins this control of long-term yields and if the situation in the Hormuz Strait ends in a fiasco, that is a sufficient catalyst. We can discuss this further tomorrow.
Original Article Link

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

Broadcom's proposed massive debt financing brings AI infrastructure's cash flow constraint to the forefront. Local approvals and enterprise deployments simultaneously prove that, beyond funding, power, licensing, and data governance remain the boundaries of expansion.
1｜Broadcom Plans Over $600 Billion in Debt Financing, Extending Credit Chain for AI Infrastructure
Reuters reported on August 21 that Broadcom is in talks with a group of lenders to raise over $600 billion in debt financing for companies like Anthropic's AI chip. Bloomberg stated that the plan may include around $300 billion in subordinated debt and a $600 billion to $700 billion senior secured portion, with a total size of up to $1 trillion. The deal is not yet finalized, with Broadcom and Apollo not responding to requests for comment, and Blackstone declining to comment.
Reuters said this debt may be issued by a special purpose entity and extend beyond the $35 billion mining expansion arrangement Broadcom, Apollo, and Blackstone reached in June. Infrastructure investments primarily in procurement, leasing, and equity are now layered with longer-term debt and collateral.
This does not signal an industry-wide shift in financing models but will draw more attention to whether leases, chip supply, and customer payments can support debt repayment. The next test of mining expansion is shifting from accessing capital to establishing contractually payable cash flows.
(Source: Reuters / Bloomberg / Blackstone)
2｜Data Center Expansion Meets Power and Approval First
According to NPR, Reuters, and The Verge, a town in Michigan passed a law prohibiting the construction of power facilities for data centers. Loudoun County, Virginia, has also begun tightening approvals for new projects. Related statistics indicate that at least 75 major data center projects in the U.S. have been delayed or canceled, involving funds exceeding $130 billion. The scope of the statistics and project status still need to be updated as local permits progress.
These resistances involve grid access, water usage, noise, and land utilization. For residents, when the tax revenue and construction jobs brought by data centers are insufficient to offset infrastructure costs, projects shift from investment attraction topics to local political issues.
Financing like Broadcom's can expedite equipment procurement but cannot replace power capacity and planning approvals. Therefore, the shortfall of AI infrastructure lies not only in chip capacity but also in whether projects can secure stable, sustainable local support.
(Source: NPR / Reuters / The Verge / Michigan State Local Legislation)
3｜Anthropic Advocates Private Deployment as Enterprise Procurement Shifts to Data Control
Anthropic announced that enterprise customers can use the service on their private cloud infrastructure without going through their servers. The latest Corporate Expenditure Observatory data from Ramp shows that OpenAI accounts for 44% of enterprise AI spending in the sample, while Anthropic accounts for 40%. These data do not represent the overall market share but indicate that the gap between the two on the enterprise procurement side is narrowing.
Competition is also entering the deployment and compliance aspects. Both companies have taken opposite stances on Massachusetts AI security legislation, and Anthropic's corporate investment division has also acquired an AI consulting company, expanding into on-the-ground services. They are no longer just comparing model invocation prices or scores.
Where enterprise customers keep their data, how to integrate into existing workflows, and who takes on compliance responsibilities will all impact procurement decisions. Model capability remains a prerequisite, but deployment architecture is becoming the watershed for turning trial users into long-term contracts.
(Source: Anthropic / Ramp / CNBC / TechCrunch)
4 | CFTC Prepares for Crypto Rules as Institutional Void Awaits Congressional Fill
The CFTC chairman has informed staff to prepare for developing cryptocurrency regulatory rules. The Senate has postponed the consideration of the CLARITY Act to September 15, keeping the timing of congressional legislation moving further back. Preparation does not equate to rules being issued, and the scope of CFTC authority, the inclusion of trading platforms, and the boundaries with the SEC still need to be seen in subsequent documents and processes.
While the bill's progress has slowed, exchanges, prediction markets, and AI companies are increasing their policy advocacy. What the industry needs is not just one regulatory agency taking the lead but a framework that can be aligned with securities regulation, consumer protection, and market structure.
If the CFTC initiates rulemaking first, it could reduce the institutional void in the short term and may open up authorization disputes and judicial review first. Regulatory certainty depends on rule text, enforcement boundaries, and Congress's final choice, rather than the mobilization signal of a single agency.
(Source: CFTC / CoinDesk / The Block / Politico)
5 | Iran Sanctions Escalate, Trade and Shipping Remain Effect Variables
U.S. Treasury Secretary Scott Benett stated that the new round of sanctions aims to weaken Iran's economic capabilities. Earlier, the UAE announced the suspension of trade, commercial dealings, and financial transactions with Iran, while an advisor to the Central Bank of Iran stated that they are prepared for more severe economic pressure. Several actions indicate that pressure has moved from spillover security risks to settlement and trade processes.
However, a sanction declaration does not automatically equate to a comprehensive blockade. Whether Iran can continue to maintain some energy and commodity flow depends on how third-party transactions, shipping insurance, and settlement channels adjust. The Axios report on the Hormuz shipping channel also indicates that parties are still trying to sustain some oil movement.
Therefore, the effect of sanctions should be measured by actual trade, shipping, and settlement data, rather than by verbal goals. The transit volume through the Hormuz Strait has not yet returned to normal, and economic pressure and energy flow are still interplaying. (Continuation of yesterday's report)
(Source: CNBC / Reuters / Axios / Central Bank of Iran)
Also Worth Knowing ↓
NVIDIA has denied reports that it will ship a new AI chip to China by the end of the year. The Information previously claimed the company had such shipment plans, which NVIDIA later denied. The back-and-forth between the report and the denial demonstrates that chip supplies to the Chinese market are still jointly affected by export policies and product arrangements. (Source: The Information / NVIDIA)
According to The Information, Meta has become Microsoft's largest AI customer, with AI compute usage on Azure surpassing Microsoft's own Copilot business. This development reflects a deepening collaboration between model training companies and cloud service providers, making it harder to distinguish between cloud vendors' customers and competitors. (Source: The Information)
SK Hynix is giving approximately $50,000 in annual bonuses to all employees. According to reports from Yonhap News Agency and Reuters, this arrangement sets a new record in the South Korean semiconductor industry. Profits driven by high-bandwidth memory demand, such as HBM, are being channeled into talent incentives and capacity expansion. (Source: Yonhap News Agency / Reuters)
Reportedly, Greg Brockman is adjusting the reporting relationships of multiple teams upon his return to OpenAI. The organizational arrangements' impact on the collaboration between research and development, product, and business teams will still need to be observed through subsequent personnel and product decisions. (Source: The Information)
Axios cites a study indicating that AI-generated content now makes up about one-third of new internet pages. The conclusion is influenced by the sample and definition of this estimation, but content moderation, source attribution, and retrieval quality are becoming common issues at the platform and tool levels. (Source: Axios)
According to CNBC and Bloomberg reports, Kevin Warsh will address debates around monetary policy space at the Jackson Hole Symposium. The market is more concerned about his public statements on policy independence, inflation, and interest rate paths rather than the event itself. (Source: CNBC / Bloomberg)
OpenAI has open-sourced the Codex Agent Framework. Developers can inspect and modify the agent harness located between applications and models, integrating Codex into their own products and workflows. This release focuses on the agent loop and integration layer, rather than the GPT model weights, expanding the competition for programming agents to the embeddability in enterprise software. (Source: OpenAI Developers)

Translation: Peggy
Editor's Note: On August 19, the U.S. Treasury Department announced an expansion of long-term bond repurchases, increasing the one-time maximum repurchase size for 10–20 year and 20–30 year treasuries from $20 billion to at least $40 billion. Previously, the 30-year bond yield had surged to around 5.34%, reaching its highest level since 2007. After the news was announced, the long-end yield quickly retreated.
This provided the market with a clear bullish signal: the Treasury Department is actively improving long bond liquidity, possibly even forming some kind of "Treasury backstop" expectation.
However, Marcus Nunes, in "The Treasury's $4 Billion Band-Aid," presented a contrarian view: while repurchases can indeed alleviate liquidity issues, if the long bond pressure stems from a larger fiscal deficit, higher bond supply, and weaker marginal demand, then a $40 billion repurchase does not address the true contradiction.
In other words, the market needs to distinguish between two things: the Treasury can make bonds trade better, but it cannot reduce the amount the U.S. government ultimately needs to finance through repurchases.
Below is the translation of the original article:
On August 19, U.S. Treasury Secretary Scott Bessent announced that the single-trade liquidity support repurchase size for 10–20 year and 20–30 year treasuries would be increased from a maximum of $20 billion to at least $40 billion. The market reacted swiftly. The 30-year Treasury yield, which had previously surged to around 5.34%, subsequently retreated significantly, with stocks, gold, and other assets rallying simultaneously.
However, the author of this article, Nunes, believes that this reaction easily leads the market to overlook a more fundamental issue: Treasury repurchases address liquidity, not the fiscal deficit.
Repurchases can improve trading but will not reduce government financing needs
Treasury repurchases are not quantitative easing.
When the Federal Reserve conducts QE, it can create base money to purchase treasuries by expanding its balance sheet; the Treasury does not have this ability. The funds used to repurchase old debt ultimately come from government cash or new debt financing.
Therefore, Treasury repurchases are fundamentally closer to debt structure management.
It can repurchase inactive old securities to increase market liquidity and, to some extent, boost demand for specific maturity bonds, but it does not change one fact: the U.S. government still needs to finance the fiscal deficit by issuing bonds.
The scale difference is particularly evident. The U.S. Treasury previously projected a need for net borrowing of $739.0 billion in the third quarter of 2026, while the size of this long-term bond repurchase has only been increased from $20.0 billion to at least $40.0 billion.
This is also why Nunes refers to it as a "Band-Aid." While $40.0 billion is enough to improve the market trading conditions for some long bonds, it is challenging to change the supply-demand dynamics of the entire U.S. bond market.
What is really weighing on long bonds is the growing fiscal supply
In Nunes's framework, the recent rise in the 30-year U.S. bond yield to above 5% cannot be simply understood as a liquidity issue. More importantly, the amount the U.S. government still needs to finance is significant.
In July 2026, the U.S. federal fiscal deficit reached $432.0 billion, a 48% year-on-year increase, hitting a record high for the month of July; the fiscal year-to-date deficit for the first ten months is around $1.8 trillion, already surpassing the full-year level of the 2025 fiscal year.
U.S. Federal Deficit—Annual Comparison
Meanwhile, the total U.S. federal debt exceeded $40.0 trillion on August 19. With the expanding debt scale and rising funding costs over the past few years, interest expenses are also increasing.
This means that the core issue the U.S. Treasury faces is not "old bonds are illiquid," but rather: who will absorb such a huge new bond supply in the future? If investors believe that future fiscal deficits will remain high, they will demand higher yields to absorb the long-term bond supply.
From this perspective, the 30-year yield breaking 5% may not be a sign of a temporary market malfunction but a repricing of U.S. fiscal and term risks.
The buying pressure issue is not something $40.0 billion can solve
Nunes also emphasizes the change in overseas demand.
According to the U.S. Treasury's TIC data, foreign holdings of U.S. Treasury securities decreased by approximately $72.1 billion month-on-month in June, with Japan, China, and the UK all experiencing varying declines. This is not enough evidence to prove that foreign investors are "massively fleeing U.S. bonds" because monthly holdings are influenced by exchange rates, custody locations, and asset allocation changes, and TIC data itself cannot fully identify the ultimate owners of securities, but it does indicate that the previously relatively stable overseas demand should not be taken for granted.
Foreign Holdings of U.S. Treasury Securities (June 2026)
Equally important is the buyer mix. If overseas official institutions become less willing to absorb U.S. debt, the U.S. will need to rely more on private investors. Private funds typically pay more attention to price and yield, meaning the market may require higher long-term rates to attract enough funds to absorb an increasingly large supply of bonds.
This is why simply increasing buybacks cannot solve the issue. While the Treasury can buy back some old debt, it cannot dictate at what price other investors will be willing to hold a significant amount of future U.S. long-term bonds.
The Real Divide: Is This a Liquidity Problem or a Fiscal Problem?
Proponents of expanding buybacks may argue that the Treasury is not attempting to address the fiscal deficit. Buybacks were originally a market liquidity tool, and as long as they can improve old bond trading and reduce market friction, they have achieved their policy objective. In this sense, criticizing buybacks by saying “$40 billion cannot solve the deficit” may itself blur the purpose of the policy tool.
However, the real issue raised by Nunes is: if the primary driver pushing up long-end yields has shifted from liquidity to fiscal supply, then continuing to use liquidity tools may inherently have limited effectiveness.
These two explanations correspond to two completely different market assessments. If recent long bond sell-offs are mainly due to insufficient market depth, deteriorating old bond liquidity, and short-term positioning impacts, then a Treasury buyback expansion may be sufficient to stabilize the market. But if the rise in long-term yields mainly reflects a persistent fiscal deficit, larger long-term bond supply, and higher term premiums, then buybacks can only smoothen the adjustment process, making it difficult to alter the ultimate yield level.
This is also the crux of this article: while the Treasury can improve the U.S. bond market's “trading problem,” it cannot buy its way out of the U.S.'s “fiscal problem.”
What truly needs to be observed next is not how much the Treasury's next buyback will increase, but whether long-term Treasury auctions will continue to receive sufficient demand, if the fiscal deficit will narrow, and if higher yields can once again attract foreign and private buying.
If these variables do not improve, then the yield retracement brought about by $40 billion is more likely to be a short-term cushion rather than a genuine reversal of pressure on U.S. long-term bonds.
[Original Article Link]

HYPE a bondi d’environ 58,5 $ à 73,7 $ au cours des deux derniers jours, marquant une hausse de 26 %. Il ne se trouve désormais qu’à environ 4 % de son plus haut historique de 77 $.
Alors que le prix s’approche d’un nouveau sommet, Hyperliquid observe également des évolutions positives.
Contrat pré-IPO entre à la SEC
La première information provient du Hyperliquid Policy Center. Il s'est associé à trade[XYZ] pour soumettre une lettre de commentaires à la SEC en réponse à la consultation publique de l’organisme de régulation portant sur la réforme du processus d’IPO, en proposant d’inclure des contrats perpétuels pré-cotés dans la discussion réglementaire.

Translation: Peggy
Editor's Note: This week, the U.S. 30-year Treasury bond yield rose to around 5.34%, reaching a high not seen since 2007. Subsequently, U.S. Treasury Secretary Besent announced an expansion of long-term bond repurchases, increasing the maximum single repurchase size of 10–30 year bonds from $20 billion to at least $40 billion, with the arrangement to be implemented between September 9 and November 4. Following the announcement, long-term yields retreated, the dollar weakened, and risk assets received some support.
Superficially, this was not a significantly large bond liquidity operation. The more worthy discussion is: why has the U.S. long-term rate risen to a level that requires a more proactive Treasury response? And is the market beginning to reinterpret the Treasury's "policy reaction function" to long-term yields?
In "Beware the Bond," as explained by Trader Joe, the recent selloff in long bonds cannot simply be attributed to inflation. The fiscal deficit continues to create bond supply, demand from traditional long-term bond buyers like Japan has shifted, and AI capital expenditure is generating a significant supply of long-term bonds in the credit markets, with these various forces collectively increasing the scale of long-duration assets that the market needs to absorb.
The author further likens Besent's expansion of long bond repurchases to the Treasury's version of "Operation Twist." This analogy captures the direction of "reducing market long-duration supply," but the two are not equivalent: the 2011 Operation Twist involved the Fed selling short-term bonds and buying long-term bonds, explicitly aiming to lower long-term rates and ease financial conditions; the current Treasury repurchase plan's official position is still to enhance secondary market liquidity and cash management. Therefore, what is truly worth noting is not the $40 billion itself, but whether this tool will increasingly take on a role in managing long-end financial conditions in the future.
Below is the translated original text:
Earlier this week, the U.S. 30-year Treasury bond yield rose to its highest level since 2007, leading to a partial retracement in U.S. stocks and a weakening dollar.
Then, Besent made a move.
The U.S. Treasury announced that it would increase the single repurchase size of some 10–30 year long-term bonds from a maximum of $20 billion to at least $40 billion, to be carried out between September 9 and November 4. Following the announcement, the 30-year Treasury bond yield retreated from its previous high of around 5.34% to around 5.2%, the dollar weakened further, and the stock market also stabilized.
U.S. 30-Year Treasury Bond Yield
The question is: Is this just a temporary fix for bond market liquidity, or does it signal a shift in the U.S. policy stance towards long-term rates?
To understand this, we first need to answer another question: Why has the long-end yield risen to this level?
Why Has the Long-End Yield Risen to This Level? It's Not Just About Inflation
The most intuitive explanation is inflation.
If investors are concerned that future inflation will remain high for an extended period, they will naturally demand a higher long-term Treasury bond yield as compensation. However, the author believes that this alone is not enough to explain recent developments.
At least from consumer surveys, there is no clear sign of runaway long-term inflation expectations. The University of Michigan's preliminary survey in August showed that the one-year inflation expectation had risen slightly from 4.2% to 4.3%, but the five-year inflation expectation remained at 3.3%. In other words, short-term inflation concerns still exist, but the "deanchoring of long-term inflation expectations" is not the only, and arguably not the most important, explanation.
Data Source: University of Michigan Consumer Survey
The long-term Treasury bond yield reflects more than just future short-term policy rates and inflation. Economic growth, term premiums, regulatory environment, how much the Treasury needs to issue in bonds, and how much long-term Treasuries insurance companies, pension funds, and foreign investors are willing to hold all affect long-end pricing.
What is currently most notable, according to the author, is that the supply of long-term bonds is continuously increasing, but traditional demand has not expanded synchronously.
The U.S. fiscal deficit means the Treasury still needs continuous funding, and whether this funding is done through short-term Treasury bills, medium-term notes, or 30-year long bonds directly affects how much duration risk the market needs to absorb.
If the Treasury relies more on short-term T-bills for funding, it reduces the long-term bond supply that the market needs to absorb, putting relatively less pressure on long-end yields. Conversely, if more funding shifts towards 10-year, 20-year, and 30-year securities, the market must absorb more duration, potentially putting greater upward pressure on long-end yields.
This is also why the debt issuance structure itself has increasingly resembled a macro variable.
Why is the Treasury Acting Now? 5.3% Long End Beginning to Impact Financial Conditions
Short-Term Debt Can Alleviate Long-End Pressure, But Liquidity Cushion is Thinning
The issue is that even short-term debt cannot be issued infinitely.
In recent years, when the U.S. Treasury issued a large amount of T-bills, a significant source of funding was the money market funds' funds originally placed in the Federal Reserve's overnight reverse repurchase agreement (ON RRP) facility. When short-term debt yields become more attractive, these funds can flow from RRP to Treasury securities, absorbing new short-term debt without significantly draining bank reserves.
However, this cushion is now nearly depleted. Federal Reserve data shows that ON RRP usage is currently close to zero on most trading days. At the same time, as of mid-year, the U.S. banking system reserves were around $3.1 trillion.
In the second half of 2025, the massive rebuild of the U.S. Treasury General Account (TGA) further drained bank system liquidity. Federal Reserve data shows that following the resolution of the debt ceiling issue, the TGA balance increased by approximately $442.0 billion at one point, leading to a notable decline in reserves.
This was also one of the backgrounds for the Federal Reserve's end of Quantitative Tightening (QT) in late 2025.
In October 2025, the Federal Reserve announced the halt of balance sheet runoff starting from December 1 and began to engage in Reserve Management Purchases (RMP) in December, purchasing short-term U.S. Treasury securities to ensure the maintenance of ample bank reserves.
These operations can easily evoke QE visually, but the policy objectives are different.
QE typically involves purchasing long-term Treasuries or MBS, actively lowering long-term yields, easing overall financial conditions; RMP primarily buys short-term securities like Treasuries, with the official aim of maintaining sufficient bank reserves and controlling short-term rates, rather than providing macroeconomic stimulus. The Federal Reserve also emphasizes that RMP does not indicate a change in monetary policy stance.
The concern is that if the Treasury continues to increase the proportion of short-term financing to reduce long-term supply, then when liquidity buffers like RRP are nearing depletion, new short-term debt may ultimately compete more with bank reserves.
At that point, the Federal Reserve may have to conduct more reserve management operations to maintain systemic liquidity. This sets up a delicate policy mix: the Treasury minimizing duration released to the market, while the Federal Reserve ensures ample reserves at the short end.
Japan and AI are Both Transforming the Supply and Demand Structure of Long-Term Bonds
The issue on the long end has another side: who will buy?
Japan has long been a key foreign investor in U.S. Treasury bonds. The latest U.S. Treasury International Capital (TIC) data shows that as of June 2026, Japan held approximately $1.116 trillion in U.S. Treasury bonds, remaining the largest foreign holder, but down about 2.3% from May.
At the same time, Japan's own long-term government bond yields are rising.
For domestic institutions in Japan such as insurance companies, banks, and pension funds, if Japanese government bonds themselves can offer increasingly attractive yields, the marginal attraction of allocating to U.S. long-term bonds may naturally decrease, especially after taking into account the USD hedging costs.
This does not necessarily mean that Japan will continue to sell U.S. bonds on a large scale, but it does mean that a structural buyer of long-term bonds that has long existed in the past may no longer consistently absorb U.S. duration as stably as before.
Another competitor comes from AI.
The AI infrastructure build-out is transitioning from a stock market story to a credit market story. Goldman Sachs research estimates that from 2026 to date alone, the entire AI-related ecosystem has issued close to $500 billion in debt; of which the hyperscale cloud players themselves have issued around $194 billion. More importantly, it's the tenor. This year, in the U.S. investment-grade credit market, about 40% of new issuances with a tenor of 15 years or longer have come from AI firms or AI-related financing.
This means that traditional long-duration funds such as pension funds and insurance companies are facing more choices. They are no longer just comparing 30-year U.S. Treasury bonds and other sovereign debt, but can also purchase long-term investment-grade bonds of large tech companies like Amazon, Google, and AI-related credit assets such as data centers and infrastructure.
From the author's perspective, this makes the core issue facing U.S. long-term bonds even clearer: the Treasury needs to sell more and more debt, while the other long-term assets that global markets need investors to absorb are rapidly increasing.
Treasury's Version of Operation Twist: $4 Billion Is Not Huge, but the Real Change Lies in the Policy Response
It is also in this context that the Bizarro expands its long-term bond buybacks.
The U.S. Treasury's regular buyback program began in 2024, with the official setting two purposes: to improve secondary market liquidity and for cash management.
Notably, the liquidity support repo primarily purchases older, less liquid securities, known as off-the-run Treasuries. The Treasury Department proactively positions itself as a potential buyer of these bonds, aiming to assist dealers in depleting inventories and enhancing the trading dynamics of these older securities. (U.S. Department of the Treasury)
Therefore, from a structural perspective, this is not a QE tool designed to cap the 30-year yield.
Moreover, a one-time purchase of at least $40 billion in the vast $30 trillion U.S. Treasury market remains relatively small. Reuters also notes that the market widely believes this scale is insufficient to address structural issues such as the fiscal deficit and long-term supply expansion.
However, the author's true focus is not on the scale but on the policy intent.
In the past, the Treasury could emphasize that repos were merely market liquidity tools; now, as the 30-year yield rapidly approaches a two-decade high, the Treasury swiftly expands long-dated bond repos, prompting market participants to question: if the long end continues to spiral out of control, will the Treasury further adjust its repo and issuance structures in the future?
This is why the author refers to the current policy as the Treasury's version of "Operation Twist."
Note: Operation Twist is often referred to as "扭曲操作" or "期限延长操作" in Chinese. Its essence is not "printing more money" but adjusting the central bank's bond maturity structure: selling short-term bonds, buying long-term bonds to depress long-term rates.
The classic 2011 Operation Twist was conducted by the Federal Reserve: selling or letting short-term bonds mature while simultaneously purchasing an equivalent amount of 6-30 year bonds, elongating the asset portfolio duration without expanding the balance sheet, reducing the private sector's holdings of long-term bonds, and lowering long-term rates.
What is happening today is not an exact replica of the same operation. The Treasury is not executing a strict "sell short, buy long" like the Fed did back then, and the expanded repos are still officially defined as debt management and liquidity tools. However, from a market duration supply standpoint, both are somewhat aligned: if the Treasury continues to repurchase more long-term old bonds while leaving more net financing pressure on the short end, the net duration that the private market needs to absorb may relatively decrease.
This is what the author refers to as the "Treasury's version of Operation Twist." More precisely, it is currently a market interpretation rather than an established new policy framework.
Can This Approach Anchor the Long End? Risks May Shift to the Dollar and Inflation
So, under what circumstances would this policy framework continue to escalate? The author argues that instead of looking for an absolute 30-year yield "red line," it is better to observe the speed of the yield increase. A 30-year yield at 5.2% or 5.3% may not be sufficient on its own to trigger a policy change; but if the market begins to see consecutive rapid jumps of around 10 basis points each time, indicating a significant deterioration in market liquidity and demand, the probability of further intervention by the Treasury Department or the Federal Reserve would increase.
Meanwhile, long-dated bond yields have become more directly competitive for funds with equities. As per the data available when the author's article was published, the nominal yield on 30-year U.S. Treasuries is around 5.2%, while the real yield on long-term TIPS is close to 3%; in comparison, the S&P 500 earnings yield is around 3.8%.
While these cannot be directly compared — the earnings yield is not a risk-free rate, and corporate earnings are expected to grow or decline in the future — when the risk-free long-term real yield rises to such a high level, it is evident that the opportunity cost that stock valuations must bear is increasing.
Therefore, the key significance of Bostic's recent move may not be in temporarily pulling the 30-year yield back from above 5.3% to around 5.2%.
It lies in the market getting a new sample of observation: when U.S. long-end yields rise rapidly, will the Treasury Department become increasingly proactive in responding through repurchase sizes and debt maturity structures?
If the answer gradually shifts to "yes," then in the future, the impact on the U.S. dollar, U.S. stocks, gold, and long-term government bonds will involve not just the Fed's policy reaction function but also this additional layer of the Treasury Department.
However, this logic also has its boundaries. If the rise in long-dated yields is primarily due to a bond supply-demand imbalance, reducing the duration that the market needs to absorb may alleviate the pressure; if inflation expectations notably rise again, continuing to expand repurchases, increasing short-term debt financing may instead make market participants concerned that policies are artificially suppressing financial conditions.
Therefore, what truly needs to be observed next is not just whether the Treasury Department will increase repurchases but whether inflation expectations, the structure of long-term bond issuances, overseas demand, and the speed of long-end yield fluctuations are all changing simultaneously.
Only when these variables continue to point in the same direction will the author's proposition that the "Treasury Department is taking over a part of long-end financial conditions management" receive further validation.
[Original Article Link]

Translation: Peggy
Editor's Note: On August 19, the U.S. Treasury Department announced an expansion of long-term bond repurchases, increasing the size of the single-day liquidity support repurchase for some 10–30 year bonds from $20 billion to at least $40 billion. Previously, the 30-year Treasury yield had briefly risen to 5.337%, reaching a new high since 2007; after the news was released, the long-term yield swiftly retreated.
The market immediately began to wonder: Is the Treasury Department showing a more explicit policy sensitivity to the rapid rise in long-term rates?
In his publication on August 20 titled "Hot Take: Bessent Makes His Mark," Stephen Innes discussed precisely this issue. He focused not on the $40 billion repurchase itself but on how this move could change the market's understanding of the U.S. policy "reaction function": when long-term yields rise high enough and fast enough, will the Treasury Department take action again and transition from a mere debt manager to another hand influencing financial conditions?
This development is significant because the policy constraints in the U.S. are increasingly concentrated at the long end. The Fed can directly determine short-term rates but cannot completely control term premia, fiscal supply, and the 30-year Treasury yield. If the Treasury Department starts more actively managing long-end market pressures, the framework of only focusing on the Fed to judge financial conditions will become incomplete.
Innes' key insight is not that the U.S. has entered yield curve control but that the market is starting to realize that Washington may have an undisclosed "policy pain threshold." Once traders believe that a certain yield level will trigger Treasury action, the pricing of future long-term rates will no longer depend solely on inflation, the Fed, and bond supply and demand but will also add a new variable: how high the Treasury Department can tolerate long-term yields rising.
Below is the translation of the original text:
U.S. Treasury Secretary Scott Bessent's latest move has added a new trading variable to the long-term U.S. bond market.
On August 19, the U.S. Treasury Department announced that it would at least double the size of the liquidity support repurchase for certain long-term nominal bonds. The single repurchase limit for 10–20-year and 20–30-year bond maturities will be increased from the current $20 billion to at least $40 billion. The new arrangement will take effect on September 9 and continue until the end of the current quarter's refinancing period on November 4. The Treasury Department's official explanation is to provide more liquidity support for long-term bond maturities.
In terms of scale, this is not a large enough operation to change the overall supply-demand dynamics of the U.S. Treasury market. The U.S. Treasury's bond repurchase mechanism itself is not QE: the Treasury mainly repurchases older, less liquid outstanding bonds to improve market liquidity, rather than injecting base money into the financial system through expanding the central bank's balance sheet as the Fed does with QE.
However, in the view of the author Stephen Innes, what the market is truly trading is not the $40 billion number, but the policy signal behind it.
Limited in Scale, Market Trades First and Foremost on "Why Now"
After the Treasury's announcement, long-term U.S. bond yields quickly fell.
On August 19, the 30-year U.S. bond yield briefly dropped by nearly 10 basis points to 5.187%. The day before, the yield had touched 5.337%, the highest level since 2007. At the same time, the long-end of the yield curve flattened, the U.S. dollar weakened significantly, gold broke $4,500 per ounce, Bitcoin rose, and the three major U.S. stock indexes all closed slightly higher.
These market changes coincided with the Treasury's announcement. However, a more accurate statement is: the market's expectations of the Treasury's policy response changed, rather than the Treasury directly lowering yields on that day by adding repurchase funds.
This is because the expanded repurchase size will not take effect until September 9. Therefore, what is most worthy of attention this time is the "signaling effect."
The author of this article points out that the Treasury could have announced this adjustment during the previous quarter's refunding arrangement, but did not do so at that time. Following that, there was a noticeable sell-off in long-term U.S. Treasury bonds over two weeks, pushing the 30-year yield to nearly a 19-year high before the Treasury separately announced the increase in the repurchase size.
The Treasury did not indicate that this adjustment was aimed at supporting a specific yield level; the official reason remains to enhance liquidity for long-term securities. However, in the author's view, traders will naturally wonder: if long-term yields rise rapidly again in the future, will the Treasury further adjust its policy?
This is precisely what he calls the "Bessent Makes His Mark" core.
From Yen Intervention to Expanded Repurchase, Market Begins to Test Treasury's "Threshold for Action"
This speculation arose quickly also because this was not the first recent event where the U.S. Treasury clearly intervened in market operations.
From late July to early August, the U.S. and Japan made a rare coordinated intervention in the foreign exchange market to support the yen. Less than a month later, the Treasury Department announced an expansion of long-term bond repurchases after long-term bond yields rose to multi-year highs. The two actions targeted different markets, and their policy objectives cannot be simply equated: the former directly addressed exchange rate volatility, while the latter was officially positioned as liquidity management in the government bond market.
However, Innes believes that they collectively influenced investors' assessment of the Treasury Department's reaction function — the market began to speculate on how much and how quickly the Treasury Department might act once asset prices experienced significant fluctuations.
This is also the essence of the so-called "Bessent Put." It is not a formal policy, nor does it mean that the Treasury Department has committed to buying bonds at a certain yield level, nor is it equivalent to Yield Curve Control (YCC).
More precisely, this is a market inference: if traders start to believe that the Treasury Department has an intolerable "pain threshold," whenever the 30-year yield approaches a high point in the future, the market may speculate on whether the policy will be adjusted again.
In other words, the market is looking for the Treasury Department's "intervention threshold": at what level will the long-term bond yield rise to prompt Washington to take action again?
The first intervention informed the market that the Treasury Department is monitoring pressure in the long end of the market. Only if similar actions are taken in the future, the market may further assess whether the Treasury Department truly has a relatively clear policy trigger range.
Whether the "Fed Put" is transitioning to the "Bessent Put" remains the author's judgment
Innes further discussed this change within the framework of the Federal Reserve under Kevin Warsh's leadership.
His assessment is that the Federal Reserve under Warsh aimed to reduce market reliance on central bank-induced suppression of volatility and allow more price discovery to return to the market itself. Therefore, if the Treasury Department becomes more willing to take action in case of sharp fluctuations in bonds or exchange rates, the familiar "Fed Put" from the past may not completely disappear but could show signs of shifting towards the Treasury Department.
It is important to make a clear distinction between facts and judgments.
The confirmed facts are: the Treasury Department recently participated in U.S.-Japan coordinated exchange rate intervention and announced an increase in the scale of long-term bond repurchases; the Federal Reserve still independently manages monetary policy. As for the idea that "policy support is shifting from the Federal Reserve to the Treasury Department," this is Innes' interpretation based on these two market operations and is not a new policy framework announced by the U.S. government.
Similarly, there is currently no evidence that the Treasury Department is implementing yield curve control. In fact, the scale of long-term Treasury bond buybacks remains small. A one-time $40 billion operation is limited compared to the over $30 trillion U.S. Treasury market, and buybacks have not addressed structural issues such as the fiscal deficit, inflation expectations, and long-term bond supply and demand that have led to rising yields.
Therefore, the "Bessent Put" is currently better suited as a concept to describe market expectations rather than an established policy tool.
Fed Still Discussing Rate Hike, Treasury Signals Adding Complexity to Trading Framework
Meanwhile, the signals from the Fed are not distinctly dovish.
The FOMC meeting minutes from July 28 to 29 showed that most members supported keeping the federal funds target rate at 3.50% to 3.75%, but "several" members were inclined to a 25-basis-point rate hike at the next meeting. The minutes also indicated that many members believed that further monetary policy tightening might be necessary if inflation did not continue to decline; ultimately, three members voted against keeping the rate unchanged and favored a 25-basis-point hike.
However, these minutes reflect the policy assessment at the end of July. Subsequently released inflation and employment data have been relatively moderate, leading the market to lower expectations of further rate hikes. This has made the current policy outlook more complex.
On the one hand, there are still voices within the Fed advocating for further monetary policy tightening; on the other hand, the Treasury Department's announcement of increasing long-term bond buybacks has prompted the market to reassess the upside potential of long-term yields, accompanied by a decline in yields and the U.S. dollar.
Innes therefore believes that simply trading U.S. financial conditions around "the next Fed rate hike or cut" may no longer be sufficient.
It is important to emphasize that the Treasury Department's announcement of expanding buybacks does not directly equate to monetary easing. A more accurate statement is that following the announcement, the decline in long-term yields and weakening of the U.S. dollar have had a certain marginal easing effect on financial conditions from a market pricing perspective; whether this effect can be sustained still depends on subsequent inflation, fiscal supply, and market demand.
This is the real issue to watch in the near future.
The Treasury Department has clearly stated that this increase in buybacks will only last until November 4 and will provide further information on future arrangements at the next quarterly refunding meeting. Therefore, for the long-term U.S. bond market, the key is no longer just the next set of inflation data or the next FOMC meeting.
The market is now also watching another variable that has not been so prominent before: if the 30-year Treasury bond yield approaches 5.3% again or even higher, will the Treasury Department adjust its tools again? In other words, what the market is currently testing is: how high does the long-term Treasury bond yield need to rise to trigger the Treasury Department's next move?
Currently, this is still a hypothesis that traders are testing, rather than a yield threshold that has received policy confirmation.
[Original Article Link]

According to PolyBeats News, on August 19th, Beijing time, an individual managed to achieve a 120% return in just 89 seconds by predicting "whether Russia would be able to occupy a certain area in the battlefield."
This individual invested $4,000 to purchase the prediction "Russia will capture Konstantinovka this month" on a platform called a prediction market. The probability of this event happening was only 43% at the time. 89 seconds later, this probability increased to 92%.
In this probability-price platform, if the event does occur, the probability will be locked at 100%, and the position value will increase accordingly. The mysterious individual placed the order, witnessed a sudden surge, and then exited at the take-profit in less than two minutes. In an era where various trading strategies abound, such stroke of luck in trading is not uncommon.
What makes this individual suspicious is that Konstantinovka was his 14th such operation.
14 Consecutive Successful Battlefield Predictions
Analysis of the account's transaction records by PolyBeats revealed that this individual's account had a total of 14 correct prediction trades related to the Russia-Ukraine conflict, realizing a total profit of approximately $17,500.
These predictions did not pertain to the same city or the same timeframe: from Maliyevka, Pokrovka, to Huliaipole, and Konstantinovka; some were about "whether Russia will capture," while others were about "whether Ukraine will re-enter." However, the operations were nearly identical as if copied from the same template.
Even more bizarrely, not only did this account accurately predict the outcome 14 times, but 11 of these trades saw the probability surge to nearly 100% within 5 minutes after the purchase.
It is difficult to consecutively guess the battlefield developments correctly 14 times and buy in before the market probability escalates. However, if the advance knowledge was not of the battlefield but of the map update time, the scenario becomes much simpler.
Prediction Markets: Breeding Ground for Insiders
All of Polymarket's Russian-Ukraine occupation market titles are almost a template:
“Will Russia take over a certain region by ___?”
While the question may seem straightforward, what truly determines the answer is not when the gunfire starts.
According to Polymarket's rules, only when the Russia-Ukraine battleground map produced by ISW—the Institute for the Study of War—labels a specific region as Russian-controlled and meets the duration requirements set by the rules, will the market move to a "Yes" outcome—corresponding to a 100% settlement probability.
In other words, even a soldier standing in the occupied territory cannot insider trade to stable long-term gains in such a market: even if the occupation indeed happens, as long as ISW, the source of settlement, does not update the map, this market will settle as "No."
In such a market, only one type of person can achieve perfect profitability like the aforementioned account: he doesn't need to know the frontline situation, understand Russo-Ukrainian geopolitics and military strategy, or even follow any international current events. He only needs to know in advance which area ISW is ready to mark as Russian-controlled, and when that update will take effect.
This group of people is the technical staff responsible for updating the battlefield status at ISW. They only need to predict the corresponding market long positions to buy into right before clicking the mouse to confirm the ISW map update, enabling them to make a substantial profit within a minute.
Sound like a conspiracy theory? Last year, ISW had already experienced a similar farce.
Employee Manipulates Open-Source Intelligence Data for Massive Profits
In November 2025, the prediction market had already previewed how this mechanism would spiral out of control.
At that time, the market question was, “Will Russia take over Mironohrad before November 15?” The rules were almost identical to the above—only when the ISW map shows any part of the Russian military controlling a specified street intersection, the outcome could be determined as “Yes.”
The most significant difference from the current rules was that there was no "waiting period" back then: as soon as the map changed, the market settlement condition could be met.
On the evening of the 15th, the probability of "Yes" was less than 3% as there were only a few hours left until the deadline—without any geopolitical news reporting the occupation of Mironohrad, everyone had already assumed that this market would ultimately settle as "No."
However, at this moment, the ISW map suddenly showed the area as occupied; before most people could even react, the probability of "yes" was instantly pushed to 100%.
Over the next few days, as traders questioned the lack of concrete evidence for this map marking, ISW publicly admitted that there had been an "unauthorized, unapproved" edit on the interactive map during the night of November 15 to 16; this edit was removed before the start of the normal workflow on the 16th and did not represent its official battlefield assessment.
The map was retracted, but the market did not retract, and those accounts that predicted "yes" when the probability was approaching zero profited greatly from this "unauthorized edit." Among them, 0x69A9 bought in when the probability was only 0.9% and ultimately achieved a 10,888.68% excess return.
Following the Mironohrad incident, ISW stated that its map would continue to be modified during workdays and should not be understood as real-time battlefield conditions; it later added "being edited" and "finalized" status indicators. Prediction markets later also implemented stricter persistence requirements: the map status had to span the next full update cycle to be used as a settlement basis.
As of now, the mastermind behind these 14 consecutive victories has not publicly revealed their identity and has not been accused by any institution; ISW has also not acknowledged any employees benefiting from map update arbitrage.
But the timeline of the 14 victories is clear: almost every time, it placed bets on a city considered "undetermined" by the market; almost every time, it waited until the market was close to certainty before selling; almost every time, it pocketed the difference between "no one knows yet" and "everyone knows."
In the past, a regular ISW employee responsible for updating the map might have been just a nameless screw in a vast open-source intelligence system. What he held in his hands was just a mark about to change color, an update about to be released, and a few seconds before the click of confirmation.
These pieces of information were originally untradeable, and few would have cared.
But when a map can determine the final market settlement, those few seconds before clicking the mouse now have a price. A person doesn't need to disclose top-secret military intelligence, manipulate the battlefield, or even change any facts; all he needs is to know about a mark that is about to appear before everyone else, turning the authority assigned to his position into profits close to certainty.
Innovation has opened a market for everyone to predict the future, but it has also opened another door: when the settlement source can be seen in advance, modified in advance, or traded in advance, you and I can all become insiders.

Emails, chat records, project documents, work orders—these used to be just digital remnants awaiting clean-up after a company shut down. Now, they are being reassessed, packaged, sold, and fed into an AI company's training pipeline.
An undertaker preserves the deceased's final dignity. The postmortem dignity of an enterprise is to prove that what they left behind still holds value.
On August 17, at a bankruptcy asset auction, Google bid $10 million to acquire all of Spirit Airlines' corporate data. Another bidder, Mercor, bid $7.5 million, falling short by $2.5 million.
The auctioned items were divided into three parts. The first part consisted of approximately 100 million employee emails. The second part included 500 million Microsoft Teams messages, totaling 600 million. The third part comprised calendars, spreadsheets, financial databases, project files, operational records, and a batch of internal software. Passenger profiles, frequent flyer information, and the like were not included.
600 million messages—if one person speaks 100 sentences a day, they would have to speak non-stop for over 16,000 years.
Calculating, each message was sold for 1.67 cents. Americans refer to a 1-cent coin as a penny, often not bothering to pick it up if it falls on the ground. In other words, a Spirit employee's sentence in Teams was worth one and a half pennies.
Going back to 1980, Spirit Airlines was born in Detroit, born from a trucking company called Charter One, which later transitioned to aviation. In 1992, it was renamed Spirit, meaning soul. Over the following 34 years, it set the benchmark for the entire industry to replicate the model of ultra-low-cost carriers in the U.S.
It once had a solid foundation—205 all-Airbus fleet, approximately 300 flights a day, and a projected 2024 revenue of around $5 billion. However, it faced a net loss of about $1.2 billion and carried nearly $9 billion in debt when it filed for bankruptcy.
On a late night in May 2026, the company announced it was ceasing operations. The next day, around 17,000 employees found out they were unemployed through the news.
Proceedings are still ongoing, and the transactions are awaiting approval from the bankruptcy court. Spirit is undergoing a liquidation-type closure under Chapter 11 bankruptcy, with no bankruptcy trustee taking over. The company is still supervised by the court and is selling off its remaining assets piece by piece. The sensitive data involving employee emails, Teams messages, etc., must first be handed over to an independent entity for processing. This entity was selected by Google, with Google also bearing the costs.
Furthermore, scouring through public reports, the bankrupt company selling internal data to an AI company is unprecedented. This deal is most likely the first of its kind in history.
The long-standing U.S. tech publication Gizmodo titled this transaction as follows: "Spirit is Dead, But Its Data Will Haunt Google's Servers for Generations."
A New Business Venture
There have always been people who handle the assets of defunct companies. Lawyers, liquidators, auction houses—doing this for decades. Aircraft, furniture, trademarks, patents—all sellable assets have long been sold off.
What's truly new is that starting this year, even a company's internal employee data has been put on the table.
This emerging business has two underlying reasons.
First, the number of defunct companies has increased.
In the first quarter of 2024, the failure rate of U.S. startups surged by 58% year-on-year, and the number of active VC firms decreased by 62% from its peak. The money hasn't decreased; it's just flowing more concentratedly towards AI. In 2024, U.S. AI startups raised a record $97 billion in funding. The capital market still has money, but it's becoming more reluctant to spend outside of AI.
As a result, a group of companies that could have continued to survive on funding are now hitting a wall earlier. In August 2024, the fintech company Tally, backed by a16z, announced its closure. Having raised a total of $172 million, with a peak valuation of $855 million, reaching Series D, it still couldn't secure the next round of funding.
Second, data has become more expensive.
The consumption of data for large-scale model training has reached unprecedented levels. The training data for GPT-4 consists of around 130 trillion tokens. For comparison, Google Books scanned over 40 million books in over forty years, which amounts to about 40 trillion tokens. In other words, the amount of data used for one GPT-4 training is equivalent to over three times the content of Google Books.
Epoch AI ran some numbers and concluded that high-quality language data from books, news, and Wikis will be depleted around 2026. The high-quality text available on the public internet is quickly running out, with synthetic data accounting for an increasing share. Next, AI companies will have to look beyond the public internet for new data. Years-worth of internal company emails, chat records, and work documents have now come into view.
As for the AI training dataset market, research institutions predict it could reach $9.7 billion by 2030. Taking into account various licenses, the total size of the pie is estimated to reach $67.5 billion.
On one side, more and more companies are closing down, leaving behind a large amount of internally generated data that was previously not priced; on the other side, AI companies have an increasing demand for data beyond the public internet. With both sides happening simultaneously, it is the first time that internal corporate data has the conditions for scalable transactions.
What truly adds value to this type of data is the development of Enterprise Agents. Gartner predicts that by 2026, 40% of enterprise applications will embed task-specific AI Agents, a figure that was less than 5% the previous year.
The training material required for Enterprise Agents is not exactly the same as that for large-scale models. While public web pages can provide knowledge, language, and final content, it is challenging to replicate a company's actual work processes. Communication and collaboration involve many real details, such as how a requirement is proposed, how several people discuss it, how tasks are assigned, how issues are addressed, and ultimately how delivery is made.
These processes are extensively recorded in a company's internal emails, chat logs, tickets, and project documents.
When such data begins to have clear buyers and use cases, things that were previously directly deleted when a company shut down now have standalone transaction value.
Body Snatcher
In the past, this job was done by liquidation lawyers, but now three new types of people have emerged.
The first is called a dissolution service provider, represented by SimpleClosure.
This company does only one thing: helping startups die gracefully. In 2023, the company just started, raising $1.5 million in pre-seed funding, and in May 2025, it secured $15 million in Series A funding, led by TTV Capital. Even Carta, which handles equity management and corporate affairs for many US startups, shut down its own closure service, invested in SimpleClosure, and handed over this part of customer demand to it.
By October 2025, SimpleClosure had already conducted funerals for over a thousand companies. Crunchbase gave it a nickname, "A Better Way To Fail," a kind of better way of failing. While American entrepreneurs love to say "fail fast," SimpleClosure argues that fast failure is not enough; it must also be dignified. The company even has a pricing calculator on its website – enter your company's situation, and it will calculate how much it costs to die once.
A funeral is never a waste. In April 2026, SimpleClosure launched the Asset Hub, specifically designed to handle intangible assets left behind after a company shuts down. In addition to brands, software, and customer lists, for the first time, items such as Slack messages, emails, and Jira tickets – internal work data – were explicitly put on the shelf.
This indicates that before Spirit, the market had already begun to attempt to price the internal data of dead companies, although at that time it was still startups, small transactions, and private dealings.
There is a ready-made case. When transcription and captioning company cielo24 shut down, they sold off Slack messages, internal emails, and Jira tickets accumulated over the past 13 years through SimpleClosure. CEO Shanna Johnson later told Forbes that this batch of data eventually sold for hundreds of thousands of dollars.
For a company that has already decided to close its doors, this was originally a batch of data that needed to be cleaned up, but in the end, it became an asset that could be recovered during liquidation.
For the data sales stage, SimpleClosure handed it over to Protege. This is a data exchange market specializing in AI training data licensing. In January 2026, they just received a $30 million investment led by a16z, and the founder is Bobby Samuels. Protege's initial entry point was medical imaging, where within 30 days, they gathered millions of images for pre-training for buyers.
Now, Protege is starting to apply this data licensing and trading ability to internal communication data of closing companies. SimpleClosure is responsible for company closure and asset organization, while Protege is responsible for finding buyers, completing data licensing, and transactions.
The second type of participant is bankruptcy courts and liquidation lawyers. For decades, they have been counting planes, tables, chairs, trademarks, and patents. Now, the list is beginning to include emails, Slack messages, and other internal data.
According to U.S. bankruptcy law, this data can be included as intangible assets in bankruptcy estates and sold under court supervision. Relevant law firms have also begun to establish specialized teams to handle data preservation, discovery, and organization in bankruptcy cases. Redgrave LLP has such a restructuring and discovery business.
The third type is e-discovery service providers responsible for technical execution. Companies like KLDiscovery, Epiq, and Consilio are usually involved in collecting, organizing, hosting, and reviewing enterprise data. The content in emails, Teams, and SharePoint needs to be exported, archived, and organized by them before being packaged into data that can enter the transaction process.
This industry itself already has a mature set of billing methods. EDRM regularly publishes pricing surveys, with common billing units including data collection per GB, hosting per GB per month, and document review fees.
Spirit’s 600 million messages eventually turned into an auction item, relying on this type of foundational work. However, compared to court documents and auction bids, this part rarely appears in public reports. The specifics of who handles it and how it is handled are usually not visible to the outside world.
Autopsy Checklist
For a corporate Agent, what it needs to learn is not just knowledge and standard answers, but also judgment, collaboration, error correction, and execution processes in the real work environment.
The final product tells the model what it has become, while internal records tell it how it was made.
This change has already been reflected in Agent training data. In the past, a single-line of code training sample might only consist of a few hundred tokens, involving modifying a few lines of code; now, an Agent training sample often needs to include the entire process of requirement understanding, file locating, code modification, and test verification.
Training data is transitioning from a single answer to full task execution records. For the Agent, the end result is of course important, but the judgments, operations, and feedback left during task completion are more valuable for training.
Compared to companies operating normally, the data of a shutting-down business is more likely to enter the transaction process. While the company is still operating, selling internal communications would involve trade secrets, employee privacy, non-compete risks, and customer relationships, and legal and management teams are usually very cautious. After entering liquidation, the company's main goal changes to recovering as many remaining assets as possible to secure more value for creditors.
That's why the same batch of internal data, which was difficult to sell while the company was alive, may be revalued during the shutdown phase.
The value of internal corporate data has long been recognized by the industry. Salesforce has always regarded the enterprise communications accumulated in Slack as important data assets, and Microsoft CEO Satya Nadella has emphasized multiple times that when enterprises use AI, a truly valuable part is their proprietary data and work context.
In the past, this data primarily served the company itself. Now, as AI companies begin to actively seek enterprise internal data, they have, for the first time, more clearly defined external buyers.
The Art of Pricing
While this business is still in its early stages, some reference prices have emerged in the market.
SimpleClosure and Protege handle data from closing startups, with individual transactions typically ranging from $10,000 to $100,000. Mercor offers quotes based on the chat logs and emails of acquired startup employees, with the highest reaching $300,000.
Spirit has taken the pricing to the tens of millions of dollars. Mercor bid $7.5 million, and Google ultimately offered $10 million, competing for around 600 million internal communications and other corporate data. Based solely on these 600 million messages, the average price per message is approximately 1.67 cents.
This unit price is not high. In 2024, Reuters reported that Photobucket negotiated licensing deals for about 13 billion photos and videos with an AI company, with prices around 5 cents to $1 per photo and over $1 per video. In the B2B data market, a single contact's information can be sold for a few cents to a few dollars depending on completeness and accuracy.
However, these prices are not yet enough to form a unified standard. The value of internal data from closing companies is currently mainly a matter of individual negotiation. Factors such as data volume, industry, time span, completeness, uniqueness, and what the buyer intends to do with it will all affect the final price. Spirit's $10 million bid appears more like one of the few publicly visible large-scale examples at present.
When compared to established data licensing markets, the gap becomes even more apparent. Reddit licensed user posts and comments to Google for approximately $60 million annually; News Corp's content licensing agreement with OpenAI is around $250 million for five years; xAI's partnership with Telegram amounts to $300 million; and Apple's purchase of Shutterstock image licenses falls between $25 million and $50 million.
These markets already have established buyers, licensing mechanisms, and pricing experiences. Transactions of internal data from closing companies are just getting started, with no clear rules yet on which data is most valuable or whether valuation should be done per item, by capacity, or as a whole.
Looking at Spirit's $10 million within the company's own scale is a different story. In 2024, Spirit's annual revenue was close to $5 billion, averaging around $13.7 million per day. The amount Google spent to acquire this data is less than what it makes in a day during normal operations.
For a bankruptcy liquidation, this is just a small recovery from the remaining assets; for an AI company, it is acquiring a set of long-unseen data containing the real operational processes of the business.
Cleansing and Handover
After the data transaction, it cannot be handed over directly to the buyer. Before the formal handover, it usually needs to go through several steps such as export, de-identification, organizing and packaging, court approval, and final delivery.
The first step is export. Slack's corporate data is usually exported as a ZIP file, including JSON files organized by channel, member information, and attachments.
Microsoft 365, on the other hand, can export emails and Teams messages through an eDiscovery tool. The Spirit transaction involves approximately 600 million internal communications, a large amount of data that requires processing in batches in practice. The specifics of whether this is carried out by an eDiscovery service provider or the buyer's engineering team have not been disclosed in publicly available documents.
Next is de-identification, which involves removing as much information as possible that can be traced back to specific employees. Common methods include identifying and masking personal information such as names, phone numbers, and addresses, or replacing sensitive fields with identifiers that do not directly correspond to individuals. In some statistical and training scenarios, additional random perturbation is applied to further reduce the possibility of re-identifying individuals.
However, de-identification does not equate to absolute anonymity. Even if names and emails have been removed, as long as there are enough professional, temporal, locational, or behavioral features retained in the text, there is still a possibility of re-identifying individuals through other information.
Therefore, the party responsible for this step in the Spirit transaction is crucial. According to the current transaction arrangement, a third-party independent entity will handle the de-identification process, chosen by Google and at Google's expense. Google has also committed not to exploit this dataset to re-identify individuals.
Data sales by bankrupt companies are not unprecedented. Over the past two decades, from Toysmart and Borders to RadioShack and 23andMe, cases have emerged involving the handling or sale of customer data during bankruptcy. Transactions of this nature involving consumer privacy usually face stricter scrutiny from courts, regulatory bodies, and state governments.
What sets Spirit apart is that this sale does not revolve around passenger lists but rather around employee emails, Teams messages, and other internal work data. Existing bankruptcy procedures have not established mature rules for dealing with this type of data as they have for consumer information.
Controversy has already arisen. Spirit's flight attendants union has raised objections to the transaction, leading the court to postpone approval. What was initially a batch of corporate data intended to be sold as bankruptcy assets has now become entangled in employee rights and the boundaries of AI usage.
If the transaction is ultimately approved, the data will enter the final settlement stage. However, there is little public disclosure about the data's journey from the finalized dataset to Google's systems. It is currently unknown how the data is transmitted, whether further cleansing will occur, and in what form it will ultimately enter product development or model training.
At this point, the program truly completes the transformation of data from bankruptcy assets to AI assets.
Buyer
The most clear-cut buyers of this type of data at present are model companies like Google and training data service providers like Mercor.
Google's official statement regarding this Spirit transaction is that the data will be used for product improvement and AI development. For Google, the value of this dataset lies in its documentation of a large enterprise's real operational processes, such as scheduling, collaboration, internal communication, project advancement, issue resolution, and management decisions.
These details are hard to obtain from public web pages. Particularly for corporate agents, beyond just the final outcome, what is more important is how tasks are progressed and completed within a real organization. The internal records left by Spirit happen to contain a significant amount of such process data.
Another bidder, Mercor, provides further insight into where this business is heading.
Founded in 2023, Mercor's three founders—Brendan Foody, Adarsh Hiremath, and Surya Midha—have been debate team partners since high school. The company initially focused on AI recruitment, using models to help companies screen and interview candidates. By September 2024, Mercor had evaluated around 300,000 job seekers, reaching a valuation of $250 million.
Subsequently, the company's focus shifted towards AI training data. In November 2025, when the three founders were only 22 years old, as the company's valuation rose, they became some of the youngest self-made billionaires globally. In the first half of 2026, Mercor's revenue exceeded $614 million, with about 90% coming from top AI labs like OpenAI. By July, the company was seeking a valuation of around $20 billion and had acquired Deeptune, a company specialized in building training environments for AI agents.
Mercor primarily acquires training data through two main paths.
One is by directly purchasing internal company records. They have made offers to acquired or closed-down startups to buy employee chat logs and emails, with individual companies receiving offers of up to around $300,000.
Another one is hiring people with real work experience. In October 2025, TechCrunch reported that an AI lab recruited former employees through Mercor, allowing them to convert their work experience into training tasks and feedback at an hourly rate of about $200. The CEO of Mercor stated that the company pays out over $1.5 million daily to individuals participating in AI training.
On one side, acquiring the company's retained work records, and on the other, acquiring the work experience held by employees. Mercor's core assets are essentially work processes from the real world.
This also explains why it appeared at Spirit's auction. For Mercor, 6 billion pieces of internal communication were another, larger-scale source of training data.
This type of business also comes with risks. In 2025, Scale AI sued Mercor, alleging that former employees took trade secrets; subsequently, the company experienced training-related information leaks and partnership suspensions. Data is both Mercor's business and its most critical asset to defend.
Lastly, there is a numerical contrast. Spirit operated under this name for 34 years, while Mercor, bidding on its internal data, had three founders who were only 22 years old at the time.
Broken Bench
The transaction for Spirit is not yet finalized, the court has not signed off, and Google has not acquired that set of data. There are still union objections, hearings, and many procedures to go through.
But this auction has already made something that was seldom discussed before more concrete.
In the past, when a company closed, many things did indeed disappear. Financial statements remained, trademarks remained, patents remained. However, a company's day-to-day experience usually did not. How a department conducts meetings, how a manager makes decisions, how dozens of people coordinate after a delay, why a process eventually changed to its current state – these things are rarely formally documented.
As a company dissolves, people leave, email addresses deactivate, chat groups close – they also disappear.
So, a company has always been a peculiar organization.
It can exist for decades, accumulating the experiences of tens of thousands of people, but what can truly be inherited is often only a very thin slice of it. The next company will have to hire again, make mistakes again, and learn many things all over again.
The potential AI is changing is exactly this. If emails, meetings, tickets, code changes, and internal discussions can truly be organized into training data, then a company's past experiences that could only be passed down by talent have, for the first time, found another way to be preserved.
Where this will ultimately lead is still difficult to determine. Perhaps in the future, companies will proactively save this data, perhaps employee contracts will be rewritten, perhaps bankruptcy laws will add new restrictions, perhaps companies will separately value even internal work records during financing and M&A.
Spirit has raised this issue early.
The word bankruptcy has a widely circulated etymology. Medieval Italian merchants conducted business sitting behind a bench. When their debts exceeded their assets, the bench was publicly destroyed. banca rotta, broken bench.
For centuries, when the bench broke, it was all over.
Spirit's bench has already broken. The 600 million messages it left behind are being repriced.
One message at one and a half pennies.
Original Article Link

Can someone who once helped Soros break the Bank of England now use the same tactics to defend the U.S. Treasury market?
Since the beginning of this year, U.S. Treasury Secretary Scott Bessent has made consecutive moves, using a series of unexpected market operations to stake his reputation on suppressing American borrowing costs. According to Bloomberg, he has become the "most proactive Treasury Secretary in decades in intervening in the financial markets".
Following the U.S.-Japan joint intervention in the yen, Bessent's latest move is to expand U.S. bond repurchases. The Treasury Department announced that it would "at least double" the scale of repurchases of 10 to 30-year Treasury bonds—and this repurchase plan was only announced just two weeks ago. On the day the news was released, the yield on the 10-year Treasury bond fell by roughly 6 basis points, the 30-year yield fell by nearly 9 basis points, and the U.S. dollar index also fell to a three-month low.
The market's reaction confirmed Bessent's judgment: he has publicly stated, "My job is to be the nation's top bond salesman, and the U.S. Treasury yield is the barometer of success."
From Shorting the Pound to Guardian of the Bond Market
To understand Bessent's strategy, we must go back to 1992.
That year, in his early twenties, Bessent worked at the Soros Fund and participated in building a short position on the pound. On "Black Wednesday," the pound was forced to exit the European Exchange Rate Mechanism, and Soros netted over $1 billion. According to media reports, a former advisor described Bessent at the time as someone who "could see market vulnerabilities that others couldn't."
Afterward, he returned to Soros as Chief Investment Officer, leading a $1 billion yen short in 2013, once again reaping substantial returns. In 2015, he founded Key Square Capital Management with $4.5 billion, betting successfully on Brexit and Trump's two election victories.
This "find the crack, then push with the trend" hunter logic has run through his entire hedge fund career.
And now, he is using the same intuition to do the exact opposite—defend a market under pressure.
This Year's Intervention Map: From Yen to U.S. Bonds
Bessent's moves this year have already formed a clear logical chain.
Step One, Yen Intervention. On July 31, the U.S. Treasury, in conjunction with Japanese authorities, intervened in the market to buy yen, marking the first direct yen intervention by the United States in nearly thirty years. According to data from the Peterson Institute for International Economics (PIIE), Japan used approximately $87 billion of its foreign exchange reserves to purchase yen in the last two days of July, with the U.S. Treasury "joining in the final stages, providing a relatively limited amount of funding but sending an important political signal of support." It is worth noting that the Treasury sold euros instead of dollars, and the European authorities were not informed in advance.
Behind this move lies an undercurrent: Japan holds around $1.1 trillion in U.S. Treasuries, making it the largest foreign holder. If Japan had intervened alone to finance the intervention, it might have been forced to sell U.S. Treasuries, further driving up long-term yields. Washington's participation allowed Japan to sell fewer U.S. Treasuries, indirectly preserving the yield curve that Secretary Benson cares most about.
Step Two, Bond-issuance Contraction Signal. Earlier this month, the Treasury hinted at a potential reduction in the issuance size of long-term bonds, signaling an expectation of supply tightening to the market.
Step Three, Enhanced Repurchase. It was announced this week that the size of long-dated bond repurchases would at least double, providing direct support to prices from the demand side.
Bloomberg cited Brad Golding, portfolio manager at Christofferson Robb & Co., as saying this is akin to "an old-school ‘clean screens’ maneuver" — a hedge fund technique involving simultaneous orders to multiple major dealers to trigger significant market volatility.
Mark Sobel, a former U.S. Treasury official now with the OMFIF think tank, told Bloomberg, "He is absolutely a radical, which harkens back to his hedge fund background." "Both he and this administration are clearly concerned about the rise in long-term yields."
Breaking "Rules and Predictability"
Benson's actions are in direct conflict with the Treasury's traditional principles.
The U.S. Treasury has long adhered to a principle of "rules-based, predictable" debt management, avoiding surprises for the market. Benson himself publicly endorsed this principle at a Treasury market conference in November of last year.
However, his recent actions have deviated from this commitment.
Gregory Faranello, Head of U.S. Rate Trading and Strategy at AmeriVet Securities, told Bloomberg, "This violates the 'rules-based, predictable' principle — but that’s the world we're in." "The signal is clear: Prevent the rise in yields."
Ironically, Bernett's predecessor Yellen adjusted the debt issuance structure in 2023 to suppress yields, with Bernett being one of the critics at the time, accusing the move of being politically motivated. Former chief economist for Trump, Stephen Miran, also co-authored a paper in 2024 criticizing the "Aggressive Treasury Issuance" (ATI).
According to Bloomberg, Miran and Nouriel Roubini wrote in the paper: "Once a party starts using ATI to stimulate the economy during the election season, all future administrations may follow suit."
Question: Can Intervention Solve Structural Issues?
The market had a short-term reaction to Bernett's actions, but economists have more fundamental questions.
As of the first ten months of the 2026 fiscal year, federal net interest payments have reached $963.0 billion, approximately $3.18 billion per day, a 14% year-on-year increase. The 10-year Treasury yield stands at 4.72%, while the 30-year yield is at 5.31%—a significant amount of old debt previously issued at under 2% is now rolling over at higher rates. The deficit for the 2026 fiscal year to date is $1.8 trillion, expanding by 5% from the previous year, with spending on Social Security, Medicare, national defense, and interest on debt all rising, while the Republican Party is still discussing further tax cuts.
Robin Brooks, Senior Fellow at the Brookings Institution, bluntly told Bloomberg: "This is not addressing the root issues—reducing debt, shrinking the fiscal deficit, but rather trying to manipulate the yield curve."
John Velis, BNY Macro Strategist, also stated: "Given current spending policies and wars, alleviating pressure at the long end will be very difficult."
The effectiveness of yen intervention is also questionable. After hitting a high of 163.98 against the dollar on July 23, the USD/JPY pair fell to 159.43 by August 17, but according to CNBC, intervention did not prevent the continuous weakening of the yen. Maurice Obstfeld of the PIIE directly stated that intervention had minimal effect, saying, "Foreign exchange intervention is not a free lunch, not even a free cake."
Guy Miller, Chief Strategist at Zurich Insurance, told Bloomberg: "This approach can only work for a while. When the Treasury clearly states its intention to continue intervening, it can indeed have a quite strong effect. But ultimately, if the profligate fiscal policy is not addressed, this is unsustainable."
Onepoint Bfg's Chief Investment Officer Peter Boockvar was more direct: "He is waging war on two giant markets at the same time - the U.S. bond market and the foreign exchange market. This is an extremely difficult battle."
The Bet on Reputation
Bridgewater's logic, as articulated in his own words, is quite clear. Referring to the Trump administration's stakes in technology and resource companies last month, he said: "What we want to do is create a market signal." Speaking on Fox Business, he said: "Essentially, it's telling investors, okay, where the puck is going to be, skate there quickly."
The problem is that in 1992, shorting the pound was about finding a systemic weakness and taking advantage of it. Now, he is facing structural pressures driven by fiscal deficits, inflation expectations, and Fed policy - things that cannot be fundamentally changed through repo operations or exchange rate interventions.
According to Bloomberg, Mark Sobel, who served in the Treasury Department for nearly 40 years, believes Bridgewater is the most radical Treasury Secretary since the early 21st century. However, he also characterized yen intervention as an unwise move, arguing that it avoided the fiscal consolidation that the U.S. truly needs.
Original Article Link

En bref
· Le Trésor américain va étendre son rachat de titres nominaux à coupon dans les secteurs 10-20 ans et 20-30 ans, en augmentant la limite par opération de 20 milliards de dollars à au moins 40 milliards de dollars.
· Cette opération vise à améliorer temporairement la liquidité des titres à plus longue échéance, en réduisant les primes de terme marginales, mais elle ne fait pas partie du programme d’assouplissement quantitatif de la Fed.
· Instruments liés : TLT, QQQ, Or, BTC et valeurs de croissance sensibles aux rendements à long terme.
Le 19 août, le Trésor américain a annoncé l’extension de ses opérations de soutien à la liquidité pour les obligations à long terme, portant la limite par opération pour les titres nominaux à coupon dans les secteurs 10-20 ans et 20-30 ans de 20 milliards de dollars à au moins 40 milliards de dollars.

Google lie des options d’achat d’actions à l’approvisionnement en puces, aux résultats cliniques et à une pression à la baisse tirant parti des ventes à découvert, transformant les attentes à long terme en cours. Les dernières évolutions à Homsud indiquent que la question de savoir si l’offre peut atteindre de manière stable le marché reste une contrainte difficile.
1｜Google lie ses options d’achat d’actions à Marvell, avec des achats assortis d’une incitation au capital intégrée dans le même contrat
Marvell a révélé qu’elle avait émis des bons de souscription à Google pour acquérir jusqu’à 58,97 millions d’actions ordinaires, au prix d’exercice de 206,58 $ par action, exerçables au plus tard en août 2033. Si tous les bons sont exercés, environ 12,2 milliards de dollars seraient calculés au prix d’exercice. Les bons sont liés à l’objectif de chiffre d’affaires de Google pour l’achat de puces personnalisées.

Le bitcoin est passé de 65 000 $ à 70 000 $ en une nuit, tandis que l’ether a connu une hausse d’environ 20 %, se rapprochant d’un sommet historique. Le battage est réel.
Le marché des cryptomonnaies était en plein essor en l’espace d’une nuit, probablement sous l’influence d’évolutions aux États-Unis.
Le 20 juillet, le président Trump a réuni une table ronde et prononcé un discours public à la Maison-Blanche, à Washington, D.C., en présence de dirigeants des entreprises américaines de la finance, des cryptomonnaies et de la technologie, ainsi que de responsables d’agences de régulation.

In mid-August, major funds successively released their 13F quarterly reports, disclosing the static holdings of the funds as of June 30.
· Odaily Note: The so-called 13F is a quarterly disclosure document that the U.S. Securities and Exchange Commission (SEC) mandates for funds with assets under management exceeding $100 million. The SEC requires that funds meeting the disclosure requirements must submit this document within 45 days after the end of each calendar quarter. Funds are required to focus on disclosing their holdings of U.S. listed stocks, call/put options, convertible bonds, and specific ETF positions held at the end of the previous quarter in the document.
Although there is a certain lag in the disclosure timing of 13F, making it unsuitable for simple copying, as a direct window into the top funds' layouts and movements, 13F still holds significant importance for exploring the ideas of on-chain high-level players. The consensus and discrepancies among different funds may well hide clues to future market trends.
In the following sections, we will analyze the 13F reports of seven top funds, including Berkshire Hathaway (Buffett, Abel), Duquesne Family Office (Stanley Druckenmiller), H&H International Investment (Duan Yongping), Himalaya Capital (Li Lu), ARK Investment (Cathie Wood), Oriental Harbor Overseas Fund (Dan Bin), Situational Awareness LP (Leopold Aschenbrenner), focusing on core holdings and major trends, hoping to add value to your investment strategy.
Berkshire Hathaway (Buffett, Abel)
Report Summary
As of June 30, Berkshire Hathaway disclosed a total of 29 holdings in its 13F report (13F only covers U.S. listed securities subject to disclosure and does not equate to the fund's total asset size), including 1 new position, 7 increases, 6 reductions, and 1 exit, with a nominal holding market value of approximately $299.3 billion.
It is worth noting that this is Berkshire Hathaway's second 13F filing under CEO Greg Abel's formal leadership. This filing marks Berkshire's first significant net buying quarter after ending the previous streak of 14 consecutive quarters of net selling, with net stock purchases amounting to nearly $20 billion.
Top Holdings
Berkshire's portfolio remains highly concentrated, with the top ten holdings accounting for about 88.47% of the total:
· Apple (AAPL): Approximately $65.95 billion, holding the top position with a 22.0% allocation;
· American Express (AXP): Approximately $51.28 billion, representing 17.1% of the portfolio;
· Alphabet (Google, GOOGL + GOOG): Approximately $37.76 billion, including around $28.16 billion of GOOGL (Class A common stock with voting rights) and $9.61 billion of GOOG (Class C common stock with no voting rights), making it the third-largest holding;
· Coca-Cola (KO): Approximately $32.51 billion, accounting for 10.9% of the portfolio;
· Bank of America (BAC): Approximately $27.54 billion, with a 9.2% allocation;
· Chevron (CVX), Occidental Petroleum (OXY), Chubb (CB), Moody's (MCO), Kraft Heinz (KHC) round out the sixth to tenth positions.
Structurally, Berkshire Hathaway has not significantly altered its core positions in the consumer, financial, and energy sectors. However, the addition of Alphabet has notably increased the weighting of the technology sector in the portfolio.
Quarterly Changes
In the second quarter, Berkshire's most significant move was undeniably a major bet on Alphabet—increasing its holdings of Alphabet's Class A shares (GOOGL) by approximately 24.54 million shares, a 45.2% increase, and also boosting its holdings of Class C shares (GOOG) by about 23.60 million shares, a staggering 658.3% surge. The combined increase of about 48.10 million shares has propelled Google into the fund's top three holdings. Apart from Alphabet, Berkshire notably increased its stakes in Delta Air Lines (DAL), Lennar Corporation (LEN), Macy's (M), among others.
On the other hand, reductions were mainly concentrated in the financial, consumer, and cyclical sectors. Bank of America (BAC) reduced its holdings by about 30.23 million shares, a 5.9% decrease, being reduced for two consecutive quarters; Capital One (COF) reduced its holdings by about 58%, Kroger (KR) by about 22%, Ally Financial (ALLY), DaVita (DVA), and Nucor (NUE) also saw declines.
Summary Analysis
Under the Abel era, Berkshire Hathaway is undergoing a subtle style shift, with the clearest signal in the second quarter being a tilt from traditional finance and consumer stocks towards tech growth. The heavy position in Google not only further breaks Buffett's stereotype of "avoiding tech stocks," but also reflects the new management's deep recognition of Google's moat in AI and search—essentially still a typical "value confirmation" rather than trend chasing.
However, Berkshire's core holdings remain unchanged: stake in "anchor" companies such as Apple, American Express, Coca-Cola are basically unchanged, and the energy sector's Chevron and Exxon Mobil have not been sold off (the decrease in percentage is mainly due to dilution from new holdings).
Overall, this quarter Berkshire Hathaway is showing a rebalancing logic of "heavily increasing positions in tech leaders, structurally reducing financial and consumer stocks, holding onto core energy positions, and tentatively positioning in real estate and aviation." While quietly pushing the portfolio towards a digital future while maintaining extremely high concentration, Abel's Buffett-esque nature of "long-term commitment" and "big bets" is still clearly visible.
Duquesne Family Office (Stanley Druckenmiller)
Although not as famous as Buffett and Berkshire, Stanley Druckenmiller may be the most noteworthy fund manager on Wall Street today.
Who is Druckenmiller? He is a legendary American macro hedge fund manager who served as the Chief Investment Officer of Soros Fund Management from 1988 to 2000, becoming Soros's most successful trader... And today, his most critical role is that both Treasury Secretary Scott Bennett and Fed Chair Kevin Wash have been his disciples (having either studied under him or had a long-term working relationship with him).
Therefore, compared to Berkshire's more long-term focused 13F, Duquesne Family Office's quarterly report is more like a macro trading map—especially in a time when AI, interest rates, and the U.S. economic outlook are all rapidly changing.
Report Summary
As of June 30, the Duquesne Family Office disclosed a total of 95 positions in its 13F filing, with 48 new stocks added during the quarter, 16 increased positions, 11 decreased positions, and 23 stocks completely sold off. The nominal value of the holdings is approximately $52.1 billion, a significant increase from the previous quarter's $33.8 billion.
Core Holdings
The 13F filing reveals that the Duquesne Family Office's top ten holdings account for approximately 45.8% of the portfolio:
· Genetic testing company Natera (NTRA) remains the largest stock holding of the Duquesne Family Office, with approximately 3.19 million shares at the end of the second quarter, valued at around $865 million, representing about 16.6% of the portfolio;
· The second and third largest holdings have shifted to TSMC (TSM) and STMicroelectronics (STM), accounting for approximately 5.4% and 4.4% of the portfolio, respectively;
· Additionally, biopharmaceutical company Insmed (INSM, held in both stock and options), Argentinean national oil company YPF Sociedad Anónima (REPYY), and Amazon (AMZN) are among the top ten holdings.
Quarterly Changes
In the second quarter, the most notable change by the Duquesne Family Office was a shift within the AI industry supply chain.
The Duquesne Family Office completely sold off Micron (MU), Broadcom (AVGO), and Intel (INTC), with Broadcom having been a new addition in the first quarter. However, this does not mean that Druckenmiller has abandoned semiconductors; instead, he continued to increase positions in TSMC (TSM) and STMicroelectronics (STM), and initiated a new position in AMD. Furthermore, the fund has begun to expand its reach into the periphery of AI infrastructure, buying into Hut 8 (HUT), Bitdeer (BTDR), Riot Blockchain (RIOT), and other Bitcoin mining companies that are transitioning towards data center operations.
It appears that Druckenmiller has chosen to cash out some of the previously highly appreciated positions, while reallocating the funds to assets he believes offer a superior risk-return profile.
Furthermore, in the second quarter, the Duquesne Family Office also reestablished its position in Alphabet (GOOGL), which it had just exited in the first quarter, and significantly increased its holdings in Amazon (AMZN) that were reduced in the first quarter—this move may be a bet that the large tech companies that had previously borne significant AI capital expenditures will gradually transition from being the "payers for AI infrastructure" to being the "beneficiaries of AI commercialization."
In addition to the aforementioned AI-related portfolio changes, the Duquesne Family Office, like Berkshire Hathaway, also expressed bullish views on the aviation sector, initiating a new position in Delta Air Lines (DAL) and increasing its stake in United Airlines (UAL); notably, in the second quarter, the Duquesne Family Office also initiated a new position in Baidu ADR (BIDU) of approximately 88,000 shares, marking its first holding of U.S.-listed Chinese stocks in two and a half years since liquidating Alibaba at the end of 2023.
Summary Analysis
If we were to condense this 13F into a single sentence, Druckenmiller has not departed from AI but is instead repositioning for the next phase of AI winners. At least based on the end of the second quarter holdings, Druckenmiller has still maintained a significant AI exposure, but has reallocated from parts of the highly crowded hardware sector to areas like cloud platforms, data centers, and AI applications.
Simultaneously, Baidu's reappearance is another signal worth noting. While the 88,000-share position is not large, the tentative reentry into Chinese tech assets after long avoidance at least indicates a renewed interest.
It is worth reiterating that due to Druckenmiller's relatively high-frequency reshuffling, the Duquesne Family Office's 13F report is more challenging to simply "copy," making it more suitable for observing directional shifts.
H&H International Investment (Neil Shen)
Turning our attention to H&H International Investment managed by the Chinese legendary investor Neil Shen. From this 13F report, it can be seen that Shen's holdings remain highly concentrated, but there are signs in the second quarter of a rotation from some highly valued tech stocks towards Chinese internet companies.
Report Summary
As of June 30, H&H International Investment disclosed a total of 18 positions in its 13F filing, with a nominal value of approximately $19.101 billion (equivalent to about RMB 130 billion), down from around $20 billion at the end of the first quarter.
It is worth noting that 13F filings only disclose long positions in US stocks, so assets held by Duan Yongping such as Tencent, Pinduoduo, Moutai, and a large number of put-option premium-selling operations are not included in this table.
Top Holdings
H&H International Investment's portfolio remains highly concentrated, with the top five largest positions accounting for over 88% of the total.
· Apple (AAPL): Approximately $7.841 billion, representing 41.05% of the portfolio, maintaining its position as the largest holding;
· Berkshire Hathaway (BRK.B): Approximately $4.618 billion, representing 24.18%, forming the "ballast" together with Apple;
· Pinduoduo (PDD): Approximately $1.91 billion, accounting for 9.99%, rising to the third-largest position, signaling a significant increase in the weight of Chinese internet companies in Duan Yongping's system;
· Tesla (TSLA): Approximately $1.42 billion, accounting for 7.44%;
· NVIDIA (NVDA): Approximately $1.26 billion, accounting for 6.58%.
Quarterly Changes
In the second quarter, the most significant move by H&H International Investment was a substantial increase in its Pinduoduo (PDD) position—adding 5.2738 million shares, a 26.71% increase, propelling it to the third-largest position; meanwhile, there was a slight increase in Berkshire Hathaway Class B shares (BRK-B), which can be interpreted as a long-term endorsement of the value investing framework. Additionally, Alibaba (BABA), which was completely sold off by Duan Yongping in the first quarter, saw a repurchase of 301.4 thousand shares (approximately $28.93 million) in the second quarter.
As for the reduction of holdings, AI and tech leaders were the main targets for profit-taking by H&H International Investment in the second quarter. NVIDIA (NVDA) saw a divestment of 7.5631 million shares, a significant 54.63% reduction; Google (GOOG) sold over 1.7 million shares, a 46.88% decrease; Microsoft reduced its holdings by 25.78%; Apple also trimmed its position by 1.8469 million shares—marking the second consecutive quarter of selling (after selling 3.4129 million shares in the first quarter); additionally, TSMC and CrowdStrike (CRWD) were completely liquidated.
It is worth mentioning that Duan Yongping is not simply bearish on the tech stocks he sold. In late July, he publicly stated, "I will add more Google," while also expressing his intention to continue seeking opportunities to buy NVIDIA at lower prices through selling Put options. In other words, the reduction in positions is more a reflection of price and margin of safety considerations rather than a complete rejection of the companies themselves.
Summary Analysis
If we were to summarize this 13F filing in one sentence—sell a bit on the way up, buy a bit on the dip, and hold onto good companies.
Overall, the second-quarter repositioning strategy of H&H International Investment appears relatively clear—slight adjustments to core holdings, increased exposure to Chinese internet companies, and profit-taking on high-growth AI assets. Apple and Berkshire Hathaway remain core holdings, but after continuous reductions, Apple's position has decreased for two consecutive quarters; meanwhile, Pinduoduo has become a new top three holding, and Alibaba has reentered the portfolio; although NVIDIA, Google, and others were reduced, it was not a bearish move but rather a realization of gains, with Duan Yongping expressing a desire to reacquire them at lower prices in the future.
This indicates that Duan Yongping's actions in the second quarter were not a full pivot towards a new theme but rather a reassessment of margin of safety amidst valuation changes, with a partial realization of gains on high-growth AI and tech leaders and a re-up in exposure to Chinese internet companies, which he is more familiar with after a long period of adjustment.
Himalaya Capital (Li Lu)
Similar to Duan Yongping, another legendary Chinese investor, Li Lu, who heads Himalaya Capital, also significantly increased his holdings in Pinduoduo (PDD) in the second quarter.
Report Summary
As of June 30, 2026, Himalaya Capital disclosed a total of 8 positions in its 13F filing, increasing holdings in 2 companies and liquidating 6, significantly reducing the number of holdings from 14 in the first quarter. However, the nominal value of total holdings increased from $3.201 billion in the first quarter to $3.703 billion, further shrinking and concentrating its positions, with the top five holdings now accounting for a high 94.77% of the portfolio.
Core Holdings
The 13F filing shows that Himalaya Capital's current 8 positions reflect a "Google Anchoring, Pinduoduo Surging, Financial Foundation" pattern.
· Google (GOOGL + GOOG): Approximately $1.775 billion, accounting for 47.64% of the portfolio, held long-term since 2020, serving as the unwavering "anchor" in Li Lu's portfolio;
· Pinduoduo (PDD): Approximately $0.821 billion, accounting for 22.17%, surged to the second largest position after a significant increase in holdings;
· Berkshire Hathaway (BRK.B): Approximately $0.555 billion, accounting for 14.98%, continues to accumulate holdings;
· East West Bancorp (EWBC): Approximately $0.358 billion, accounting for 9.68%, position unchanged.
· In addition, Crocs (CROX, 2.90%) and Tencent Music (TME, 1.50%) remained unchanged, while Apple (AAPL) is left with a mere 0.88% as a minimal observation position.
Quarterly Changes
In the second quarter, Himalaya Capital only added two new positions, liquidating the other six existing positions in one go, a resolute decision rarely seen among top funds.
Pinduoduo (PDD) was the most significant addition by Himalaya Capital in the second quarter — adding 6.1531 million shares, a staggering 133.53% increase, with holdings soaring from 4.608 million shares to 10.7611 million shares, ending the quarter with a market value of approximately $0.821 billion, becoming the second largest position. Furthermore, Berkshire Hathaway (BRK.B) saw an increase of 23.46%, effectively entrusting more funds to Buffett and Abel's management.
Meanwhile, Himalaya Capital completely exited 6 positions in one go—Bank of America (BAC), Occidental Petroleum (OXY), S&P Global (SPGI), Moody's (MCO), MSCI (MSCI), H&R Block (HRB). This signifies Li Lu's decisive withdrawal from traditional finance, index data services, and the oil and gas sector. Particularly noteworthy is Bank of America, which was once one of its core anchors, aligning with Berkshire Hathaway's "bank reduction" this quarter, but Li Lu chose a complete exit instead of a slight reduction.
Summary Analysis
If the second quarter of Li Lu's operations were to be condensed into a sentence, it might be — a significant reduction in positions outside the Circle of Competence, further concentrating limited funds on a few truly understandable companies.
On one hand, there is the maintaining of a heavy position in Google (GOOGL, GOOG) and at the same time increasing the position in Pinduoduo (PDD), with the former representing Li Lu's long-term view on the U.S. tech leader and the latter on Chinese internet assets; meanwhile, the liquidation of holdings in financial, energy, and index service-related assets also implies that Li Lu is further shrinking his investment portfolio, with the current concentration of positions being very close to his classic value investment style.
Of course, similar to the situation with Duan Yongping, the 13F will only disclose securities in the U.S. market that meet the criteria, so this document cannot represent Li Lu's entire investment portfolio; assets in the Hong Kong stock market such as BYD, Postal Savings Bank of China, CRRC Corporation, etc., will not appear in this 13F.
ARK Investment (Cathie Wood)
Report Summary
As of June 30th, ARK Investment managed by the "female Buffett," "WoodSis" Cathie Wood, disclosed a total of 191 positions in the 13F, including 15 new holdings, 78 additions, 96 reductions, and 6 liquidations, with a nominal holding market value of approximately $15.4 billion, a significant increase of over $2 billion compared to the previous quarter.
In the second quarter, ARK Investment's repositioning was more aggressive, showing Wood's typical high turnover style.
Core Holdings
The 13F report shows that ARK Investment's top ten core holdings account for a total of 39.45% of the portfolio, with a relatively moderate level of concentration, displaying a relatively high risk preference in the core asset portfolio.
· Tesla (TSLA): About $11.61 billion, holding percentage 7.5%, still the largest core position but has been reduced for the third consecutive quarter;
· AMD (AMD): About $8.2 billion, holding percentage 5.3%, a core holding in the AI computing power sector;
· SpaceX (SPCX): About $7.65 billion, holding percentage 5%, a new entry this quarter directly into the core position;
· Tempus AI (TEM): Approximately $5.8 billion, position weight 3.8%, focused on precision medicine + AI diagnostic benchmark;
· Robinhood (HOOD): Approximately $5.25 billion, position weight 3.4%, fintech and retail investor ecosystem.
Quarterly Changes
In the second quarter, ARK Investment's most significant move was the "rush funding" of SpaceX (SPCX). On the first day of SpaceX's listing on June 12, ARK's various ETFs collectively bought approximately 3.29 million shares, which had increased to 4.478 million shares by the end of the quarter, with SPCX's weight in the fund's portfolio reaching nearly 7% at one point.
Aside from SPCX, in the second quarter, ARK Investment's investment landscape also showed a clear trend of deepening its industry chain. Cerebras Systems (CBRS) was newly included, representing its bet on AI computing architecture; Google (GOOG) saw an increase, strengthening the exposure to platform AI; the purchase of X-Energy (XE) signaled the fund's early layout of nuclear energy as a baseload power source in the AI era; and the additional position in Eli Lilly (LLY) reinforced the focus on life sciences...
Regarding reductions, the most noteworthy action was the continuous reduction in Tesla (TSLA) for three consecutive quarters, indicating ARK's assessment of Tesla's diminished relative position in the AI narrative.
Summary Analysis
Overall, ARK Investment's actions in the second quarter demonstrated a restructuring logic of "actively embracing SpaceX's listing, expanding AI and energy boundaries, reducing traditional core holdings, and maintaining high turnover innovation hunting."
ARK Investment still maintains a relatively high risk appetite among the top funds, focusing more on finding technologies that can transform industry structure—from AI computing power to aerospace, nuclear energy, and life sciences, "disruptive innovation" remains the fund's most enthusiastic theme.
Oriental Harbor Overseas Fund (Bin Du)
Compared to previous funds, the latest 13F report of the Oriental Harbor Investment Master Fund managed by Chinese renowned investor Dan Bin can be considered quite "aggressive" — instead of simply adjusting a few stocks, it almost completely revamped its AI holdings.
Report Summary
As of June 30, the Oriental Harbor Investment Master Fund disclosed a total of 13 positions in its 13F report, including initiating 7 new positions, increasing 1 position, decreasing 5 positions, and fully exiting 6 positions, with a total nominal value of approximately $1.65 billion, representing a growth of about 45.6% from around $1.133 billion in the previous quarter.
Core Holdings
The 13F report indicates that by the end of the second quarter, the top five core holdings of the Oriental Harbor Investment Master Fund accounted for approximately 73% of the total portfolio, further increasing the portfolio concentration, with AI hardware and semiconductor industry-related targets comprising over 70% of the portfolio weight, signaling a significant shift in the portfolio style.
· Google (GOOG): Approximately $371 million, with a roughly 23% allocation, maintaining its position as the largest core holding;
· Intel (INTC): Approximately $258 million, with a 16% allocation, making a new entry and landing as the second-largest position in the second quarter;
· NVIDIA (NVDA): Approximately $217 million, with a 13% allocation, slipping from first to third in holdings percentage;
· SanDisk (SNDK): Approximately $176 million, with an 11% allocation, entering as the fourth-largest position;
· Micron (MU): Approximately $170 million, with an approximately 10% allocation, witnessing a doubled-up increase;
· Positioned sixth to tenth are AMD (8.9%), MRVL (7.9%), TSM (4%), ARM (3.2%), and AVGO (1.5%).
Except for Google, nine out of the top ten core holdings belong to AI computing infrastructure or the semiconductor industry chain, as the "shovel-selling" hardware logic has completely replaced the previous focus on core software platform holdings.
Quarterly Changes
This quarter, Oriental Harbor Overseas Fund launched a "saturation attack" on the AI hardware industry chain, newly acquiring 7 related targets in one go—Intel (INTC), SanDisk (SNDK), AMD, Marvell Technology (MRVL), ARM, Broadcom (AVGO), Lumentum (LITE), all focusing on computing chips, storage, optical communication, and semiconductor upstream. Among them, Intel directly became the second largest position, SanDisk and AMD entered the top six holdings, with a clear intention to increase exposure to the storage sector.
While fully embracing hardware, Ben has also "cut ties" with existing holdings. Nvidia (NVDA), TSMC (TSM), Amazon (AMZN) have all been reduced; although Google remains the top holding, compared to the first quarter, 155,200 shares have been sold, and the 2x long Google ETF has also been completely liquidated, significantly reducing its overall weight in the portfolio.
In addition, the Oriental Harbor Overseas Fund has also completely sold off Apple (AAPL) and Tesla (TSLA), the two leading consumer electronics companies, in the second quarter, perhaps indicating that Ben is not optimistic about the short-term recovery of the consumer end. At the same time, the fund has also liquidated its position in stablecoin issuer Circle (CRCL).
Summary Analysis
Overall, the most crucial information in this 13F report from Oriental Harbor Overseas Fund is that Ben is still betting that AI CapEx will continue to flow to the hardware side. While the market is still debating whether there is an AI bubble, Ben believes that the hardware segment, the "selling shovels" part, is still the most certain direction.
This is also in line with Ben's recent public statements—he still believes that AI is the super trend of the next decade and has stated that the market still does not fully grasp AI's long-term potential. During the storage market's major correction at the end of July, Ben even made a high-profile statement that "you must dare to buy when there is a big drop and have already used up all remaining ammunition."
Situational Awareness LP (Leopold Aschenbrenner)
For the story of the "AI Stock God" Leopold Aschenbrenner and his fund Situational Awareness LP, we have provided a detailed interpretation in the article "Today, the world finally understands why the 'AI Stock God' has fallen."
Due to its outstanding performance in the past few quarters, the 13F of Situational Awareness LP was highly anticipated by the market. However, it is regrettable that the fund experienced a dark moment at the end of July—due to a significant pullback in AI-related stocks combined with high leverage, the fund suffered major losses, was forced to liquidate large-scale public market positions, and has already packaged most of its stock portfolio at a discount for sale to Citadel, managed by Ken Griffin.
Breaking down Situational Awareness LP's 13F filing reveals an even more regrettable story. The fund had built a bearish options position on chip and storage industry leaders with a nominal value of over $8 billion by the end of the first quarter, intending to effectively hedge against the current market downturn. However, due to an early shift to an overall bullish outlook, Leopold Aschenbrenner himself dismantled this insurance wall, ultimately leading to the fund's once glorious performance being discounted and taken by Ken Griffin.
The AI Narrative Remains Unchanged; What's Changing is Where the Money Flows
Putting together the 13Fs of these seven funds seems to point to a relatively clear consensus — AI remains at the center of top investors' vision, with funds significantly repricing around the AI industry chain. The growth of AI seems undisputed, but the real question is, in the next phase, who can truly turn investment into profit?
Berkshire Hathaway has begun a major stake in Google, Druckenmiller is rotating within the semiconductor sector and recommitting to cloud computing and AI infrastructure; Bin, on the other hand, is more aggressive, significantly shifting the overseas portfolio towards chips, storage, and optical communication; ARK continues to explore next-generation growth assets in AI, SpaceX, energy, and life sciences. Meanwhile, Duan Yongping and Li Lu are not simply chasing the AI trend but, after reducing holdings in tech stocks with significant gains, are reallocating funds to assets they are more familiar with and consider to have a higher margin of safety.
Mere observation of each fund's 13F makes it difficult to deduce a clear "next big stock," but the capital movements of different funds are themselves a rotation game regarding AI investment logic — from GPUs, chips, storage, networking, optical communication, data centers, power, cloud computing, commercialization... From selling shovels to providing computing power, and finally sharing in the commercialization dividend of AI, funds are continuously seeking the next profit realization point on the AI value chain.
It is worth emphasizing once again that all these positions are as of June 30, 2026. The 13F itself has a disclosure lag of up to 45 days, especially pertinent for investors like Druckenmiller and Bin who reposition quickly. Thus, rather than treating it as a "homework copying list," it is better to view it as a snapshot of how different funds are interpreting the market in the next phase.
Original Article Link

En bref
· Une partie de la capacité quotidienne de transit par le détroit d’Ormuz est tombée à des niveaux extrêmement bas, mais le Brent n’a pas augmenté de façon constante au-dessus de 100 $.
· Le marché parie temporairement que les libérations de stocks, le transbordement, les exportations alternatives et les détours des acheteurs peuvent absorber une partie de l’impact.
· Actifs liés : pétrole brut Brent/WTI, ETF sur l’énergie, sociétés de pétroliers, raffineries indépendantes chinoises, chaîne d’approvisionnement en diesel, or.
Depuis août, le suivi du transport maritime et des informations médiatiques ont montré qu’une partie de la capacité quotidienne de transit par le détroit d’Ormuz est tombée à des niveaux extrêmement bas, les statistiques indiquant peu ou pas de pétroliers y passant. Toutefois, le Brent n’a pas augmenté de façon constante au-dessus de 100 $. Après un bref pic fin juillet, il a récemment passé plus de temps autour de 90 $.

Le 30 juin, la société HYPE Treasury Hyperliquid Strategies (PURR) a été officiellement incluse dans l’indice Russell 3000 et l’indice Russell 2000. Plus tôt, PURR avait été inclus dans l’indice S&P Global BMI. À la suite de la cotation d’ETF liés à HYPE, cette étape marquante est également perçue comme une preuve supplémentaire du positionnement d’institutions de Wall Street sur des tokens HYPE et dans l’écosystème Hyperliquid.
Récemment, avec les dernières divulgations du dépôt 13F, davantage d’institutions d’investissement ont soit initié une position, soit augmenté leurs avoirs dans PURR, des informations sur ces actions ayant été rendues publiques. En prenant cela comme un indice, Odaily Planet Daily examine indirectement comment des institutions représentatives du secteur accroissent leur exposition à HYPE. (Note : La valeur totale des participations ci-dessous est basée sur le calcul du cours de l’action PURR au 30 juin.)