The reversal of yen arbitrage detonates liquidity, and AI misallocates funds to Bitcoin

According to Woofun AI, Maelstrom's chief investment officer Arthur Hayes attributed the shift in global macro-liquidity to the collapse of Japanese arbitrage trading and the inevitable correction of AI capital mismatches in The Rollup podcast broadcast on September 8, 2026. Hayes pointed out that this structural change is forcing the Federal Reserve to accelerate the creation of dollar liquidity, thereby providing strong upward momentum for crypto assets, particularly Bitcoin and Ethereum.

The ripple effect of Japan's policy shift began with Finance Minister Katayama Satsuki's statement in mid-late July. She asked domestic institutions to re-evaluate asset allocation standards, reduce holdings of foreign assets, and increase investment in Japanese domestic assets. This directive directly points to the Japanese Government Pension Investment Fund (GPIF), which is Japan's largest pension fund and has a quasi-government nature. At the time, the USD/JPY exchange rate hovered between 160 and 163. Looking back at history, the last time GPIF drastically adjusted its asset allocation was after 2012, and individuals and businesses have followed this change. At that time, Abe promoted Abenomics, stimulated the economy through banknote printing, and hoped that GPIF would increase the allocation ratio of foreign securities and reduce the allocation ratio of domestic securities. It took him two years to get official GPIF approval, which included removing opponents and appointing supporters of this direction.

Since then, GPIF issued an allocation framework for increasing foreign assets and reducing domestic assets. The market was launched. The dollar rose against the yen, the yen weakened, Japanese investors began to invest overseas, and others followed suit. Therefore, Hayes initially believed that GPIF may not start selling US Treasury bonds or buying Japanese treasury bonds until two or three years; this is not something that requires immediate attention. But then came the first intervention of the yen: Scott Bessent sold the euro, bought the yen, and proposed that the single counterparty quota limit for the Federal Reserve's foreign and international monetary authority repurchase facility (FIMA Repo Facility) should be lifted. He was actually pressuring Kevin Warsh to fulfill his duties and lift this cap.

This means that GPIF and other institutions do not have to sell US Treasury bonds, but can use US Treasury bonds as collateral to obtain US dollar loans from the Federal Reserve, then sell US dollars and buy yen in the foreign exchange market, and finally bring the capital back to Japan. This is only part of the puzzle, as Warsh is also required to convene the relevant financial subcommittee and have the committee agree to do so. Afterwards, the US Treasury proposed increasing the treasury bond repurchase scale by 20 billion US dollars, but compared to the approximately $40 trillion bond market, this is nothing.

Bessent said again last week or earlier this week that the Bank of Japan needs to raise interest rates more quickly. Not that no one has said anything like that; the key remains what action he is prepared to take. There's also the G20 meeting this week. Hayes believes that some kind of agreement may have been reached on the sidelines during the meeting, and the Japanese side has finally received the information. Bloomberg reported that GPIF held an unplanned meeting in August. August is Japan's holiday month, and it's unusual to have unplanned meetings at this time.

We don't know what was discussed at the conference, but previously the Japanese government asked it to increase its asset allocation in Japan, and Bessent also asked Japan to increase its domestic assets and sell US assets. Since then, USD/JPY fell from 160 to 155 in one trading day, and EUR/JPY also fell by about 3 yen during the Asian trading session, which is a very large fluctuation. Hayes believes there may soon be an announcement: either the FIMA repurchase facility limit has been raised, or GPIF has begun to adjust the allocation weights of domestic and foreign assets. The crypto market and other markets reacted overnight.

Meanwhile, Waller said that inflation doesn't seem to be that serious, and the Federal Reserve probably shouldn't raise interest rates. Looking at these things together, the goal is to weaken the US dollar and strengthen the yen. This has always been one of the top goals of the Trump administration, which wants to reshape the global trade structure. To do this, the yen must rise in value. The yen is probably the most undervalued currency in the world other than the renminbi. It is difficult for the US to do the same against China, but it can influence Japan because Japan depends on America's security guarantees. Hayes believes this is the reason for the rise in the crypto market. The market has been digesting all kinds of information until now, at last, a substantial change has taken place. In the absence of clear news, USD/JPY dropped from 160 to 155, indicating that something had changed. Therefore, he believes that the market has already started. Crypto assets and other assets rose overnight, while the S&P index remained broadly flat or fell, and technology stocks and AI trading did not rise significantly. This indicates the logic of liquidity behind it. More information may be revealed in the next few days or weeks, proving that an agreement was indeed reached during the G20 period, and corresponding arrangements will be introduced to reduce the dollar and push up the yen by creating dollar liquidity.

Hayes refers to Japanese society as “Japan Inc.” (Japan Inc.) , which operates the world's largest Japanese yen arbitrage transaction. If you look at Japan's consolidated balance sheet and include private sector assets, you will find that Japan has actually been buying foreign assets in Indian yen. With the depreciation of the yen and the rise in assets held by Japan, such as US technology stocks, the overall performance of Japan was good. Some people only focus on a single indicator, such as debt-to-GDP ratio, but Japan should be viewed as a whole. Although Japan claims to be a capitalist society, it has strong communal and socialist characteristics, and capitalism is more like an external form.

At the end of the day, there is a “Japan Corporation” in Japan. The yen arbitrage is a nationwide transaction, and Japan is also the biggest participant in this transaction. When GPIF is required to switch, “Japan Co., Ltd.” will follow suit: sell foreign bonds and stocks, sell foreign currency, buy yen, bring capital back to the country, and invest in Japanese treasury bonds, local companies, and real estate. This is exactly the order issued by the government. It will take a while to launch, but once it starts, you shouldn't stand on the opposite side of this trend. The problem facing the US is that Japan has held these assets for the past 30 years, which has boosted the US market. When the entire US system relies on financial benefits from rising stock markets and continued bond issuance, how should this deal be withdrawn?

America's response was only to print money and take over the deals Japan made in the past. Japan's strategy in the past was that it didn't matter if the dollar rose to 200 against the yen, as long as it was able to re-inflate the domestic economy and get rid of the problems left by the housing bubble of the 1980s through inflation. The US is currently adopting a similar strategy: even if the US dollar index (DXY) falls to 50, it is acceptable as long as it can become an industrial power again and reduce the debt-to-GDP ratio from about 100% to about 30% after the previous adoption of a similar strategy. The two are essentially the same deal. It takes a long time to form, but once launched, it's hard to buck the trend.

Hayes pointed out that the time when the US monetary environment was really limited was only from December 2021 to October 2023. Since then, Janet Yellen has begun issuing more short-term treasury bonds and bonds, and has taken away $2.5 trillion from reverse repurchase facilitation. For holders of crypto assets and other assets, the market has re-entered an upward phase since then. As the host said, AI transactions are a “liability exemption card” for them. Over the past 50 or 60 years, America has printed a lot of money. According to normal mathematical logic, the cost of accruing interest and the size of debt are growing exponentially, and it is almost impossible to solve it by economic growth alone. But now there's a new thing called AI.

The narrative is that as long as AI is developed and the AI competition with China is won, the debt problem will disappear and productivity will increase dramatically. That's why Kevin Warsh, Trump, Scott Bessent, and everyone else are talking about AI. Only in this way can they explain to voters that they don't have to worry about how much money the government has spent, nor do they have to worry about spending a higher share of GDP than at any time except during the war or pandemic, because the US has AI and will win the AI competition. But these people don't even know what AI actually means; they just accept the narratives Dario, Sam, and Elon are peddling to them. AI will also be incorporated into the same deal.

If AI is the only reason the government uses to explain how to solve the deficit problem and why not worry about spending, what will the government do once large AI laboratories are under pressure because the unit economic model is not established? It will bail out these companies, and the way to bail out is to invest more money. As a result, Japan-related trade structures and European issues will push the US to create more money; AI has given the government a reason to save face. The government has already wasted trillions of dollars on these illusory chatbots, which will also be the reason it continues to inject huge sums of money into the market. The combination of these two aspects will help crypto assets reach new highs.

Hayes believes that change is now beginning. A friend just sent him the cover of the latest issue of “The Economist,” which portrays Nvidia (NVDA.US) CEO Jensen Huang as an amazing wizard, as if Nvidia (NVDA.US) had no cash flow issues, no revolving financing, supplier financing, or accounting tricks like “Enron 2.0". Just add an AI chatbot, and it's the best company in history. Hayes believes this is a sign from the top of the market. When “The Economist” tells you one thing, you should do the opposite, because their judgment is very foolish. The current situation is very favorable for Bitcoin and gold, as politicians can no longer stop spending. Otherwise, they must admit that they have made huge mistakes in the past, including issues surrounding data centers, social media, and technology companies' use of user data.

If the government admits there is a problem with the direction of AI and changes policy, it must remove support for the industry, let Elon and others bear their own capital costs, no longer provide special regulatory arrangements, and stop using the nationalist narrative of competition between China and the US to solicit more investment for loss-making companies. At that time, the enterprise will either make money or not. Companies such as Anthropic should also disclose real profits, not just revenue figures; if they continue to burn money, they should explain the unit economic model of the reasoning business. Prospective IPOs or secondary market investors need to see this information. But that's clearly not going to happen, because politics doesn't work that way. That's why Bitcoin, gold, and other similar assets will perform well: we're already in a phase where capital is wasted. The government will use large amounts of newly created funds to renew these loans to cover up previous mistakes because it is unable to admit that it has wasted huge sums of money.

According to data compiled by Woofun AI, Hayes confirmed that the capital allocation to AI is a capital mismatch, and the funds used to deal with this mismatch will eventually flow to digital assets. He couldn't pinpoint who would buy Bitcoin. He believes that many venture capital institutions will instead suffer serious losses. They told investors that they had a very high book return for investing in these AI labs. Maybe Anthropic can go public, but it needs to go public soon as more people are questioning it. The OpenAI situation is more difficult; it requires government bailouts or some form of merger. Sam Altman had to design an impressive financial engineering in order for it to close the deal. As for Anthropic, it depends on whether Dario Amodei can do it. But many venture capital institutions' funds are actually locked in.

If these companies' stock prices drop 50% to 60% after listing, and liquidity disappears, without government bailouts, Hayes doesn't know how these institutions will achieve the DPI promised to investors. So it's not “people in the AI world are going to invest in crypto assets”. People in the AI world don't have cash; they only have a few book assets. If central banks continue to push up these assets, they may be able to exit and get cash and then buy crypto assets. But a more appropriate understanding is that the central bank undertakes widespread balance-sheet expansion to cover up capital mismatches. Bitcoin was born for this. What happened in 2009? Policymakers are expanding balance sheets to mask capital misallocations in the housing sector. This time it's essentially the same, only on a larger scale, and the object becomes AI debt.

Hayes believes Bessent is a firefighter. What really sets the narrative is the market: a rise in the 10-year US Treasury yield to 4.8% will set the story, and a rise of the dollar to 160 against the yen will also set the story. Bessent was the only competent person in charge. He is faced with many plates spinning at the same time and needs to try not to let them fall, so he can only reach an agreement here and another agreement there. Policymakers are completely controlled by the market and can only do their best under all the imbalances that have accumulated over the past few decades. At the end of the day, these problems can be traced back to the post-World War II system. They are the result of continuous interaction of various events over the past 100 years, which eventually formed today's situation. As a result, individual politicians are important, but they aren't that important because in the end, they can't beat math and compound interest. Operationally speaking, Warsh is the chairman of the Federal Reserve. He controls the balance sheet and can also create money. But at the end of the day, you can refer to former Federal Reserve Chairman Arthur Burns's speech “The Anguish of Central Banking” (The Anguish of Central Banking).

He gave this speech in 1979, and Hayes can't remember where. Burns was Volcker's previous chairman of the Federal Reserve, and financial historians generally think he let inflation go. The central point of the speech was that when the Chairman of the Federal Reserve took office, he believed in sound monetary policy and claimed that he would defend the independence of the Federal Reserve. But at the end of the day, they are still a subsidiary part of the US system. Americans vote to elect politicians to implement specific spending plans; what right can the Federal Reserve Chairman object? So, no matter how much you believe your duty is to protect the independence of the Federal Reserve and the value of the dollar, the real responsibility is still to cooperate with the government spending that the American people voted for. Ultimately, you'll always print money, and you'll always meet the president's demands in some way. History has proven that regardless of whether the Republican Party or the Democratic Party is in power, the results are the same. It doesn't matter what you said before you sit in this position; once you sit down, your superiors will always ask you to create money in some way. Warsh allegedly stepped down as Federal Reserve governor around 2011 due to his opposition to quantitative easing. Over the next 15 years, he made many tough remarks in the private sector, and his opinions didn't influence policy at the time. Now it's time for the Federal Reserve, what did he do? He set up a working group, which ultimately submitted only one report.

Hayes predicts Warsh will keep interest rates unchanged. They can easily find a third-order rate of change from one of the government's inflation indicators, and this indicator doesn't include what people actually spend money to buy, and then claim that this indicator is declining year over year, so they can keep interest rates unchanged. Meanwhile, the nominal growth rate of the US economy in the last quarter was about 8%, while short-term interest rates were only 3.5% to 3.75%. This is a textbook-style operation. By keeping interest rates unchanged, Warsh can continue to be hawkish, and on the other hand, it can also claim that reserve management purchases are not real balance sheet expansion or quantitative easing, but only an operation to deal with technical issues in the repurchase market. Most American voters don't really understand the buyback market; this statement may be enough to get through. Bessent, on the other side, will continue to try to maintain balance in all aspects and prevent problems from breaking out in his own hands.

Hayes believes Bitcoin could break through an all-time high before the end of the year. But at the end of the day, before the US midterm elections, the government still couldn't reveal its true intentions too clearly. The biggest concern of American voters is affordability, and Trump must try to explain why the various easing measures the administration is taking are not part of printing money. Hayes doesn't know if Bitcoin actually helps Trump if it rises to $50K the day before the election. For crypto asset holders around the world, we certainly want this to happen. On the one hand, the structural factors we discussed earlier require the government to create money, and we know this will happen; on the other hand, American politicians must also manage a clear political timeline. They must not let outsiders think they are pushing the Federal Reserve to print money. As Scott Bessent said in the Wall Street Journal review article, until now, most Americans have viewed the Federal Reserve as the maker of inequality. Therefore, the government must maintain the appearance that it still cares about the purchasing power of the money earned by American taxpayers. Therefore, Hayes is very bullish and fully concurs with the views of previous guests. However, the market may be very volatile. It may first rise rapidly, then trade sideways for a period of time, or even fall somewhat, and then continue to rise. As we move towards large-scale money printing step by step, the market will experience this rhythm over and over again. Will Clemente recently stated that he has never been so confident about the long-term outlook. Although it will still fluctuate in the short term, the direction of financial suppression is already clear. Hayes' view on the timeline and prospects of the Bitcoin and crypto markets is that the 2028 US presidential election will be a key point.

According to Hayes, Ethereum is the least popular large market cap token in the market. At the end of the day, if you want to take a higher risk than Bitcoin, but don't want to drop 75% overnight due to protocol issues, then Ethereum is the right choice. It's also the worst-performing large-cap coin in the last cycle, and hasn't even broken past its 2021 all-time high of close to $5,000. As a result, Hayes thinks Ethereum's risk-reward is excellent. This is why in this round of rising liquidity, ETH is his largest position. He also holds other assets with similar logic, but his positions are significantly smaller, such as ether.fi and Ethena.

Hayes said in a recent article that EUR/JPY is currently the most interesting indicator because it is the only leading indicator that the rate of liquidity creation in the US dollar has accelerated in the short term. Anyone who understands Bitcoin's logic knows that the creation of dollar liquidity is at the core of fiat currency depreciation transactions. However, the acceleration of liquidity creation is a second-order change, that is, money printing is beginning to accelerate. Hayes explained that EUR/JPY can tell us in advance that money supply growth will accelerate. Bessent often sends signals to the market, and the market trades these signals ahead of time. He recently used “potential energy” and “kinetic energy” to describe this relationship in an interview. He hinted to the market that the Ministry of Finance will buy back long-term treasury bonds, and the market will trade early, which will accumulate a large amount of potential energy. When the policy is finally implemented, will “buy expectations, sell facts”? Or is the actual scale of money creation sufficient to deliver on his signal and push the market higher than currently anticipated? Hayes apparently increased positions before the Federal Reserve's balance sheet rose. The current balance sheet is indeed rising, but compared to the COVID-19 pandemic or 2009, the magnitude is not extreme. That's why Bitcoin only went from around $6.3 million to $8 million, which isn't a particularly big increase. A real crisis is needed to resolve the perceptual issues brought about by the massive expansion of the Federal Reserve's balance sheet.

The crisis on the yen side is that the side that has long yen in the EUR/JPY deal, including individual Japanese investors such as GPIF, Nomura Securities, and “Mrs. Watanabe”, are selling foreign assets because the Japanese government has asked them to do so. To prevent these institutions from directly selling assets, they need to be provided with loans to finance them through repurchase. This is a pillar of balance sheet expansion, and it will also drive the yen to appreciate. The problem on the euro side is the buyback market. Large French banks, led by BNP Paribas, Crédit Agricole, and Société Générale, account for about 20% of the buyback market. If there is a problem with the euro, the first market to be hit will be France, because Japan holds a large amount of French debt. If Japan cannot sell US assets because the US has a large number of military bases in Japan, then it can sell European assets, and the first to sell will be French assets, including French treasury bonds OAT and Bank of France bonds. As the situation worsens in France, France cannot legally print its own banknotes under the rules of the Eurosystem. However, new French politicians may think that they are French presidents, not European Union presidents, and that their duty is to meet the needs of the French people and country. France needs more money, and it also needs to depreciate within the euro system. If it does not withdraw from the euro, the French government may ask the Bank of France to implement quantitative easing domestically. This is not legal under EU rules, but the French government may act on the grounds that it will save the country's bond market. The EU may say to Le Pen and Mélenchon: “I am capable of creating the euro and saving the French bond market, but since you don't want to bow to me, I won't buy French bonds.” Since both sides are unwilling to compromise under their respective power structures, a de facto “soft frexit” (soft frexit) may eventually develop. This is the euro's bearish logic. The Federal Reserve has switched to quantitative easing since December of last year to support the buyback market. The buyback market is financing short-term US Treasury bonds, and currently the person issuing the most short-term treasury bonds is Scott Bessent, so in the end, these issues are still part of the same deal.

If EUR/JPY falls from about 182 to 140 or even 120, the French banking system will have serious problems, which can only be solved by creating money. This could also mean the end of the Eurosystem, because France cannot unilaterally print money without leaving the ECB as a central institution. If French banks are concerned about capital controls or some kind of “quasi-euro lira” monetary system in their country, they will need to withdraw from the US repurchase market and return capital to the country. As a result, commercial bank balance sheets that the Federal Reserve thought it could rely on no longer exist. The Federal Reserve must therefore increase reserve management purchases, and it is already doing so. The Federal Reserve can explain this is not quantitative easing for technical reasons such as a long period of time, and hopes that the American public cannot understand what it really means. This is how the euro's bearish side creates money, and why Hayes believes that EUR/JPY can reflect whether the two specific factors that forced the Federal Reserve to rapidly increase the money supply have already started. What can really push Bitcoin to $250,000 or $500,000 is these actual changes, not just Bessent's statement on future policies.

Hayes called Scott Bessent “Buffalo Bill Bessent,” a term derived from the serial killer Buffalo Bill in “The Silence of the Lambs.” Hayes called him the “country's serial killer”: if you do business with Iran, sanctions will come at your door. Hayes believes AI assets are still likely to rise, but will not perform as well as scarce assets. If you're an AI investor and have already made a lot of money with AI, that's because the rate of change in the speed of AI development was very high in 2025 and 2026. But that stage is now over. This is not to say that AI-related assets won't rise, but rather that they may not increase as much as in the past because the market has already entered the “whether it was right to invest so much money before” stage, and the market usually peaks at this stage. Nasdaq is still likely to rise another 40%, 50%, or even 60%, but at the same time, Bitcoin may rise to $1 million, gold may rise to $15,000, and defensive assets such as ExxonMobil (XOM.US) may also rise several times.

AI assets will continue to rise, but they may not perform as well as other assets. As to which other assets to choose, it depends on individual perceptions and preferences. Hayes is clearly focused on the crypto market. In this macro environment, he believes that Bitcoin is the fastest running horse; this is one end of the barbell. What needs to be considered on the other side is what assets will benefit when people begin to believe that politicians will not continue to print money. Hayes believes this will not happen until the 2028 US presidential election at the earliest. The opposition Democratic Party may then propose to raise taxes, because the richest people have made a lot of money in this round of the market, while ordinary people have become poorer and inflation is rising. Whether or not the Democrats actually raise taxes in the end, the point is that the market is worried that they will win the election. They are likely to win because American politics often swing back and forth like a pendulum. At that time, the market may start to worry that future money printing will not be as large as previously anticipated, and investors will need to allocate it on the other end of the barbell. As far as Hayes is concerned, he holds a large interest in a volatility hedge fund, which expresses this view through options trading.

Investors can also choose other types of businesses: when money printers are shut down and the monetary environment tightens again, these companies can still perform well. The moderator pointed out that one end of the barbell is fiat depreciating assets in the crypto market, such as Bitcoin and Zcash. They have more room to rise, but the retracement could also be significant, particularly Zcash. At the other end are assets such as Hyperliquid and ether.fi. After investors are unlocked, they have a better token economy model and repurchase mechanism. The downside support may be higher, but there is less room for growth, because there is an upper limit on fundamental valuations, and only business expansion can further drive the token price. The host believes that the industry has finally entered the stage of “being able to produce actual results”, which is why they think the bear market is over. In addition to policy changes, the token itself has also begun to have real business and good performance. Many of the early players have made money and left, but the industry seems to have reached a stage of maturity similar to the one after the collapse of the internet bubble in 2001. Since then, the asset has been rising for a long period of 25 years, and there have been slow growth, repeated bottoming out, and sharp retractions in the process. The crypto industry will still have cycles, but both in terms of products and external exports, it seems to be entering a long-term stage of development, and there is more substantial value in the industry than in the past. Hayes agreed. As for HYPE, Hayes still doesn't think its risk-reward is good. That's not to say it won't go up; it will definitely go up. But with the same risk capital, it may not have risen as much as Ethena. That's his current opinion on Hyperliquid. As for Zcash, it also depends on the progress of formal verification and other work.