Arthur Hayes' new article: The AI siphon has ended, but I still believe in ETH

Original Title: Situationship

Original Author: Arthur Hayes, Co-founder of BitMEX

Original Compilation: Azuma, Odaily Planet Daily

Looking around, humanity has transformed the Earth's natural environment into something entirely different. Some changes are awe-inspiring, while others are shocking, but without exception, they all began with an idea in the minds of one or more evolved primates—namely, humans.

As our brains process vast amounts of information every day, we continuously construct various narratives to make the world coherent and meaningful. For this reason, the narrative itself ultimately shapes reality.

For investors, to predict the future fluctuations of market prices, it is essential to understand which "collective illusions" market participants collectively believe in. The same company, under unchanged future cash flows, can receive vastly different valuation multiples simply because the market believes in different stories.

The simplest way to achieve a "valuation re-rate" for a previously dull and boring company is to replace it with a new narrative that aligns with current market trends, making investors willing to chase it at any cost.

Is There a Bubble in AI?

This leads to the core question of whether AI is in a bubble.

However, before discussing the AI bubble, it is worth answering a more fundamental question: "What exactly are we investing in?"—to put it in relationship terms, "What exactly is our relationship?"

At least from my somewhat "Luddite" perspective, the key lies in how the market defines AI capital expenditure (AI CAPEX)—does it belong to technology or real estate?

The mainstream narrative in the market today believes that this multi-trillion dollar AI infrastructure buildout belongs to "technology," and therefore should enjoy extremely high growth valuations.

But my view is quite the opposite. AI CAPEX is essentially just another boring real estate investment. The only difference is that this time, the data centers are filled not with office buildings, but with computing power. This computing power will ultimately give rise to silicon-based lifeforms, driving the development of human civilization, and its significance may even surpass that of the railway revolution.

It is crucial to distinguish whether it is "real estate" or "computing power" because many newly matured hedge fund managers, banks, private credit funds, and even governments mistakenly believe they are lending to tech giants like Apple, rather than providing real estate financing to Lehman Brothers.

I believe the reason the AI bubble will eventually burst is that financial intermediaries will overbuild data centers and all the supporting infrastructure needed for data center construction, including energy, electricity, and everything required for AI chip training and inference.

Thus, the AI bubble resembles a 2008-style credit bubble, rather than a profit bubble like the 2000 internet bubble.

During the 2000 internet bubble, most publicly listed internet companies had virtually no revenue, let alone profits, such as Pets.com, so that bubble was essentially a "profit not materializing" issue.

In contrast, the 2008 financial crisis was different. The real trigger for the crisis was the slowdown in the rise of U.S. housing prices, which raised concerns among banks and financial institutions about the solvency of mortgage assets, leading to a credit crisis.

The AI bubble will follow a similar logic. The real turning point will not be when AI leading companies stop making money, but when the growth rate of data center construction begins to slow down, or when cloud computing giants (Hyperscalers) lower their future data center construction guidance.

Even if AI leading companies can still earn huge profits, their forward valuation multiples will still contract due to declining growth expectations. The first to fall will be those AI companies with the weakest credit conditions and the highest leverage.

Subsequently, these risks will quickly transmit to the balance sheets of financial institutions that hold large amounts of AI debt assets and are similarly highly leveraged. Ultimately, the government will once again step in under the guise of "national security" to ensure that these over-leveraged AI companies and the financial institutions behind them do not collapse.

And this misallocated capital will ultimately flow into the crypto market… sending Bitcoin to the moon again.

Credit Risks of AI CAPEX

Whenever someone claims "AI is in a bubble," AI bulls almost always cite the "Jevons Paradox" as a counterargument. Jevons argued that when the price of a commodity falls, its usage will increase significantly, thereby expanding the overall market size, even exponentially.

If you believe that AI capital expenditure itself represents computing power demand, then according to the Jevons Paradox, there is indeed nothing to worry about. As computing power costs continue to decline, demand for AI token consumption applications and AI agents will grow exponentially, making lending to AI infrastructure a sure-win business.

But I believe this is actually a misinterpretation of the Jevons Paradox. To understand why the Jevons Paradox does not mean that all credit flowing into AI CAPEX will yield returns, let's take a look at what a cloud computing giant (Hyperscaler) is doing when it builds a data center.

Essentially, it is first undertaking a real estate development project. It constructs a building to house server cabinets and then procures the latest generation of semiconductor chips for AI model training and inference. As industrial technology—especially semiconductor manufacturing—continues to advance, the floating-point operations per second (FLOPs) provided per kilowatt-hour of electricity will continue to grow exponentially.

In a few years, whether it's Nvidia, AMD, Intel, Huawei, or SMIC, they will all launch new generations of AI chips that are far more efficient than today’s. At that point, the same data center will be able to produce over 1000 times the intelligence while consuming less power.

This means that two things can be true at the same time: on one hand, the construction of physical infrastructure like AI data centers could very well experience supply saturation; on the other hand, the consumption of AI tokens could still grow exponentially.

So, the real question to ponder is, what exactly do you want to hold—a real estate business, which is what today's cloud computing giants are doing; or the AI application layer?

A common rebuttal from AI bulls is that cloud computing giants (Hyperscalers) are both landlords and tenants. They rely on the massive cash flow generated from the "selling attention" business of the Web 2.0 era to provide credit support for building data centers; at the same time, they leverage their AI capabilities to sell "wisdom apples" from the Garden of Eden to the world.

If you truly believe this story, then I hope you hold their stocks, not their bonds. There’s a reason bonds are called fixed income—no matter how successful a company ultimately becomes, the best outcome for creditors is merely to recover their principal and earn a bit of interest.

If Google successfully bets on AI and creates revolutionary products that change the course of human civilization, sending its stock price soaring, then shareholders certainly have reason to cheer; but bondholders will still only receive their principal.

Conversely, if Google ultimately becomes a "data center landlord" renting out a large number of depreciated Nvidia GPUs but fails to generate enough income to repay its debts and interest, then creditors will suffer significant losses. And how much a data center filled with outdated chips will be worth is also highly questionable.

The CFOs of cloud computing giants and Wall Street financiers are not foolish. They know they are essentially in the real estate business. Therefore, they must find some "greater fools" who believe they are investing in high technology rather than real estate.

These greater fools include insurance companies under alternative asset management giants like Apollo, as well as taxpayers in various countries who will ultimately foot the bill for the government’s implicit guarantees of AI credit.

If you carefully examine those deliberately convoluted financial statements, you will find that the massive debt issued to finance AI CAPEX is almost entirely off-balance sheet, with little clear connection to the core profitable businesses that support stock valuations.

How we define AI CAPEX determines how we understand the entire AI investment cycle. It is this narrative that explains why capital can become severely misallocated and why the scale of this bubble may exceed that of the railway bubble back in the day.

More importantly, because the AI bubble is a credit bubble rather than a profit bubble, when the crisis erupts, the government will inevitably step in to rescue those last greater fools who mistakenly bought traditional real estate debt as new tech equity assets.

Do not assume that the recent adjustments in the AI sector, especially in high-leverage markets like South Korea, mean that the AI bull market has ended. On the contrary. The truly crazy "blow-off top" phase may have just begun.

Just last week, the Federal Reserve had the opportunity to respond to inflation, which remains above trend levels across various statistical measures, by raising interest rates, but it did not do so; instead, it chose to remain on hold, with even former Chairman Powell casting a vote in favor of maintaining the current interest rate.

So, for those cryptocurrency players who have been forgotten by the market and can only struggle in a sideways bear market, does the allocation of AI credit matter? It matters because it determines how governments will fill the financial holes created by the out-of-control AI CAPEX investments—what methods they will use, why, and how much money they will print.

The following content of this article will revolve around this theory and explain why the government will ultimately have no choice but to print money to rescue the market.

As the growth rate of AI CAPEX slows while the scale of credit continues to expand, Bitcoin will complete its bottoming process and initiate a long-term upward trend. By the time decision-makers finally realize that the AI GDP growth they have high hopes for is essentially just another ordinary real estate bubble, they will have to initiate monetary easing on a scale that even exceeds that of the 2008 global financial crisis (GFC).

Ultimately, this will drive Bitcoin to break through $1 million, or even higher.

The Second Derivative Determines Everything

I constantly remind myself: "What we are really trading in investments is not growth itself, but the acceleration of growth."

In other words, what we are concerned about is the second derivative—whether growth is accelerating or decelerating.

This is quite intuitive. When an asset is still in an accelerating growth phase, people will continuously weave stories about its infinite possibilities in the future. Thus, phrases like "I would rather see a cloud computing giant go bankrupt than miss the opportunity to create AGI (Artificial General Intelligence)" will emerge in the market.

However, any growth will eventually enter a deceleration phase. The problem is that the price movements of most assets often follow this pattern:

• Acceleration phase: Prices continue to reach new highs;
• Deceleration phase: Prices fluctuate sideways;
• Only when growth itself (i.e., the first derivative) turns negative do prices truly begin to fall.

No one can accurately predict how long it will take from the start of growth deceleration to actually entering negative growth, but many investors, including myself, subconsciously believe that—even if growth has begun to decelerate—asset prices can still rise indefinitely.

If the AI bubble is essentially a credit bubble, then the importance of the second derivative becomes even more pronounced. Because society's willingness to continue financing AI CAPEX is based on one assumption—AI investment will forever maintain accelerating growth.

Once this acceleration disappears, continuing to increase debt will become increasingly dangerous, but the reality is that before truly taking a punch, no one knows when to stop.

Either wait for a financial crisis to erupt; or wait for "Ken Griffin"—who recently took over the AI stock positions at low prices—to take all your assets at the market bottom.

Therefore, even if investment growth has begun to slow down, the scale of credit often continues to grow. Only when AI CAPEX budgets truly begin to decline will the market welcome that classic "Wile E. Coyote" moment—when the character has already run off the cliff but only realizes there is no ground to support him when he looks down, and then falls instantly.

At that point, the market will begin to recognize who has become over-leveraged due to holding large amounts of junk AI CAPEX debt.

Let’s apply this logic to the U.S. subprime crisis. One of my favorite courses in college was about U.S. housing policy and the mortgage market. The instructor had served as Deputy Secretary of Housing during the Clinton administration. Coincidentally, I took this course in the spring of 2008—right when Bear Stearns collapsed, making it quite timely.

The core point conveyed in this course was that the government, in pursuit of the social equity goal of "homeownership for all," continuously encouraged more people to buy homes, leading to sustained credit expansion. However, by 2006, many first-time homebuyers were actually unable to afford the monthly payments after the loan interest rates reset. The only way they could continue to make payments was if housing prices continued to rise at an increasingly rapid pace.

Of course, I am still waiting for the government to fulfill its promise of "forty acres and a mule" (a historical land compensation promise in the U.S.). If that’s the case, why not just print money to build houses?

The following four-panel chart shows:

• S&P 500 Index;
• U.S. construction loans and construction activity;
• Case-Shiller National Home Price Index.

By the end of 2005, the growth rate of U.S. housing prices had begun to slow, which corresponded exactly to the peak of actual construction investment spending (the orange line in the first chart). However, real estate credit (the purple line) continued to flow into the market until the stock market peaked and began to show slight corrections.

From 2006 to 2007, it could be said to be a "no man's land" before the entire crisis erupted, as housing prices continued to rise, but the growth rate was constantly slowing; subsequently, the stock market peaked in mid-2007 (the pink dashed line in the chart); the real "Wile E. Coyote moment" occurred in August 2007—when three credit hedge funds under BNP Paribas collapsed; the crisis then spread, ultimately collapsing Bear Stearns and Lehman Brothers in September 2008… and prior to that, the S&P 500 index had already fallen about 50% from its peak.

What truly triggered the financial collapse was that investors finally discovered who held those toxic "Frankenstein" financial derivatives. Ultimately, the government had to take over the debts and equities of these institutions to avoid a new Great Depression.

This is very important because the following discussion on how the government will rescue the AI industry will return to this logic.

The second chart is also worth noting. It shows that the starting point of capital misallocation coincided with the slowdown in housing price growth. If new credit was still being used to build more housing, the problem would not be severe, but if the entire system began to rely on borrowing new debt to pay off old debt, then risks would begin to accumulate. The rising ratio of construction loans to construction investment spending is the best reflection of this process.

Now, let’s apply the same analytical framework to AI. The key variable here corresponds to the CAPEX spending plans of various companies.

The current market believes that real estate (here referring to AI) is technology; the more technology investment, the higher future profits will be. Therefore, the market rewards those cloud giants that announce increased capital expenditure budgets by driving up stock prices.

I expect that the growth rate of AI CAPEX announced will begin to slow down from mid to late 2027, and by 2028, the market will clearly enter the "growth deceleration phase."

At the same time, a seemingly contradictory phenomenon will occur: although the growth rate of CAPEX begins to decline, the scale of credit flowing into AI will continue to expand. The reason is that lenders believe they are investing in technology, not real estate. Coupled with the fact that governments are constantly emphasizing the need to dominate in global AI competition, continuing to provide financing for all projects related to AI CAPEX seems to be the most reasonable choice.

Thus, 2027 will become a "no man's land" similar to 2006 to 2007. The recent significant adjustments in AI stocks are merely a normal correction within a bull market. The true peak of the AI bubble will appear next year.

After that, the market will begin to reward those cloud computing giants that "first exit the arms race" and actively reduce CAPEX budgets. Unlike the early bubble phase from 2022 to mid-2026, future cloud computing giants will find it increasingly difficult to rely on their free cash flow to support AI investments. They will have to rely more on issuing bonds and increasing stock offerings to raise funds.

The pressure on balance sheets will also force management to seriously consider: "Is it really worth it to continue borrowing money to build more data centers just to accommodate more continuously depreciating chips?"

At least for American cloud computing giants, cutting-edge AI models from China will completely extinguish their "silicon deity" fantasies, especially when they are priced lower with comparable performance. After all, if two products are of the same quality or only slightly inferior, the vast majority of people will choose the cheaper one.

As the intelligence generated per kilowatt-hour from AI chips continues to grow exponentially, and under competitive pressure from China, the cost per token continues to decline, a rational cloud computing CFO will not continue to worsen their balance sheet merely to build more data centers.

Even according to the Jevons Paradox, the demand for AI tokens will eventually experience explosive growth, but this growth will not come fast enough to offset the negative impacts of the massive debt issued years ago. Ultimately, the market will first punish the participants with the worst credit. At that point, people will truly realize how much capital has been wasted in this AI investment wave.

I cannot predict which cloud computing giant will be the first to overreach, prompting bond investors to collectively exclaim, "Oh shit!"

However, before discussing why banks, knowing the massive risks of AI investments, still have to continue lending, let’s first take a look at the chart below. It shows the scale of CAPEX investments that major cloud computing giants have already committed compared to their cash on hand.

What supports the entire narrative of the AI bull market is actually trillions of dollars of leverage. Among these companies, one will eventually fall from grace like the once-celebrated AI genius Leopold Aschenbrenner. The difference is that when that happens, the ones coming to their rescue will not be those greedy and neurotic East Coast hedge fund managers from Wall Street, but the money printing machines in the hands of Warsh and U.S. Treasury Secretary Bessent.

The Dilemma of Bank Credit

Many believe that the slowing growth rate of AI CAPEX means the AI bubble is nearing its end. If that were the case, why would banks continue to lend?

The reason is actually quite simple: First, because it is profitable; second, because the government wants them to do so; third, because they know that even if the loans eventually turn into bad debts, the government will step in to rescue them.

Louis-Vincent Gave of Gavekal Research published an interesting article last week. He argued that Warsh's interest rate policy actually follows a very simple logic—actively steepening the yield curve.

This has two benefits. First, it makes lending more profitable for banks; second, it gradually dilutes the massive debt burden of the U.S. through inflation.

Ultimately, banks will continuously create new loans, which means creating new money. This new funding will provide financing for the re-industrialization of American manufacturing while continuing to support AI construction. This line of thought aligns closely with Treasury Secretary Bessent's recent emphasis on "Hamiltonian Economics."

If we look at all objective economic indicators, the Federal Reserve should have raised interest rates at its most recent meeting, but the fact is it did not. On the contrary, long-term Treasury yields quickly surged afterward.

Odaily Note: The yield on 30-year U.S. Treasury bonds rose rapidly after the Federal Reserve remained on hold.

Many believe this is a policy mistake by the Federal Reserve, but from the perspective of banks, this is a godsend.

The reason is simple. Banks can almost finance at the cost of the federal funds rate, while the Federal Reserve deliberately keeps this rate below the nominal economic growth rate and even below the actual inflation rate.

Subsequently, banks lend this money in the form of long-term loans to AI data center developers, rare earth mining companies, military manufacturers, etc. The steeper the yield curve, the higher the net interest margin banks can earn.

As shown in the chart of commercial and industrial loan scales below, banks are increasingly motivated to continue creating new money through lending.

Odaily Note: The white line represents the difference between the 10-year U.S. Treasury yield and the effective federal funds rate (reflecting the steepness of the yield curve); the yellow line represents the balance of commercial and industrial loans at U.S. commercial banks.

From a political perspective, this is a sustainable Federal Reserve policy. Even though the Trump administration's Justice Department has investigated and even prosecuted some Federal Reserve governors (like Lisa Cook and Powell), these voting members still support maintaining short-term rates at negative real levels.

In other words, Warsh effectively has a "volunteer coalition," which includes people from the Trump camp as well as those who have been deeply affected by "Trump Derangement Syndrome" (TDS) and are still within the system.

From a monetary policy perspective, the recent actions of the Federal Reserve also allow Treasury Secretary Bessent to issue short-term Treasury bills (T-Bills) at yields below the nominal economic growth rate. If the market cannot absorb the massive weekly issuance of Treasury bills, then the RMP (Reserve Management Program) will print money to fill the demand gap.

To suppress those "disobedient" long-term Treasury yields that continue to rise, Bessent can also implement Treasury buybacks—first issuing short-term Treasury bills that the Federal Reserve monetizes, and then using those funds to buy back 10-year or 30-year Treasury bonds, thereby lowering long-term rates.

It is worth noting that Warsh, who is known for advocating a reduction in the Federal Reserve's balance sheet, shows no intention of limiting or even stopping the expansion of the RMP project. This is all just a Kabuki-style UFC performance on the White House lawn.

If you are a credit approval officer at a "too big to fail" (TBTF) bank and hope for future promotions and raises, you will almost certainly approve loan applications belonging to "key industries" like AI, military, etc.

The reason is simple. This both increases bank profits and aligns with the policies of the Federal Reserve and the Treasury. Even if the loans ultimately blow up—mathematically, this possibility is quite high—the government will certainly roll out a "bazooka-sized" rescue plan quickly.

There is virtually no downside risk. This is how "window guidance" operates in the U.S.

I believe that a 2026 version of the "Treasury-Fed Accord" has already quietly occurred; it just hasn't been formally announced. Otherwise, how else can you define the current situation?

• The Federal Reserve maintains negative real interest rates;
• The Federal Reserve prints money to purchase Treasury bills issued by the Treasury;
• The Treasury encourages banks to lend to key industries;
• Once loans go bad, the ruling government will step in with implicit guarantees.

If this isn't a collaboration between fiscal and monetary policy, I don't know what else to call it. So, I am currently extremely bullish on the market; the real large-scale money printing has not yet ended.

U.S. Sovereign Wealth Fund

Now, let’s boldly stretch our imagination. What if the U.S. government not only rescues banks after a crisis occurs but also proactively buys AI company stocks at the first sign of a crisis? After all, thought experiments that stretch the imagination are always interesting.

In fact, the Trump administration's rescue of AI has already begun. Under the guise of "national security" and "U.S.-China competition," the U.S. government has started borrowing money and directly purchasing equity in so-called "key industry" companies like rare earths and semiconductors.

This is essentially an operation to increase dollar liquidity, which can also be understood as equity QE (Quantitative Easing). Because these dollars, which originally sat in government accounts, are being directly injected into the financial markets.

Here are some examples of how the U.S. government has used funds borrowed from the CARES Act, CHIPS Act, and the Department of Defense budget to directly hold equity in relevant companies.

Unfortunately, for us crypto investors whose wealth entirely depends on changes in the scale of money printing, there is little room left under the current legal framework for the government to continue making similar equity investments.

However, the Trump administration and Treasury Secretary Bessent have clearly shown an attitude that as long as the law allows, they will not hesitate to use borrowed money to buy AI stocks.

Thus, a new question arises: Is there a way to print money to buy AI stocks in advance without waiting for a crisis to erupt, and without needing Congressional approval?

The answer is yes! And this is precisely the most interesting part.

According to the Federal Reserve Act, in so-called "emergency and exigent circumstances," the Federal Reserve can print money directly and provide unlimited liquidity loans to special purpose vehicles (SPVs) established by the U.S. Treasury.

During the 2008 financial crisis and the 2020 pandemic, the Treasury used the "Exchange Stabilization Fund" (ESF) to support first-loss equity, which was then financed by the Federal Reserve to purchase various financial assets to stabilize the market.

Currently, there is still about $28 billion in the ESF account. Bessent could easily use this funding as initial capital for a new SPV, with the justification of maintaining national AI security.

According to past practices, the Federal Reserve is usually willing to provide up to 10 times leverage for SPVs. This means that Bessent could theoretically leverage about $280 billion to invest in AI companies that have yet to turn a profit. Of course, compared to today’s AI companies with market capitalizations in the trillions, $280 billion is far from "heavy firepower."

So, can the scale be further expanded? For example, could the Treasury simply establish an SPV without any first-loss capital buffer, allowing the Federal Reserve to provide unlimited loans directly? Technically, this can be done, but it means the Federal Reserve would have to endure immense political pressure—because the outside world would perceive it as secretly conducting unlimited-scale equity quantitative easing.

So, does the Federal Reserve really care about political pressure?

The answer is both yes and no. New Chairman Warsh has consistently emphasized that AI will soon become a miracle that enhances U.S. productivity. In other words, ideologically, he believes in the grand narrative painted by AI entrepreneurs.

If Trump tells him that to save Sam Altman and OpenAI, the government must directly buy stocks, the reason being that there are not enough retail investors willing to spend real money to purchase a yet-to-be-profitable cutting-edge AI company; meanwhile, Anthropic, founded by Dario Amodei, is not only profitable but also performs better.

Then, Warsh would likely act without hesitation. Of course, according to protocol, approving SPV loans still requires the votes of three other Federal Reserve governors. But considering that in the most recent meeting, members including Cook and Powell have already aligned with Warsh (supporting the maintenance of interest rates), if Warsh truly pushes the Federal Reserve down this path, I see little substantial resistance.

After all, compared to theoretical concerns about whether to print money, the investment returns in personal stock accounts are always more persuasive.

If the Treasury uses the printed money to support those soon-to-be-listed AI star companies in issuing new shares, it is essentially preemptively realizing the unrealized profits on the books of early investors and employees. This is the purest form of "liquidity creation."

Because before the government steps in to support, this capital does not exist. It is precisely the government's willingness to provide buy orders for those already unreasonable primary market valuations that allows this book wealth to truly "realize."

From an accounting perspective, the government can gain two benefits.

First, as long as an AI company has the government's backing, investors will flock to it. After all, following those with the power to print money in stock trading will almost always yield profits, at least in the initial stages. Thus, this SPV will quickly accumulate massive unrealized gains on its books. Trump could easily package these book profits as government "earnings," even claiming they could theoretically offset the fiscal deficit. If AI truly is the most important technological revolution in human history, then merely based on the paper gains in the stock market, from an accounting perspective, it could even "erase" the entire U.S. fiscal deficit.

Second, those newly minted millionaires, billionaires, and even trillionaires will have to pay federal and state capital gains taxes when they sell their stocks. This new tax revenue can also reduce the fiscal deficit, allowing the government to borrow less and further claim that the U.S. debt-to-GDP ratio has declined. At least initially, the bond market will believe this story, leading to a decline in Treasury yields, and the market will reward the government for this "accounting magic."

However, I must emphasize that Trump did not invent the philosopher's stone. He is merely postponing the problem, hoping that it will ultimately be picked up by the next administration—preferably another Republican government.

Why will this model ultimately lead to disaster? We can conduct a simple thought experiment. Suppose you want to become a billionaire overnight without wanting to work. So, you spend a few thousand dollars to register a company and issue a total of 1 billion and 1 shares.

Then, you sell one share to your mother for $1. Since the latest transaction price is $1, on paper, the other 1 billion shares you hold are worth $1 billion. Next, you take this "wealth" to the bank, hoping to borrow $100 million to buy a mansion, a Lamborghini, and various luxury goods. The bank will simply tell you: "Not a chance."

You would be confused. Because in your view, the loan-to-value (LTV) ratio for this loan is only 10%, and the risk seems very low, but the bank's answer is simple:

"If you need to sell these stocks to repay the loan in the future, there is simply no liquidity in the market."

Putting this logic back into the AI SPV context is the same. If the SPV has become the largest single shareholder of an AI company, and other investors are buying stocks merely because the government is involved, then once the government is ready to exit, there will be no real buyers in the market. Not only that, but when those politicians—like Ro Khanna, Nancy Pelosi, etc.—start selling stocks, all investors will rush to sell before the government, causing the book profits originally used to "offset national debt" to vanish instantly. They will not only turn into actual losses but will also further increase government debt.

Worse still, the Treasury will still have to repay the money it borrowed from the Federal Reserve. Therefore, for this SPV, it is essentially an investment that can only buy, not sell. The Federal Reserve can only continue to roll over the SPV's loans, ensuring that it never triggers a margin call.

Ultimately, to maintain the entire system, the expansion of the Federal Reserve's balance sheet will become permanent.

However, this is not something Trump needs to worry about. Politically, he has already profited on both sides—on one hand, those unprofitable U.S. AI companies have received funding to continue competing with China; on the other hand, the book "wealth" created by AI has driven tax revenue growth and stimulated current economic activity.

Meanwhile, unrealized gains combined with new tax revenues create an illusion that "the U.S. debt-to-GDP ratio is declining." Thus, the market is willing to continue lending to the U.S. government at lower rates.

The U.S. government can do this now to preemptively prevent the AI bubble from bursting. Of course, it can also wait until the growth rate of AI CAPEX slows down and the market begins to sell off AI stocks comprehensively before stepping in to rescue.

Since the U.S. government has already begun directly purchasing corporate equity, why not buy more? By combining "bank window guidance-style lending" with "government directly buying AI stocks," it is theoretically possible to ensure that an AI credit crisis never occurs. At least not before the 2028 U.S. presidential election.

Some may ask, after all this, if so much money has already been printed from 2022 to now, why hasn’t Bitcoin broken through $126,000? Hold on, the next section will tell you the answer.

When Will Bitcoin Bottom?

The bottom of the last cycle occurred when the market discovered that the white boy Sam Bankman-Fried had stolen FTX customer funds, and CZ helped facilitate this discovery.

At the same time, ChatGPT was commercialized, and the AI wave began.

Starting in October 2023, the U.S. liquidity environment changed, as funds from the overnight reverse repurchase agreement (RRP) tool continued to flow out, leading to an increase in dollar liquidity. Subsequently, bank credit and government borrowing also began to grow. Bitcoin rose as a result and peaked in October 2025; however, Bitcoin did not continue to rise, as it only increased about 2 times compared to its previous historical high, because AI credit and AI stocks absorbed the newly added fiat liquidity.

As AI capital expenditure (CAPEX) expansion accelerated, consuming all available fiat liquidity, Bitcoin—this point seems obvious in hindsight—fell by 50%.

By mid-2026, the liquidity environment will reverse. The growth rate of AI capital expenditure announced for the next 18 months will begin to slow down, but the channels of bank and government credit will just begin to create dollars and funnel them into the AI industry.

If banks cannot fulfill their "patriotic duty" to continue providing credit, the government will strongly push them to lend to AI. If that still fails, the government will reduce the risks of banks lending to the AI industry by providing equity support to specific AI companies and procurement commitments (offtake agreements) similar to Intel and IBM.

Bitcoin will bottom out in this early stage of credit misallocation, and the financialization process of AI—where the amount of dollars and renminbi chasing quality AI projects in the market exceeds the scale of truly quality projects themselves—will ultimately lead to capital misallocation.

As I write this article in late July 2026, I do not know at what price Bitcoin will ultimately bottom; perhaps the bottom has already appeared.

The market needs time to digest the concerns brought about by Strategy (formerly MicroStrategy) selling Bitcoin. At the same time, the market also needs to find a new narrative; if Strategy cannot continue to issue stocks or find investors to purchase its preferred shares and use those funds to continue buying Bitcoin, then why would Bitcoin still rise?

Perhaps Bitcoin will oscillate between $60,000 and $70,000 for a while and may dip to $50,000. However, during all this, the waste of AI capital will continue to accelerate, laying the foundation for Bitcoin to bottom out and slowly rise afterward.

If my view is correct—that the scale of truly valuable AI capital expenditure projects is smaller than the scale of credit flowing into "AI"—then Bitcoin's price will ultimately reflect this excess liquidity. This will help Bitcoin bottom out, even if digital asset treasury companies (DATs) like Strategy can no longer purchase Bitcoin in a way that increases the Bitcoin-per-share value through the stock and corporate bond markets.

I will continue to monitor several indicators to validate this logic:

• Whether the growth of AI capital expenditure slows down;
• Whether the scale of AI loans increases;
• Whether hyperscale cloud computing companies increase off-balance sheet commitments.

If we enter a phase of capital waste in this round of AI credit boom, then the next question is: What will regulators and governments do? Will they print money in advance? Or will they wait for the final crisis to erupt due to a lack of political space and then intervene through rescue plans?

Fortunately, as long as we hold Bitcoin in a non-leveraged manner, we do not care when the rescue arrives. Because we know that due to the distortions in government incentives, they will ultimately choose to print money to save the system. The current scale of the AI capital expenditure credit frenzy is already comparable to the proportion of GDP during the railway construction period. This means that the scale of capital misallocation has already exceeded that of the U.S. subprime crisis.

Therefore, the scale of future rescues will exceed the trillions of dollars printed by the Federal Reserve and major global central banks between 2009 and 2013. Bitcoin was born in response to the "irresponsible rescue of bankers" during the subprime crisis. If you think about it carefully, this is quite astonishing. And this time, Bitcoin already exists; it may fulfill the dreams of many people—rising to $1 million or even higher.

Given the current dismal state of the crypto capital market, imagining such a future is not easy. But in my view, this creates an interesting asymmetrical opportunity. Maelstrom has long held a large amount of Bitcoin.

Aside from Bitcoin, what new narrative could drive a large-cap token up in the next six months? Ethereum is currently the most hated and forgotten large-cap "junk coin" in the market. It hasn't even broken through the historical high of $5,000 from 2021, while most of the top ten junk coins have already surpassed that.

In my view, the next narrative is that companies like Robinhood will launch RWA (Real World Assets) chains using customizable Ethereum Layer 2 solutions like Arbitrum. Ethereum will become the securities settlement layer for these chains. Therefore, even if the actual gas revenue flowing to Ethereum constitutes a small proportion of the entire system, ETH will still be the junk coin driving "tokenization of everything."

I am a critic of RWA. Maelstrom often receives numerous financing pitches for junk projects, with teams claiming they will ride the wave of asset tokenization, while on the other hand, traditional finance (TradFi) loves to discuss: "In the future, all assets will be tokenized and run on some private or public chain."

I firmly believe that if this future truly arrives, then these TradFi RWA projects must operate on public blockchains. And Robinhood launching its own chain using Arbitrum will reduce the professional risks for TradFi practitioners—they can replicate the same model and ultimately build Ethereum-based solutions.

This narrative is very strong. And ETH, as a junk coin, is the second-largest crypto asset by market cap and has existed since 2015, so it has the Lindy effect (the longer something exists, the higher the probability it will continue to exist) second only to Bitcoin. Coupled with Tom Lee from Bitmine endorsing ETH for institutional investors, it allows fund managers to bet on the trend of capital market tokenization.

My rough target price for ETH by the end of 2026 is $5,000, which is about 2.6 times the current price. I like this trade because I can take a relatively large nominal position while being willing to accept the risk of ETH plummeting 75% one day due to some technical flaw, which is a very low risk.

Additionally, ETH is highly liquid. Therefore, even though it occupies a large proportion of the Maelstrom portfolio, I can still exit in a matter of minutes. Finally, I will also sell out-of-the-money puts to earn extra income while accepting the risk of buying ETH at a discounted price if it falls below the strike price.

The AI bubble once siphoned liquidity from the crypto market, but that situation has ended. As the market narrative gradually shifts from "investing in AI at any cost" to "what is my investment return," and finally to "when will I recover my principal"… those governments that have bet their entire economic policies on AI will begin to worry—perhaps this bubble really could burst.

To avoid admitting their mistakes and to prevent this outcome, they will engage in massive capital misallocation, and this scale of capital misallocation will ultimately create a cryptocurrency bull market the likes of which we have not seen since 2021.