Article by: Arthur Hayes
Compiled by | Odaily Planet Daily (@OdailyChina); Translator | Azuma (@azuma_eth)
Looking around, humanity has transformed the Earth's natural environment into something else. Some changes are admirable, others are shocking, but without exception, they all began as an idea in the mind of one or more evolved primates — humans.
As the brain processes a vast amount of information daily, we constantly construct various narratives to make the world coherent and meaningful. Precisely because of this, narratives themselves ultimately shape reality.
For investors, predicting the future ups and downs of market prices requires understanding which "collective delusions" market participants collectively believe. The same company, with unchanged future cash flows, may receive drastically different valuation multiples simply because the market believes a different story.
And the easiest way to achieve a "valuation reset" for a previously dull, boring enterprise is to give it a new narrative that aligns with current market hotspots, making investors willing to chase it at any cost.
Is There an AI Bubble?
This leads to the core question of judging whether AI is currently in a bubble.
However, before discussing the AI bubble, it's more worthwhile to answer a more fundamental question: "What exactly are we investing in?" — to put it in terms of relationships, "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) — is it Technology or Real Estate?
The mainstream market narrative today is that this multi-trillion-dollar AI infrastructure build-out belongs to "technology," and therefore deserves high growth valuations.
But my view is exactly the opposite. AI CAPEX is essentially just another boring real estate investment. The only difference is that this time, data centers house not office buildings, but computing power. This computing power will ultimately give rise to silicon-based lifeforms, driving the development of human civilization, possibly even surpassing the significance of the railway revolution.
The reason we must distinguish between "real estate" and "computing power" is that the newly matured hedge fund managers, banks, private credit funds, and even governments mistakenly believe they are lending to tech giants like Apple, not providing real estate financing to Lehman Brothers.
I believe the ultimate reason for the AI bubble burst will be financial intermediaries overbuilding data centers, and all the supporting infrastructure required for data center construction, including energy, power, and everything needed for training and inferencing AI chips.
Therefore, the AI bubble is more like a 2008-style credit bubble, not a 2000 internet bubble-style earnings bubble.
During the 2000 internet bubble, most listed internet companies had little to no revenue, let alone profits, like Pets.com, so that bubble was essentially a "failure to deliver earnings" problem.
The 2008 financial crisis was different. What triggered the crisis was the slowing pace of US home price appreciation, which raised concerns among banks and financial institutions about the repayment ability of mortgage assets, so it was a credit crisis.
The AI bubble will follow similar logic. The real turning point will not be AI leading companies stopping profits, but a slowdown in data center construction growth, or hyperscalers lowering their future data center construction guidance.
Even if AI leading companies continue to generate massive profits, their forward valuation multiples will still contract due to declining growth expectations. The first to fall will be those AI companies with the most fragile credit profiles and the highest leverage.
Subsequently, these risks will rapidly spread to the balance sheets of highly leveraged financial institutions holding significant AI debt assets. Ultimately, governments will intervene again in the name of "national security" to ensure these over-leveraged AI companies and their supporting financial institutions don't collapse.
And this misallocated capital will ultimately flow into the crypto market... sending Bitcoin to the moon (to da moon) again.
The Credit Risk of AI CAPEX
Whenever someone suggests "AI is in a bubble," AI bulls almost always counter with "Jevons Paradox" as a rebuttal. Jevons argued that when the price of a commodity falls, its usage increases significantly, causing the overall market size to continue expanding, even exponentially.
If you believe AI capital expenditure itself represents computing power demand, then according to Jevons Paradox, indeed there's nothing to worry about. As computing costs keep falling, demand for AI token-consuming applications and AI Agents will grow exponentially, so lending to AI infrastructure is naturally a money-good business.
But I think this is actually a misreading of Jevons Paradox. To understand why Jevons Paradox doesn't mean all credit flowing to AI CAPEX will be repaid, let's look at what a hyperscaler is actually doing when building a data center.
Essentially, it's first undertaking a real estate development project. It builds a structure to house server racks, then procures the latest generation of semiconductor chips for AI model training and inference. And as industrial technology — especially semiconductor manufacturing — continuously advances, the floating-point operations per kilowatt-hour will continue to grow exponentially.
A few years from now, whether it's Nvidia, AMD, Intel, Huawei, or SMIC, they will all launch new-generation AI chips far more efficient than today's. At that point, the same data center, consuming less power, will be able to produce 1000x more intelligence.
This means two things can be true simultaneously: On one hand, physical infrastructure like AI data centers could become saturated with supply; on the other hand, AI token consumption could still grow exponentially.
So, the real question worth considering is: Do you want to hold the real estate business — what hyperscalers are doing today; or the AI application layer?
The common rebuttal from AI bulls is that hyperscalers are both landlords and tenants. They rely on massive cash flows from Web2.0 "selling attention" businesses to provide credit support for issuing debt to build data centers; meanwhile, they use their AI capabilities to sell the "apple of knowledge" from Eden to the world.
If you truly believe this story, then I hope you are holding their stocks, not their bonds. There's a reason bonds are called fixed income — the best outcome for creditors, no matter how successful the company becomes, is just getting their principal back plus some interest.
If Google successfully bets on AI and creates revolutionary products that transform human civilization, sending its stock price soaring, shareholders certainly deserve to cheer; but bondholders will still only get their principal.
Conversely, if Google ends up as a "data center landlord" renting out lots of depreciated Nvidia GPUs, but fails to earn enough revenue to service its debt and interest, creditors will suffer heavy losses. And how much a data center filled with obsolete chips is worth is highly questionable.
The CFOs of hyperscalers and Wall Street financiers aren't stupid. They know they're really in the real estate business. Therefore, they must find some "bag holders" who believe they're investing in high-tech, not real estate.
These bag holders include insurance companies under alternative asset management giants like Apollo, and ultimately the taxpayers of various countries who will foot the bill for the government's implicit guarantee of AI credit.
If you carefully read through those deliberately obscure financial statements, you'll find — the massive debt issued to finance AI CAPEX is almost all off balance sheet, with almost no clear connection to the core profitable businesses supporting stock valuations.
How we define AI CAPEX determines how we understand the entire AI investment cycle. It is this narrative that explains why capital misallocation occurs severely, and why the size of this bubble might surpass that of the railway bubble.
More importantly, since the AI bubble is a credit bubble, not an earnings bubble, when the crisis erupts, governments will inevitably step in to save the final bag holders who mistakenly bought traditional real estate debt thinking it was new-tech equity assets.
Don't think the AI bull market is over because of recent adjustments in the AI sector, especially in highly leveraged markets like South Korea. On the contrary. The truly crazy "blow-off top" might have just begun.
Just last week, the Fed had the chance to raise interest rates to combat inflation that remains above trend under various statistical measures, but it didn't. It chose to remain on hold, with even former Chairman Powell voting to keep rates unchanged.
So, what does it matter for forgotten crypto players stuck in sideways bear markets whether AI credit allocation is good or bad? It matters because it determines the future manner, reasons, and amount of money printing by various governments to fill the financial holes created by out-of-control AI CAPEX investment.
The rest of this article will develop this theory and explain why governments ultimately can only choose to print money to rescue the market.
As AI CAPEX growth slows while credit continues to expand, Bitcoin will find its bottom and start a long-term uptrend. When policymakers finally realize the AI GDP growth they pinned their hopes on is essentially just another ordinary real estate bubble, they will be forced to launch monetary easing on a scale even larger than the 2008 Global Financial Crisis.
And ultimately, this will push Bitcoin past $1 million, and even higher.
The Second Derivative Decides Everything
I always have to remind myself: "Investing doesn't really trade on growth itself, but on the acceleration of growth."
In other words, what we focus on is actually the second derivative — whether growth is accelerating or decelerating.
This aligns with intuition. When an asset is still in an accelerating growth phase, people keep weaving stories about its infinite future possibilities. Thus, you hear bold statements like, "I'd rather hold a hyperscaler until bankruptcy than miss the chance to build AGI."
However, all growth eventually enters a deceleration phase. The problem is, the price action of most assets often follows this pattern:
- Accelerating growth phase: Prices keep hitting new highs;
- Decelerating growth phase: Prices move sideways;
- Only when growth itself (the first derivative) turns negative do prices truly start falling.
No one can accurately predict how long the lag between growth starting to decelerate and actually turning negative will last, but many investors, myself included, subconsciously think — even if growth has started decelerating, asset prices can still rise indefinitely.
If the AI bubble is essentially a credit bubble, the importance of the second derivative is even more prominent. Because the entire society's willingness to keep financing AI CAPEX is based on an assumption — AI investment scale will forever keep accelerating.
Once this acceleration disappears, continuing to increase debt becomes increasingly dangerous, but the reality is, until you actually get punched in the face, no one knows when to stop.
Either wait for a financial crisis to erupt; or wait for "Kenny G" (Odaily note: likely a reference to Ken Griffin recently taking over distressed AI stock positions at low prices) to scoop up all your assets at the market bottom.
Therefore, even if investment growth has started slowing, credit scale often continues to expand. Only when AI CAPEX budgets actually start declining does the market reach that classic "Wile E. Coyote" moment — the character has already run off the cliff but only realizes he has no ground beneath him when he looks down, then instantly plummets.
Then, the market will start identifying who is over-leveraged from holding large amounts of junk AI CAPEX debt.
Let's apply this logic to the US subprime crisis. My favorite class in college studied US housing policy and the mortgage market. The professor had served as Deputy Secretary of Housing under the Clinton administration. Coincidentally, I took this class in spring 2008 — right when Bear Stearns collapsed, making it incredibly relevant.
The core message of the class was that the government, to achieve the social equity goal of "homeownership for all," kept encouraging more people to buy homes, leading to continuous credit expansion. However, by 2006, many first-time homebuyers were actually unable to afford the monthly payments after their mortgage rates reset. The only way they could keep paying was if home prices continued rising at an ever-increasing pace.
Of course, I'm still waiting for the government to deliver on my "40 acres and a mule." Might as well just print money to build houses.
The four-panel chart below shows:
- S&P 500 Index;
- US construction loans vs. construction activity;
- Case-Shiller US National Home Price Index.
By the end of 2005, US home price appreciation had started slowing, which coincided with the peak in actual construction investment spending (orange line in the first chart). However, real estate credit (purple line) kept flowing into the market until the stock market peaked and started showing mild corrections.
2006 to 2007 was essentially the "no man's land" before the crisis erupted. Home prices were still rising, but at a decelerating pace; then, the stock market peaked in mid-2007 (pink dashed line in the chart); the real "Wile E. Coyote moment" happened in August 2007 — three credit hedge funds under BNP Paribas collapsed; the crisis then spread, ultimately taking down Bear Stearns and Lehman Brothers in September 2008... By then, the S&P 500 had already fallen about 50% from its high.
What triggered the financial collapse was investors finally discovering who held those toxic "Frankenstein" financial derivatives. Ultimately, the government had to take over both the debt and equity of these institutions to prevent a new Great Depression.
This point is very important, as we'll return to this logic when discussing how governments will save the AI industry later.
The second chart is also worth noting. It shows that the starting point of capital misalignment was precisely when home price appreciation started slowing. If new credit was still being used to build more houses, the problem wouldn't be too severe. But if the entire system started relying on new debt to repay old debt, then risks began accumulating. The rising ratio of construction loans to construction investment spending perfectly illustrates this process.
Now, apply the same analytical framework to AI. The key variable here is the CAPEX spending plans of various companies.
The current market believes that real estate (here, AI) is technology; the more tech you invest in, the higher future profits will be. Therefore, the market rewards hyperscalers announcing increased capital expenditure budgets by pushing up their stock prices.
I expect the announced growth rate of AI CAPEX will start slowing in mid-to-late 2027, and by 2028, the market will clearly enter a "decelerating growth phase."
Simultaneously, a seemingly contradictory phenomenon will occur: while CAPEX growth starts declining, credit flowing into AI will still continue expanding. The reason is that lenders believe they are investing in technology, not real estate. Plus, with various governments constantly emphasizing the need to dominate the global AI race, continuing to finance anything related to AI CAPEX seems the most rational choice.
Thus, 2027 will become the "no man's land" similar to 2006-2007. The recent sharp correction in AI stocks is just a normal pullback within a bull market. The true peak of the AI bubble will appear next year.
After that, the market will instead start rewarding hyperscalers who are "first to exit the arms race" and proactively cut CAPEX budgets. Unlike the early bubble phase from 2022 to mid-2026, hyperscalers will find it increasingly difficult to rely on their own free cash flow to support AI investment in the future. They will have to depend more on issuing bonds and equity offerings to raise funds.
The pressure on balance sheets will also force management to seriously consider: "Is it really worthwhile to keep borrowing to build more data centers, just to house more chips that keep depreciating?"
At least for US hyperscalers, Chinese frontier AI models with lower prices but comparable performance will completely extinguish their "silicon deity" fantasies. After all, if two products are of the same quality, or just slightly inferior, most people will choose the cheaper one.
As the intelligence generated per kilowatt-hour by AI chips continues growing exponentially, and per-token costs keep falling under Chinese competitive pressure, a rational hyperscaler CFO won't continue worsening their balance sheet just to build more data centers.
Even if, according to Jevons Paradox, AI token demand eventually explodes, that growth won't come fast enough to offset the negative impact of the massive debt issued years earlier. Ultimately, the market will punish the worst credit participants first. Then, people will truly realize how much capital was wasted in this AI investment wave.
I cannot predict which hyperscaler will overplay its hand first, causing bond investors to collectively shout: "Oh shit!"
However, before discussing why banks continue lending to AI despite knowing the huge risks, take a look at the chart below. It shows the scale of committed CAPEX investment by major hyperscalers versus the cash on their books.
What underpins the entire AI bull market narrative is actually trillions of dollars in leverage. Among these companies, one will eventually fall from grace like the once market-beloved AI genius Leopold Aschenbrenner. The difference is, the ones rescuing them won't be the greedy, neurotic East Coast hedge fund managers of Wall Street, but the printing presses in the hands of Warsh and Treasury Secretary Bessent.
The Bank Credit Dilemma
Many believe the imminent slowdown in AI CAPEX growth means the AI bubble is nearing its end. If that were true, then why would banks keep lending?
The reasons are simple: First, because it's profitable; second, because the government wants them to; third, because they know even if the loans go bad, the government will bail them out.
Louis-Vincent Gave of Gavekal Research published a very interesting article last week. He argued that Warsh's interest rate policy actually follows a very simple logic — actively steepening the yield curve.
Doing this has two benefits. First, it makes bank lending more profitable; second, it helps gradually inflate away the massive US debt burden.
Ultimately, banks will keep creating new loans, i.e., creating new money. This new money will finance US re-industrialization and continue supporting AI build-out. This thinking aligns closely with Treasury Secretary Bessent's recent mentions of "Hamiltonian Economics."
Judging by all objective economic indicators, the Fed should have raised rates at its recent meeting, but it didn't. Instead, long-term Treasury yields subsequently rose rapidly.
- Odaily note: The 30-year US Treasury yield quickly rose after the Fed held steady.
Many view this as a policy mistake by the Fed, but from a bank's perspective, this is a heaven-sent gift.
The reason is simple. Banks can fund themselves at almost the Federal Funds Rate, and the Fed deliberately keeps this rate below nominal economic growth, even below actual inflation.
Then, banks lend these funds as long-term loans to AI data center developers, rare earth mining companies, defense manufacturers, etc. The steeper the yield curve, the higher the net interest margin banks can earn.
And as shown in the commercial and industrial loan chart below, the more incentive banks have to keep creating new money through lending.
- Odaily note: White line shows the 10-year Treasury yield minus the effective federal funds rate (reflecting yield curve steepness); Yellow line shows US commercial bank commercial & industrial loan balances.
From a political perspective, this is a sustainable Fed policy. Even though the Trump administration's DOJ investigated and even prosecuted some Fed governors (like Lisa Cook and Powell), these voting members still support keeping short-term rates at negative real levels.
That is to say, Warsh essentially has a "coalition of volunteers," including Trump-aligned people and those within the system who still suffer from "Trump Derangement Syndrome."
From a monetary policy perspective, the Fed's recent actions also allow Treasury Secretary Bessent to issue short-term T-Bills at yields below nominal economic growth. If the market cannot absorb the massive weekly T-Bill issuance, the RMP (Reserve Management Program) will print money to fill the demand gap.
To suppress those "disobedient," persistently rising long-term Treasury yields, Bessent can also implement Treasury buybacks — first issue short-term T-Bills monetized by the Fed, then use these funds to buy back 10-year or 30-year Treasuries, thereby lowering long-term rates.
Notably, Warsh, known for advocating shrinking the Fed's balance sheet, now shows no intention of restricting, let alone stopping, the RMP program's balance sheet expansion. It's all just Kabuki UFC theater performed on the White House lawn.
If you are a credit officer at a "Too Big To Fail" bank hoping for future promotions and raises, you will almost certainly approve loan applications from "critical industries" like AI, defense, etc.
The reason is simple. It both increases bank profits and aligns with Fed and Treasury policy direction. Even if the loans eventually blow up — and mathematically, the probability is quite high — the government will definitely quickly roll out "bazooka-sized" rescue packages.
There's almost no downside risk here. This is how "window guidance" actually operates in the US.
I believe the 2026 version of the "Treasury-Fed Accord" has already quietly happened, just not formally announced. Otherwise, how else would you define the current situation?
- The Fed maintains negative real interest rates;
- The Fed prints money to buy Treasury-issued T-Bills;
- The Treasury encourages banks to lend to critical industries;
- When loans go bad, the ruling government implicitly guarantees a bailout.
If this isn't fiscal and monetary policy coordination, I don't know what is. So, I am extremely bullish on the market now. The real massive money printing is far from over.
US Sovereign Wealth Fund
Now, let's use our imagination more boldly. What if the US government not only bails out banks after a crisis, but proactively buys AI company stocks when crisis signs emerge? After all, thought experiments with wild imagination are always interesting.
In fact, the Trump-era bailout of AI has already begun. Under the banners of "national security" and "US-China competition," the US government has started borrowing money and directly buying equity in so-called "critical industry" companies like rare earths and semiconductors.
This is essentially a dollar liquidity-increasing operation, also interpretable as Equity QE. Because these dollars, originally sitting in government accounts, are directly injected into financial markets.
Below are some examples where the US government uses borrowed funds from the CARES Act, CHIPS Act, and Defense Department budgets to directly hold equity in relevant companies.
Unfortunately, for us crypto investors whose wealth depends entirely on the scale of money printing, under the current legal framework, there's very little room left for the government to continue similar equity investments.
However, the Trump administration and Treasury Secretary Bessent have clearly shown an attitude that, if the law allows, they wouldn't hesitate to use borrowed money to bottom-fish AI stocks.
Thus, a new question arises: Is there a way to preemptively print money to buy AI stocks before a crisis erupts, without requiring Congressional approval?
The answer is: Yes! And this is precisely the most interesting part.
Under the Federal Reserve Act, under so-called "emergency and exigent circumstances," the Fed can directly print money and provide unlimited liquidity loans to special purpose vehicles (SPVs) established by the US Treasury.
During the 2008 financial crisis and the 2020 pandemic, the Treasury used the Exchange Stabilization Fund (ESF) to support first-loss equity, then had the Fed lend to this SPV to purchase various financial assets to stabilize markets.
Currently, the ESF account holds about $28 billion. Bessent could completely use these funds as initial capital for a new SPV, again citing national AI security as the reason.
Following past precedent, the Fed is typically willing to provide up to 10x leverage to an SPV. That means Bessent could theoretically leverage about $280 billion into unprofitable AI companies. Of course, compared to today's AI companies with market caps in the trillions, $280 billion is far from "heavy firepower."
So, can the scale be further expanded? For example, could the Treasury set up an SPV with no first-loss capital buffer at all, having the Fed directly lend unlimited amounts? Technically possible, but doing so means the Fed must withstand immense political pressure — because outsiders would view it as secretly conducting unlimited-scale equity quantitative easing.
So, does the Fed really care about political pressure?
The answer is, both yes and no. The new Chairman Warsh has consistently emphasized that AI will soon become a miracle for boosting US productivity. That is, ideologically, he himself believes the grand narrative painted by AI entrepreneurs.
If Trump tells him that to save Sam Altman and OpenAI, the government must directly enter the market to buy stocks. The reason is, there aren't enough retail investors willing to put real money into a frontier AI company not yet profitable; meanwhile, Dario Amodei's Anthropic is already profitable and has stronger model performance.
Then, Warsh will most likely do it without hesitation. Of course, procedurally, approving SPV loans still requires three other Fed governors to vote yes. But considering that in the recent FOMC meeting, including Cook, Powell, and others, have already sided with Warsh (voting to keep rates unchanged), if Warsh truly pushes the Fed down this path, I hardly see any substantive resistance.
After all, compared to theoretical concerns about whether to print money, personal stock portfolio returns are always more persuasive.
If the Treasury uses printed money to support upcoming AI star companies issuing new shares, it's actually cashing out 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 provide a backstop, this capital simply didn't exist. It's precisely the government's willingness to provide a bid for valuations lacking fundamental justification in the private markets that truly "monetizes" this paper wealth.
From an accounting perspective, the government gains two benefits.
First, as long as an AI company has government backing, investors will flock to it. After all, trading stocks alongside someone with a printing press almost always makes money, at least initially. Thus, this SPV's books will quickly accumulate massive unrealized gains. Trump could completely package these paper gains as government "profits," even claiming they could theoretically offset the fiscal deficit. If AI truly is the most important technological revolution in human history, then paper gains in the stock market could, in accounting terms, even "eliminate" the entire US fiscal deficit.
Second, the millionaires, billionaires, and even trillionaires thus created must pay federal and state capital gains taxes when selling stocks. This new tax revenue can also reduce the fiscal deficit, letting the government borrow less, and further claim that US debt-to-GDP is falling. At least initially, the bond market will believe this story, so Treasury yields will fall, and the market will reward this government "accounting magic."
However, I must emphasize, Trump did not invent the Philosopher's Stone. He's just kicking the can down the road, hoping the next administration — preferably Republican — will deal with it.
Why will this model inevitably lead to disaster? Let's do a simple thought experiment. Suppose you want to become a billionaire overnight without doing any work. So, you spend a few thousand dollars registering a company, issuing a total of 1 billion and 1 shares.
Then, you sell 1 share to your mother for $1. Since the latest transaction price is $1, on paper, your other 1 billion shares are worth $1 billion. Next, you take this "wealth" to a bank, hoping to borrow $100 million to buy mansions, Lamborghinis, and various luxuries. The bank will directly tell you: "Don't even think about it."
You'd be confused. From your perspective, the loan-to-value ratio is only 10%, seemingly low risk, but the bank's answer is simple:
"If we need to sell these stocks in the future to repay the loan, there's no market liquidity."
Apply this logic back to the AI SPV, it's the same. If the SPV has become the largest single shareholder of an AI company, and other investors buy stocks only because the government is in it, then once the government prepares to exit, there won't be any real buyers in the market. Not only that, when politicians — like Ro Khanna, Nancy Pelosi, etc. — start selling stocks, all investors will rush to sell before the government, so those paper gains originally intended to "offset national debt" will instantly vanish. Not only will they become real losses, they will further increase government debt.
Worse, the Treasury must still repay the money it originally borrowed from the Fed. So for this SPV, this is essentially an investment that can only be bought, not sold. The Fed can only keep rolling over the SPV's loan, ensuring a margin call is never triggered.
Ultimately, to sustain the entire system, the expansion of the Fed's balance sheet will become permanent.
However, this isn't a problem Trump needs to worry about. Because politically, he wins both ways — on one hand, unprofitable US AI companies get funds to continue competing with China; on the other, the paper "wealth" created by AI drives tax revenue growth and stimulates current economic activity.
Simultaneously, unrealized gains plus new tax revenue create the illusion that "US debt-to-GDP is falling." Thus, the market is willing to keep lending to the US government at lower rates.
The US government could do this now, preemptively preventing the AI bubble from bursting. Of course, it could also wait until AI CAPEX growth slows and the market starts a full-scale sell-off of AI stocks before stepping in to rescue.
Since the US government has already started directly buying corporate equity, why not buy more? Combining "bank window-guidance lending" and "government direct buying of AI stocks" could theoretically ensure an AI credit crisis never happens. At least not before the 2028 US presidential election.
Some might ask, after all this talk, if so much money has been printed from 2022 till now, why hasn't Bitcoin broken $126,000? Don't worry, the next part gives the answer.
When Will Bitcoin Bottom?
The bottom of the last cycle occurred after the market discovered that the white boy Sam Bankman-Fried stole FTX client funds, with CZ helping to facilitate this discovery.
Simultaneously, ChatGPT launched commercially, and the AI wave began.
Starting October 2023, the US liquidity environment changed. As funds from the overnight reverse repo facility (RRP) kept flowing out, dollar liquidity began increasing. Subsequently, bank credit and government borrowing also grew. Bitcoin thus rose, peaking in October 2025; but Bitcoin didn't continue rising, it only rose about 2x from its previous all-time high, because AI credit and AI stocks absorbed the new fiat liquidity.
As AI capital expenditure (CAPEX) expansion accelerated, swallowing all available fiat liquidity, Bitcoin — obvious in hindsight — fell 50%.
Mid-2026, the liquidity environment reversed. The growth rate of announced AI CAPEX for the next 18 months will start slowing, but bank and government credit channels are just starting to create dollars and funnel them to the AI sector.
If banks fail to fulfill their "patriotic duty" to keep providing credit, the government will strongly push them to lend to AI. If that still fails, the government will reduce banks' lending risk to AI through equity support to specific AI companies and offtake agreements similar to those for Intel and IBM.
Bitcoin will bottom during the early phase of this credit misallocation. The financialization process of AI — i.e., the amount of dollars and yuan chasing quality AI projects exceeding the scale of genuinely quality projects themselves — will ultimately cause capital misallocation.
Writing this article in late July 2026, I don't know at what price Bitcoin will ultimately bottom; perhaps the bottom has already appeared.
The market needs time to digest concerns over Strategy (formerly MicroStrategy) selling Bitcoin. Also, the market needs to find a new narrative. If Strategy cannot keep issuing stock, or find investors to buy its preferred shares and use these funds to keep buying Bitcoin, then why can Bitcoin still rise?
Perhaps Bitcoin will fluctuate between $60,000 and $70,000 for a while, possibly dipping to $50,000. However, during all this, AI capital waste will keep accelerating, laying the foundation for Bitcoin's bottoming and subsequent gradual rise.
If my view is correct — that the scale of valuable AI CAPEX projects is smaller than the credit flowing to "AI" — then Bitcoin's price will ultimately reflect this excess liquidity. This will help Bitcoin bottom, even if digital asset treasury (DAT) companies like Strategy can no longer buy Bitcoin in a Bitcoin-per-share accretive manner through stock and corporate bond markets.
I will keep observing several indicators to verify this logic:
- Whether AI CAPEX growth slows;
- Whether AI loan volume increases;
- Whether hyperscalers increase off-balance-sheet commitments.
If we enter the capital waste phase of this AI credit boom, then the next question is: What will regulators and governments do? Will they print money preemptively? Or, lacking political space, wait for the eventual crisis and then intervene with bailouts?
Fortunately, as long as we hold Bitcoin non-leveraged, we don't care when the bailout arrives. Because we know that, due to distorted government incentives, they will ultimately choose to print money to save the system. The AI CAPEX credit frenzy now already scales proportionally to GDP similar to the railway construction era. This means the scale of capital misallocation already exceeds the US subprime crisis.
Therefore, the future bailout scale will exceed the trillions printed by the Fed and global major central banks between 2009 and 2013. Bitcoin was born precisely in response to the "irresponsible bailing out of bankers" during the subprime crisis. If you think about it, that's an amazing thing. And this time, Bitcoin already exists, and it might achieve many people's dream — rising to $1 million or higher.
Given the current dismal state of crypto capital markets, it's not easy to imagine such a future. But in my view, this creates interesting asymmetric opportunities. Maelstrom has long held significant Bitcoin.
Besides Bitcoin, what could be the new narrative driving a large-cap token higher in the next six months? Ethereum is currently the most hated, most forgotten large-cap "shitcoin" in the market. It hasn't even broken its 2021 all-time high of $5,000, while most top-ten shitcoins by market cap already have.
In my view, the next narrative is that enterprise RWA (real-world asset) chains, like Robinhood's, will use customizable Ethereum Layer2s like Arbitrum. Ethereum will become the securities settlement layer for these chains. Therefore, even if the actual Gas revenue flowing to Ethereum is a small fraction of the entire system, ETH is still the shitcoin powering the "tokenization of everything."
I am a critic of RWA. Maelstrom often receives many garbage project fundraising pitches, with teams claiming they'll ride the asset tokenization wave. But on the other hand, traditional finance loves discussing: "In the future, all assets will be tokenized and run on some private or public chain."
I strongly believe that if this future arrives, these TradFi RWA projects must run on public blockchains. And Robinhood using Arbitrum to launch its own chain will lower the career risk for TradFi practitioners — they can replicate the same model, ultimately building Ethereum-based solutions.
This narrative is very strong. And ETH, as a shitcoin, is the second-largest crypto asset by market cap, existing since 2015, so it possesses Lindy effect second only to Bitcoin. Plus, Tom Lee's endorsement for institutional investor ETH allocation provides fund managers a way to bet on the tokenization of capital markets trend.
My rough year-end 2026 target for ETH is $5,000, roughly a 2.6x increase from current prices. I like this trade because I can deploy a sizable nominal position while accepting the risk that ETH might drop 75% one day due to some technical flaw — a very low probability.
Furthermore, ETH is extremely liquid. So even if it occupies a large proportion in Maelstrom's portfolio, I can still exit within minutes. Finally, I will also sell out-of-the-money puts to earn extra yield, while accepting the risk of buying ETH at a discount if it falls below the strike price.
The AI bubble once sucked liquidity from the crypto market, but that's over. As the market narrative gradually shifts from "invest in AI at any cost" to "what's my return on investment," and ultimately to "when do I get my principal back"... governments that bet entire economic policies on AI will start worrying — maybe this bubble could really burst.
To avoid admitting their mistakes and prevent this outcome, they will engage in massive capital misallocation, and capital misallocation of this scale will ultimately create a crypto bull run unlike any we've seen since 2021.
