Arthur Hayes: Dilution Is Poison — How Do AI Giants' IPOs Trigger a Credit Freeze and Crypto Liquidity Blowback?

Author: Arthur Hayes

Compiled by: TechFlow

Amid the intertwining of tech euphoria and capital bubbles, markets are focusing their attention on the super IPOs expected from space and AI leaders SpaceX, along with Anthropic and OpenAI. However, veteran macro trader Arthur Hayes presents a brutal, near-mathematical-reversal logic in this article: current AI unicorns rely heavily on going public to qualify for corporate bond issuance to support astronomical compute Capex; yet, due to extremely low initial float and subsequent massive insider unlock sell-offs, these stocks will face an extremely brutal dilution storm in the secondary market.

Hayes draws lessons from the historical tragedy of early Zcash investors in 2016, meticulously calculating Anthropic's staggering 83% dilution rate and OpenAI's fragile cash flow situation. He points out that once shares break below their IPO price triggering a reversal in market sentiment, AI corporate bond credit spreads will spike abruptly or even fall into junk status, thereby cutting off refinancing channels for the upstream supply chain.

More critically, this credit cooling is directly impacting the balance sheets of US commercial banks. As bank credit and securities financing contract, USD fiat liquidity will be systematically drained, exerting a transmissive suppression on crypto assets. Hayes clearly outlines defensive trading strategies before the ultimate global central bank printing press intervenes to save the day, as well as the structural dividends brought to the on-chain autonomous Agent ecosystem by the oversupply of compute spot market.

(All views expressed in this article are solely those of the author, should not be taken as investment decision basis, nor should they be considered recommendations or advice for participating in investment trades.)

I opened my eyes. It was the morning of Saturday, October 28, 2016, during Tokyo's Halloween weekend.

Our group gathered at the Wyndham Guy Hotel to board the flight, becoming yet another band of foreign travelers spending a raucous bachelor weekend in Tokyo. When I stumbled back safely to my comfortable king bed at the hotel in Roppongi, I felt immense relief.

The night before, we wandered around Minato Ward for a long time. My current Chief of Staff and I took a comical photo at a "reputable" venue, both utterly drunk, posing like characters from the movie Lethal Weapon poster.

That Friday late at night, the Zcash genesis block was born, kickstarting one of the craziest trading days in cryptocurrency history.

Similar to Bitcoin, purchasing physical ZEC required buying it directly from miners. Back then, as now, the privacy narrative was incredibly hot, and everyone desperately wanted a share of the first scalable private digital cash network.

However, the core issue was that the initial supply was completely zero. Circulation could only slowly increase block by block as miners continuously mined and sold ZEC to cover operating costs.

Poloniex was the platform dominating altcoin trading at the time. My memories of trading on Polo and writing code against its incredibly terrible API are still vivid today.

If you were lucky enough to get a response from customer service, it felt like listening to God's direct commandments. Years later, Circle acquired Poloniex at a hugely expensive premium, only to run it terribly until Justin Sun cleaned up the mess. But that's beside the point. Players, always remember: price is what you pay, value is what you get.

Since Poloniex was the only place to buy ZEC at the time, when tiny amounts of ZEC started listing sporadically, impatient speculators swept them up at absurdly outrageous prices.

As the chart shows, the vast majority of players who bought ZEC during the first few days and weeks of trading remain deeply underwater and at a severe loss to this day.

Chart: Historical price trend of Zcash since its launch in 2016

Traders might counter that ZEC's performance must be adjusted for circulating supply to assess its true results.

Although ZEC's total market cap is indeed at historical highs, as a trader, you never trade total market cap; you trade the actual unit price of each token.

This is a key difference many investors fail to fully grasp. As token supply rapidly expands, even if total market cap rises, the unit price of each token or stock will be severely suppressed, ultimately leading to abysmal returns on investment.

Turning our gaze to the AI sector: what do the public offering dynamics of this combo—SpaceX plus the upcoming IPOs of Anthropic and OpenAI—portend? Great riches or poverty? The answer is obvious: retail traders will only face poverty!

I believe, much like early ZEC investors, if you buy these companies' IPO stock at the offering price, you are destined to suffer heavy losses as insiders unlock and mass-sell.

This will deliver a heavy blow to the entire macro environment, because the entire AI trade rests entirely on market vibes. Whether you believe in the lofty future profitability and ubiquitous penetration rate of AI determines whether you're willing to lend funding to the industry.

And all these vibes stem from the single objective metric measuring AI success: whether stock prices are rising or falling. So, setting aside whether CEOs of SpaceX, Anthropic, and OpenAI are deceiving or committing financial fraud, I know full well that once shares break below IPO price due to massive insider unlock sell-offs, the market will assume they are all lying.

After presenting these brutal mathematical calculations that will utterly shatter investor illusions, I will detail macro trading strategies amid deteriorating global liquidity, right up until the ultimate rescue arrives.

SpaceX and Compute Heavyweights

The three AI companies we focus on are SpaceX, Anthropic, and OpenAI. This is because they consume the vast majority of compute resources in the process of producing intelligent tokens.

Once their demand for compute slows down, the entire AI macro narrative will collapse. Let's delve into the float supply dynamics of these companies' stocks, examining their current and future price action. But before crunching the numbers, I want to emphasize why stock prices are so crucial to the AI narrative.

These companies must all invest hundreds of billions of dollars in massive capital expenditures (Capex) to drive the godlike silicon-based intelligence they create.

Because they were previously non-public, issuing corporate bonds at the required scale was virtually impossible. They had to rely on a publicly listed hyperscale cloud provider or semiconductor giant for endorsement or debt guarantees. Compared to the massive debt issued, their equity financing was negligible.

However, once these labs go public, they gain direct qualification to issue corporate bonds.

Although SpaceX is not a pure AI lab, being public gave it an investment-grade credit rating and enabled it to successfully issue billions in debt since its IPO. As these companies tap into institutional capital pools, credit spreads narrow and issuance scales grow.

Whether institutional investors are willing to accept lower yields depends entirely on their expectations of the AI labs' future viability.

Observing their publicly traded stock prices is the most direct way to gauge whether the market believes in the AI narrative and the profitability of specific labs.

If stock prices fall—and given the pressure from massive pending sell orders, this is almost mathematically inevitable—debt investors would still demand higher yield premiums if they're willing to inject capital.

It is precisely this reflexive logic that drives Elon Musk, Dario Amodei, and Sam Altman to tirelessly push their respective companies toward going public.

They will use every legal means to accelerate listings because without going public, their heavy-capital businesses dependent on high debt will face complete suffocation.

SpaceX Equity and Unlock Model

The chart below compares the percentage change in SpaceX stock price from its IPO date to the present with the change in the total number of shares issued.

It is evident that since going public, its stock price has remained in sideways consolidation, while free float increased significantly by 50%.

Chart: Comparison of SpaceX Price Movement and Total Shares Issued

From now until the end of June 2027, approximately an additional 8.6 billion shares are expected to be unlocked, meaning the float will face another ~50% dilution, with the vast majority of incremental shares originating from Elon's personal holdings vesting.

Additionally, if the rumored SpaceX-Tesla merger transaction materializes, it could subject Class A common shareholders to nearly an extra 50% equity dilution.

The only thing that can save bulls is the miracle of the company unimaginably generating massive free cash flow.

Achieving accounting profit alone is far from enough; Elon must also clearly guide the market that these cash flows will be used for share buybacks or dividend payouts.

If this isn't achieved, facing certain colossal dilution, investors have absolutely no reasonable justification for holding the stock.

Having analyzed the first, let's look at the pure AI representative: Anthropic.

Anthropic Equity and Unlock Model

Anthropic is a pure frontier AI lab. It focuses on training closed-source large models and sells smart access externally on a per-token billing basis.

This is a one-time consumption business model. Historical past expenditure on tokens does not guarantee continuous future consumption. This is the key insight many investors fail to grasp. They wrongly try to apply the SaaS industry's Annual Recurring Revenue (ARR) metric, when in reality Anthropic should be viewed more as an oil and gas company.

Setting aside profitability quality, what kind of stock supply landscape faces its upcoming IPO?

After meticulously studying the recently filed and updated S-1 prospectus (frankly completed by my AI Agent), I calculated that investors participating in the initial IPO will face a staggering 83% equity dilution by June 2027.

Speculators, please weigh this number carefully: a massive 83% dilution! It is truly jaw-dropping.

By June next year, Anthropic is absolutely unlikely to be profitable; even if it somehow turns a profit, the S-1 prospectus explicitly states in black and white that there are no dividend payout plans in the near term.

All available cash flow will be fully invested into the approximate $900 billion in compute spend listed in the prospectus and future potential data center construction.

Driven solely by this crazy dilution timeline, mathematically speaking, there is virtually no possibility for its stock price to rise between the listing date and June 2027.

However, investors often blindly trust in miracles when faced with Anthropic. Assuming its IPO market cap is $1.5 trillion, to keep the stock price flat without dropping, the entire market would need to inject about $1.245 trillion in new external capital to completely absorb the potential insider sell-off wave.

This is not absolutely impossible, but the probability of it occurring is extremely slim.

Next, let's examine the most worrying among them: OpenAI.

OpenAI Valuation Dilemma

Before investors draw harsh lessons from the continuously bleeding cash flow AI super IPOs, assuming OpenAI manages to list first, its performance is doomed to be the worst among the three.

Compared to Anthropic, OpenAI faces a more severe cash burn rate, inferior product experience, and higher customer concentration risk. But even as the bottom runner, it still requires colossal capital to build its godlike silicon-based infrastructure.

According to latest disclosures, OpenAI is currently attempting to raise $20 billion at a $1.25 trillion valuation, already down from the previously rumored $1.5 trillion valuation and $30 billion target. If the UK Financial Times report holds true, this round faces heavy resistance.

If Anthropic successfully completes its IPO with a sturdy post-listing stock price, OpenAI's passive situation might see brief relief.

Although OpenAI's official S-1 filing hasn't been seen yet, it is entirely reasonable to infer that its stock issuance structure will fully replicate the model of SpaceX and Anthropic.

They will offer a low-float portion of about 5% for public sale, and allow massive employee and early VC insider cash-out selling within the following six months.

Given the prevalent blind valuation arms race among Silicon Valley giants, Sam Altman will likely strive to price OpenAI around $1.5 trillion to compete toe-to-toe with former rival Dario.

Assuming its IPO wraps up in January, by June 2027 it will face the exact same scenario as Anthropic: needing hundreds of billions to trillions in new external liquidity just to barely maintain its stock price without breaking below IPO.

Sideways Means Down: The Mathematical Limits of Capital Absorption

As I stated earlier, the core purpose for these companies choosing an IPO is to open corporate bond financing channels, thereby freeing themselves from reliance on guarantees from hyperscale cloud providers and chip giants to independently raise expansion capital.

But to inject confidence into potential bond buyers, stock prices cannot merely hold steady; they require sustained strong upward trends. Yet facts prove that merely maintaining sideways movement demands absorbing extremely vast volumes of external market liquidity.

The most likely script plays out thus: secondary market investors are not as stupid as these CEOs imagine, and absolutely will not resign themselves to endlessly absorbing the colossal chips thrown by insiders.

Therefore, even if revenue barely meets expectations (accounting profit is out of the question altogether), the continuous surge of massive supply pressure will gradually erode stock prices, causing them to slide and bleed steadily.

Once these companies lose external support and face capital markets independently, how wide will the spreads of their issued corporate bonds widen relative to the average high-yield bond spread?

SpaceX will almost certainly be unable to maintain its investment-grade credit rating, while Anthropic and OpenAI will inevitably meet the fate of being relegated to junk bonds.

The core crisis lies in the fact that once the capital market sees the true financing cost at the final compute consumption terminal, all participants upstream in the industrial chain will face the grim dilemma of being forced to issue Capex bonds at extremely punitive interest rates.

Thereby, the negative reflexive cycle will completely shatter the last line of defense for the AI stock bull narrative. The wider the credit spreads, the higher the interest expenses for infrastructure construction, which in turn swallows up all potential future profit margins.

Too Late: Credit Ratings and Corporate Bond Freeze

Those blindly bullish AI investors must sober up. I simply don't care whether AI is practical now, whether revenue will beat estimates, or fabricated enterprise adoption statistics cooked up by some institutions.

Because today, through real-time stock prices, I can intuitively see the authentic market's pricing attitude toward SpaceX, Anthropic, and OpenAI, which is akin to an apocalyptic disaster for the bull narrative.

The root of the disaster lies in the fact that their stock supply dumped onto the market is too massive that, mathematically, there is absolutely no capacity for the stock price to sustain an upward trend.

Bond investors will closely reference the sluggish performance of the stock market. Even if they remain willing to participate in subscriptions, they will demand exorbitantly strict excess yield compensation.

Not only that, bond covenant terms will become extremely stringent, such as mandating large cash buffers on the balance sheet, or triggering strict limitations on early debt repayment.

In short, by the time the so-called savior god finally arrives, the believers have long dispersed and fled the church.

This is undoubtedly major bearish news for all holders of AI-related debt. If the credit market freezes completely due to being flooded with massive junk-grade high-yield debt, the entire vertical industrial chain will plunge into liquidity shock, followed immediately by a downgrade storm from rating agencies.

Bank Balance Sheet Contraction: The Transmission Path from AI to Crypto

As crypto investors, why must we closely monitor the bursting of the AI bubble and the crashing stock prices of the big three? The core reason is that as AI credit spreads widen and bond issuance stagnates, fiat liquidity (especially USD liquidity) will undergo systematic contraction.

Chart: Trend of US Commercial Bank Balance Sheet Size Changes

The liquidity source driving our held crypto assets to continue rising essentially can only come from Federal Reserve or US commercial bank balance sheet expansion.

But at least at this stage, the Fed is effectively idle. The core decision-makers, including Kevin Warsh and Scott Bessent, consistently prefer having the commercial bank system generate credit rather than the Fed injecting liquidity directly.

It is precisely the phased expansion of US commercial bank balance sheets that provided crucial liquidity support for the crypto rebound that kicked off in August. However, as of last week, the size of US commercial bank balance sheets has already seen a clear month-on-month contraction.

Although this is merely a single period's data point, under the shadow of widening AI credit market spreads, its latent risks are enough to warrant vigilance.

The expansion of commercial bank balance sheets previously greatly fueled the AI bubble: on the lending side, banks opened credit channels wide for various AI and affiliated enterprises; on the securities financing side, banks continuously provided leverage financing for secondary market speculators; on the proprietary trading side, banks themselves were also heavily buying tech stocks and bonds.

From a global perspective, unless you're in the highly profitable defense industry, almost all capital is betting on the AI bubble. If the downtrend in the benchmark companies' stock prices bruises AI vibes and causes the credit market to cool and freeze, the most optimistic outcome for commercial bank balance sheet size is merely sideways consolidation, or more likely, entering a contraction trajectory.

This is precisely the critical link in the transmission of risk from AI to the crypto market.

Piercing the Camouflage: Maintaining Patience Before the Ultimate Rescue

Ultimately, the core issue does not lie in whether these companies are squandering investor funds to construct silicon-based gods.

The truly fatal hidden danger is that the secondary market liquidity price, mathematically guaranteed to drop due to colossal equity dilution, has already become a catalyst for clarity known to the entire market.

Faced with real-time plunging stock price trends, the market's belief that AI assets only go up will be thoroughly shattered. Bleak market trends will force capital markets to re-examine the absurd assumptions previously built on the AI macro narrative.

Regardless of whether these labs can eventually achieve accounting profits under GAAP, regardless of whether AI ultimately evolves into a general-purpose technology, the undeniable objective fact right before us is: the sole objective thermometer measuring AI commercial success is continuously plummeting downward to the right.

If these unicorns can continue evading the strict scrutiny of the public markets for the next few years until the technology truly generates positive cash flow, they will certainly choose to do so.

But reality is cruel: without the backing of the corporate bond market, they absolutely cannot raise the hundreds of billions or even trillions in massive capital needed to keep burning cash. To step into the arena of debt financing, going public is an inescapable rigid threshold.

Their time is completely exhausted. Facing the high-pressure scrutiny and transparency of public markets, any speculative tricks relying on bluster will be laid bare.

Before a dramatic reversal occurs in fiat liquidity, Maelstrom will choose to withdraw to the sidelines and patiently wait, observing the collapse and disintegration of super AI IPOs under the crushing weight of massive dilution.

The ultimate rescue will inevitably arrive. The core question is merely when and in what manner central banks will intervene, which we will explore further in subsequent articles. Reversing the hawkish policy stance of the Fed and other central banks often requires the actual sounding of financial crisis alarms.

The AI bubble tugs at the lifelines of the global economy: from underlying data center construction to raw material exports and advanced semiconductor manufacturing, the global industrial chain is deeply bound to AI development. Without this infrastructure wave, the global economy will plunge into severe recession.

Therefore, major central banks and fiscal regulators will ultimately be forced to reopen the printing doors to contain the crisis. At that time, Bitcoin and crypto asset holders will usher in an extraordinarily glorious liquidity feast, provided you can maintain ample patience in the present.

During the observation phase, as more data centers gradually come online, compute will pour continuously into the spot market. Even if training and reasoning large models show negative returns on a unit economics basis, Anthropic and OpenAI will have no choice but to grit their teeth and continuously purchase compute.

The sharp decline in compute prices will provide an extremely cheap raw material foundation for the exponential explosion of AI Agent autonomous intelligent workflows. Ecosystems such as decentralized compute networks and Genlayer, the AI arbitration layer project incubated and guided by Maelstrom, will significantly benefit from this. Investors can also seize the tokenized shares of the proxy economic adjudication layer during its upcoming TGE.

May all comrades stay sharp in the trenches and calmly weather the storms.

(1) ZEC: Zcash privacy cryptocurrency ticker/code; (2) ARR: Annual Recurring Revenue.