Original author: Jia Liu

Low circulation, big narrative, and high market value are becoming the common characteristics of this round of financial market speculation.

Less than half a year after Zhipu rang the bell on the Hong Kong Stock Exchange, its stock price once rose 25 times. But if you look at its share structure, you will find a more critical, but easier to ignore, number: in the early days of listing, only about 17.35 million shares of Zhipu could be freely traded on the market, accounting for less than 4% of the total share capital. For a company with a market value of one trillion Hong Kong dollars, the chip pool for daily transactions is actually only 30 to 40 billion Hong Kong dollars.

This is a typical but not the only case, and can even be said to be the epitome of this round of market play.

SpaceX went public more than ten days ago with a valuation of US$1.77 trillion and only 4.3% of its publicly traded shares. In order to cooperate with its listing, Nasdaq directly abolished the 10% minimum public shareholding threshold that had been implemented for decades. SPCX's market capitalization peaked at more than 2 trillion US dollars, but its daily trading volume was only about 100 million US dollars.

Cerebras, an American AI chip company, sold only about 15% of its issued shares during its IPO in May, and the price rose to more than twice the issue price on the first day. Figma, the issuance and sale of old shares combined accounted for less than 10% of the total share capital, and rose 250% on the first day.

Low circulation, big narrative, high market value. The structure that the crypto market has been playing with for several years is now being replicated across the board by traditional stock markets. Similar structures appeared in U.S. stocks, Hong Kong stocks, and A-shares at the same time, and the narrative extended from AI, chips, and large models to stablecoins.

The era of pricing based on financial reports is over again

In February 2000, a sock puppet dog appeared in a Super Bowl commercial. That's a 30-second ad spot that Pets.com paid $1.2 million for. At the time, it had annual revenue of less than $6 million and losses of more than $60 million. The company liquidated nine months later, and the sock puppet became the dot-com bubble's most iconic tombstone.

The market lessons of that generation have been written into almost all investment textbooks: valuation without income support is a bubble, and narrative cannot replace financial reporting.

This lesson dominated the market for the next two decades. DCF, PE, PEG, discounted free cash flow, and pricing methods based on financial report data have become orthodox. Buffett was deified again after the 2008 financial crisis. "Buy without reading financial reports" has become synonymous with speculation.

But when we look at the new technology track from 2025 to 2026 today, we will find a fact: the most popular companies in these industries are actually losing money.

For example, CoreWeave, an AI computing infrastructure company invested by NVIDIA, will have revenue of US$16 million in 2022 and US$5.1 billion in 2025, an increase of more than 300 times in three years. Revenue grew at an impressive pace, but net losses also widened from $31 million to $1.2 billion. In the first quarter of 2026, the company had revenue of $2.1 billion, a net loss of $740 million, and a debt-to-equity ratio of 10.7. By traditional bank credit standards, such a balance sheet is not healthy. But after going public, its stock price rose by 190%.

The situation is similar with Nebius. This company was formerly known as Yandex in Russia. After splitting, it turned to AI cloud services. Revenue in the first quarter of 2026 was $399 million, a year-over-year increase of 684%, but the adjusted net loss was still $100 million. Its shares are up more than 510% in the past 12 months.

Turn your attention back to the Chinese market.

Zhipu’s full-year revenue in 2025 was 724 million yuan, approximately US$100 million, but its net loss was 3.182 billion yuan, which was 4.4 times its revenue. In other words, for every 1 yuan it makes, it spends far more than 1 yuan on computing power and research and development. AI Hong Kong stock MiniMax, which was listed in the same batch of IPOs, rose 109% on the first day, and subsequently rose by more than 700%. The full-year revenue was US$79.038 million, about 560 million yuan, which was less than Zhipu.

Similarly, Hong Kong-listed GPU company Biren Technology, A-share domestic GPU Muxi Shares, and Science and Technology Innovation Board Moore Thread increased by 120%, 693%, and 425% respectively on the first day of listing. These new stocks with astonishing gains also suffered serious losses or no profits.

If you look at these companies using PE, many of them don’t even have calculation prerequisites because profits are negative. Using PS, the Wisdom Spectrum is more than 1,200 times, and SpaceX is about 95 times. Using DCF, if the discount rate and terminal growth rate change slightly, the conclusion may change from 100 billion to 10 billion. The sensitivity of the model is so high that it loses its guiding significance. Damodaran, the author of the DCF textbook, himself valued SpaceX at US$1.2 trillion, which is 30% lower than the IPO pricing. He himself admitted that when dealing with this generation of IPOs, parameter fine-tuning will lead to dramatic fluctuations in results.

Some people will say that PE was not considered a factor in the early days of the Internet, and Amazon took 20 years of losses before making a profit. This is nothing new. Yes, but there's a key difference between this round and the dot-com era: The market isn't even pricing in proxy metrics for PE now, but is trading on pure narrative.

Although investors in the Internet era do not look at PE, they look at user growth, GMV, and page views. In essence, they are still using a set of quantifiable intermediate indicators to anchor valuations. Today’s AI companies also have indicators such as ARR, but ARR cannot explain Zhipu’s price-to-sales ratio of 1,200 times. The hype around the supply chain has long since broken away from the pull of financial fundamentals, pricing all expectations for the next three to five years into the present.

The old pricing framework begins to fail in the face of a new class of assets. Financial markets around the world and the investment logic of investors have also undergone tremendous changes.

Model weights, algorithm capabilities, developer ecology, and computing power scheduling capabilities are the real core assets of AI companies, but none of them can be written into the balance sheet. The programming capabilities of GLM-5.2 made Vercel CEO say "almost shocked", which will not be reflected in Zhipu's profit and loss statement. CoreWeave is sitting on a $100 billion order backlog, but that doesn't change the fact that it posted a net loss for the quarter. Nvidia's GPU is called the oil of the AI ​​era, and the pricing of oil has never only looked at the current quarter's production, but also based on reserves, demand curves and geopolitics.

The core assumption of traditional pricing methods is that future cash flows can be extrapolated from historical financial data. This assumption is very useful in industries such as consumer goods, finance, and real estate.

But the revenue curve of an AI company is not linearly extrapolated. It depends on the jump in model capabilities, the network effect of the open source ecosystem, and the sudden switch of policies and industry cycles. After the release of GLM-5.2, the narrative status of Zhipu can change overnight; Llama's open source has rapidly amplified Meta's AI influence; the US's restrictions on Chinese chips have turned Bi Ren and Mu Xi from marginal companies to "domestic alternative leaders." These variables are difficult to write into any financial model in advance.

At the same time, the market’s tolerance for narrative dominance is also increasing, because in the past few years, people who believe in narrative have really made money.

Those who bought Nvidia without reading the financial report at the beginning of 2023 made ten times the profit. At the beginning of 2026, people who bought Wisdom without looking at financial reports made 24 times their profits. When a "wrong" method continues to produce "correct" results, the market will correct its own methodology rather than correct the results.

The money that supports high market value is actually not much

One of Nasdaq's own studies looked back at data from 1980 to 2020: In the 1980s, the average float of U.S. stock IPOs was about 30% of the total equity. By 2020, that number had dropped to about 20%.

J.P. Morgan gave a more macro figure in its June 2026 report: the new shares issued in the IPO, plus the shares of early investors allowed to be sold after the ban was lifted, only accounted for about 1% of the total market value of the entire market.

IPO floats are getting smaller and smaller. This is a trend that has lasted for almost thirty years.

Nasdaq also found that there is a clear inverse relationship between circulation and first-day gains. In those years when the circulation volume is smaller, the first-day increase will be greater.

The same characteristics can also be seen in the U.S. stock IPO samples compiled by us from 2024 to 2026. Defining low liquidity as "current circulation/total share capital is less than 30%", among the samples for which first-week performance can be calculated, low-float IPOs accounted for 67.4% of the first-day gains, 65.2% of the gains on the third trading day, and 63.6% of the gains on the fifth trading day.

The corresponding proportions of non-low-trading IPOs are only 47.9%, 48.9%, and 49.6%.

With fewer chips to buy, the same buying momentum will be greater and the price elasticity will be stronger.

The reason is simple. The same buying order of 1 billion will be a wave if it hits the circulation of 20 billion, and it will be a tsunami if it hits the circulation of 3 billion. The shrinkage of circulation volume from 20% to 3% is not a linear change, but a qualitative change in price elasticity.

Newly listed companies are increasingly inclined to low circulation, because this is the result of maximizing the interests of all parties.

Let’s look at the founder first. The smaller the circulation plate, the more stable the control. SpaceX’s Musk controls about 85% of the voting power through Class B shares, and a float of 4.3% on the public market means outside investors have virtually no governance influence. He can serve as CEO, CTO and chairman at the same time, he can merge xAI into SpaceX without shareholder approval, and he can take the company's strategic direction completely into his own hands. The smaller the float, the weaker the voice of external shareholders, and the greater the freedom of founders.

Scarcity also directly drives up market capitalization figures. A company's market capitalization is not determined by its total shares, but by multiplying the price of its last trade by its total equity. If only 3% of the chips are trading, and these 3% are chased up to a ridiculous price, the market value of the entire company will be calculated based on this price.

The book value of the 97% of untraded stocks in the hands of founders and early shareholders has all inflated. This expanded market value can be used to raise funds, make mergers and acquisitions currency, and attract talents. SpaceX went public with a valuation of $1.77 trillion. This number will appear in all job postings and will appear on the table of all cooperation negotiations.

This phenomenon is not exclusive to small-cap stocks.

Figma (FIG) is a collaborative design software platform with only 2.36% of the chips in circulation. It rose 250% on the first day, 168.48% on the third day, and 173.7% in a week.

Circle (CRCL) is the stable currency and blockchain financial infrastructure company behind USDC. It has 13.68% of the listed chips, rising 168.48% on the first day, rising 271.77% on the third day, and rising 278.06% in a week.

Bullish (BLSH) is a digital asset trading platform and market infrastructure company, with 19.78% of listed chips in circulation, an increase of 83.78% on the first day, an increase of 87.95% on the third day, and an increase of 60.84% in a week.

Cerebras (CBRS) is an AI computing power infrastructure company with 13.66% of listed chips, rising 68.15% on the first day, 60.35% on the third day, and 57.13% in a week.

Look at investment banks. IPO "first-day gain" is the core indicator for measuring underwriting success. Media headlines, client reviews, and the reputation of investment banks are all linked to this number. The smaller the circulation, the easier it is to achieve the first day's increase. Goldman Sachs helped SpaceX design a 4.3% float, which rose 19% on the first day. Everyone said it was a great IPO. If the circulating supply is 20%, and the same size of buying orders is dispersed into five times the chips, it may only rise by 4%, and the media headlines will be completely different.

The incentive structure of investment banks is naturally biased towards low circulation - the smaller the circulation, the better the first-day increase, and the greater the reputation of the investment bank.

Then there are cornerstone investors. The cornerstone system of Hong Kong stocks is essentially a transaction: "I will help you lock in the chips, and you will guarantee distribution for me." The benefit of cornerstone investors is to get a certain IPO share (without worrying about being reduced or drawn), but the price is that they cannot sell for 6 months. But this price often turns into a reward - because the cornerstone has locked up most of the circulating market, there are very few chips left for trading, and the stock price can easily be pushed up.

When the ban is lifted after 6 months, if the stock price has risen several times due to low circulation, Cornerstone's return will far exceed that of a normal IPO. The cornerstone system ties together "helping the company lock up chips" and "making more money for oneself", and the interests of both parties are completely consistent.

The 11 cornerstones of Zhipu (Gaoyi Assets, Taikang Life, GF Fund, etc.) took away 70% of the few circulating shares, resulting in less than 4% of the final circulating shares. All locked down for 6 months. While they are helping Zhipu lock in circulation, they are also helping themselves create a scarcity premium.

So we may even see an institutional turning point from the Nasdaq trading platform, the abolition of the 10% minimum public shareholding threshold.

This rule existed for decades. A listed company must have at least 10% of its shares in public hands to ensure sufficient liquidity in the market and protect the interests of public investors. The S&P 500 is more stringent, requiring that the public shareholding ratio of constituent stocks should not fall below a certain level. MSCI requires 15%. The Russell Series requires 5%.

The effect of this precedent is far-reaching. If Nasdaq can do away with the 10% threshold for SpaceX, what obstacles remain for the next company that wants to go public with 3% float? If the largest trading platform in the United States decides that low liquidity is acceptable, will other trading platforms follow? The cornerstone system of the Hong Kong Stock Exchange already allows most IPO chips to be locked up. If Nasdaq also lets it go, will there be a global competition: which trading platform is more friendly to low circulation, so as to attract the best IPO targets?

First-level investment, second-level hedging, the stock market begins to reproduce the old routine of the currency circle

In the 1990s, as the options market slowly matured, zero-cost collars became standard equipment for the wealthy. You hold a stock, buy a put option to protect on the downside (costing money), and sell a call option to earn back the cost (collect money). The two sides cancel each other out, locking in a price range without spending any money. In the late 1990s, Michael Dell used variable prepaid forward contracts to cash out part of his Dell shares. This did not trigger taxes or reduce the number of positions, but he received cash in advance.

But it used to be used by a few super-rich people and founders. Now after SpaceX went public, wealth management companies have publicly promoted this plan to thousands of employees. The scale is completely different. Wealth managers such as Bernstein and Mercer are now directly issuing guides to teach SpaceX employees how to make collars. This level of popularity has never been seen before.

Bernstein’s report contains a very sobering set of data. They looked back at all U.S. IPOs that raised more than $50 million over the past decade and found that six months after the lock-up period ended, the median return was a decline of about 10%. One in 10 IPOs fell by at least 62% within six months of lifting the ban. The conclusion is straightforward: if you are a SpaceX employee and hold locked shares, statistically speaking, when you can sell them, the price will most likely be lower than now. So you should use derivatives to lock in gains before the ban is lifted.

Michael Burry, the man who made hundreds of millions of dollars by shorting the U.S. subprime mortgage market in 2008, publicly said after SpaceX went public that he had studied put options and wanted to short them, but found that the price was too expensive to bear, and ultimately neither went long nor short. Even the "big short" thinks the cost of shorting is too high, which shows that too many people are trying the same transaction, raising the option price to an absurd position.

In addition to the above methods, whether it is collar, pair trading or unlocking arbitrage, they all have a common premise: the company has been listed. The stock already has a public market price, the options chain has come out, and the short-selling mechanism has been established.

But if you are someone who has been in the crypto market for a few years, you will find that these changes are not new at all.

In 2024, Binance Research, the largest trading platform in the crypto market, issued a report titled "Low Float & High FDV: How Did We Get Here?" The report listed a group of tokens that had just been launched at the time. The lowest circulation was only 6% of the total supply, and the highest was no more than 20%.

The ratio of market cap to fully diluted valuation for newly issued tokens in 2024 is the lowest in the past three years. When I launched the company, I only put in a small amount of chips, and the valuation soared to the sky. In the early days of their launch, these projects relied on very few tradable chips to drive up prices. It seems that the market value has equaled those of Layer-1 and DeFi blue chips that have been running for several years. Then once it is unlocked, the price goes all the way down.

This is the lesson that the crypto market has taught everyone in two years: low circulation can push prices up, but it cannot be sustained. Because the locked position will eventually be unlocked, and unlocking is supply. If the supply does not meet the corresponding demand, the price will fall. Binance Research has done some calculations: From 2024 to 2030, it is expected that $155 billion in tokens will be unlocked. For these tokens to maintain current prices, the market would need an additional $80 billion in buy-side liquidity.

Memento Research statistics show that of the 118 major token issuance events (TGE) in 2025, 100 of them, or 84.7%, have current prices lower than the fully diluted valuation at launch. The median decline was 71%.

CryptoRank wrote very straightforwardly in its 2025 year-end summary: "The token issuance model with low circulation and high FDV continues to hinder the arrival of the alt season, because most of the upside space has been taken away by private equity and early investors in advance, leaving very limited opportunities for the public market."

The price formation mechanism of low-circulation tokens in the crypto market and low-circulation IPOs in traditional finance in the early stages of listing/launching is almost exactly the same: a small amount of chips, a lot of narrative, buying orders far exceed the tradable supply, and the price is pushed to a position far beyond the fundamentals.

Will the decline of crypto altcoins be repeated in the stock market?

In fact, not necessarily. After all, the stock market has index passive funds that provide continuous buying, deeper institutional participation, and more diverse sources of funds. These are structural supports that the crypto market does not have. Let a low-float IPO in the stock market rise for several months.

Of course, the stock market still cannot escape the day when the ban is finally coming.