329 nhà giao dịch tạo nên 'định giá' 1,2 nghìn tỷ đô la cho Anthropic, sự lo lắng về AI cuối cùng cũng có một cái giá

Original author: Sleepy

Yesterday, I came across an article with the headline: "The New Global AI King is Born! Anthropic's Valuation Rockets Past $1.2 Trillion, Surpassing OpenAI for the First Time."

This headline perfectly captures the zeitgeist. It has AI, an underdog story, a new king being crowned, and a number so large it boggles the mind. It's like a gong. When the gong sounds, it's hard not to look up.

Where did this $1.2 trillion valuation come from? It actually comes from the on-chain Pre-IPO market.

The so-called on-chain Pre-IPO market doesn't trade the common stock you'd see in a brokerage account. It's more like a designed "pre-IPO exposure" instrument. Someone tokenizes, uses an SPV, or creates a synthetic structure to slice up the future listing expectations of a private company and facilitate trading on-chain. It opens a window for ordinary investors that was previously hard to access and provides the market with a real-time price. This is where Anthropic was priced at $1.2 trillion.

Over the past two years, the feeling AI has given ordinary people is often not "I'm participating in a new era," but "a new era has passed me by." Nvidia went up, cloud providers went up, large model companies went through rounds of financing, but the real core equity was mostly locked away in the private market. We could see the ship, but we couldn't grab a ticket. So any ticket that might lead to companies like OpenAI or Anthropic comes with its own filter.

But the more this is the case, the more necessary it is to take that number out of the headline, put it on the table, and see exactly where it came from. Anthropic might be one of the most worthy AI companies to study right now. But the problem is, a great company, a great era, and an aggressive price don't automatically merge into the same thing.

On the crypto exchange Jupiter, Anthropic's Pre-IPO token had a 24-hour trading volume of only $1.39 million, with just 329 traders in the past 24 hours. And it was precisely this $1.39 million and 329 traders that illuminated a trillion-dollar illusion.

But I don't want to discuss whether Anthropic is worth its valuation, or whether on-chain Pre-IPO asset trading is problematic per se. I want to first understand a more fundamental question: What conditions must a price meet to qualify as a "valuation"?

The Birth Certificate of a Price

In February 2026, Anthropic closed its Series G funding round. It raised $30 billion at a $380 billion valuation, led by Singapore's sovereign wealth fund GIC and hedge fund Coatue Management. A month later, OpenAI also announced the closing of its latest funding round, $122 billion at an $852 billion valuation, with major investors including SoftBank, Microsoft, and other institutional investors.

How were these sets of numbers generated?

Take Anthropic's Series G as an example. GIC is a sovereign wealth fund managing over $700 billion, and Coatue manages over $60 billion as a global tech-focused hedge fund. Each has dozens of people on their due diligence teams, spending months analyzing Anthropic's technology architecture, revenue curve, customer retention rates, and competitive landscape. The final $30 billion investment came with a full set of legal terms, including anti-dilution protection, liquidation preferences, information rights, and board observer seats. If Anthropic underperforms or heads south, these terms ensure GIC and Coatue can get their principal back first.

What they bought wasn't just a number; it was a whole set of legally enforceable rights structures.

What about the $1.2 trillion on Jupiter? A few hundred traders, a daily volume of just over a million dollars. The token behind it carries no promises or obligations from Anthropic. What you buy is not a fraction of the company's ownership, but an on-chain bet receipt.

Both prices are presented identically in news headlines, both called "valuation of XX billion."

In 1985, financial economist Albert Kyle published his seminal paper, "Continuous Auctions and Insider Trading," introducing the concept of "market depth," measured by λ (lambda) to gauge the price impact of a unit of capital inflow. In a deep market, buying $100 million might only cause a 0.1% price fluctuation. In a shallow market, $50,000 can move the price by 20%. The larger the λ, the greater the price impact of a single trade, and the thinner the consensus information carried by the price itself.

Anthropic on Jupiter: a liquidity pool depth of about $1 million supporting a $1.2 trillion implied valuation. The ratio of liquidity to valuation is roughly 1:1,200,000. If someone wanted to sell $10 million worth of their position at that $1.2 trillion valuation, the entire liquidity pool would be drained ten times over. This price is not executable; it exists only on the books, un-cashable in the real world.

If it were simply treated as a reference indicator for observation, that would be fine. The problem is it wasn't treated that way. It became the argument for "officially surpassing OpenAI," the headline for "the birth of a new global king," and a cognitive input for countless readers.

This phenomenon of packaging marginal prices from thin markets as broad consensus isn't new. It's been happening for almost four hundred years.

The Tavern in Haarlem

February 3, 1637. Haarlem, Netherlands.

In a small tavern, about thirty people sat around a long table. Following the custom in Amsterdam and Haarlem at the time, these informal tulip bidding gatherings happened several times a week, usually in the back rooms of taverns. The participants were merchants and flower brokers from within the circle, who knew each other well.

That day's lot was a *Semper Augustus* bulb. Its red and white petals were considered a masterpiece of creation, with only about a dozen known to exist in all of the Netherlands. Bidding lasted an evening, and the final price was 10,000 Dutch guilders.

In 1637 Amsterdam, a canal-side townhouse cost about 5,000 guilders. A skilled craftsman's annual income was about 300 guilders. One bulb equaled two mansions, equaled 33 years of a craftsman's wages.

This price was born from just thirty people, in an enclosed space, fueled by alcohol. There were no external constraints, no market-maker obligations, no information disclosure requirements. Bidders fueled each other's excitement and faced no obligations beyond payment for their bids.

The next day, the transaction price was recorded in a pamphlet printed in Haarlem. The pamphlet traveled via postal routes to Leiden, Rotterdam, Utrecht, and other cities. Farmers and small merchants who read it had no way of knowing how that number was generated. In their eyes, the price in print was the market price. Some began hoarding common bulb varieties based on this, believing the entire market would rise.

On February 6th, at a tulip auction in Alkmaar, bidding suddenly stopped. Then Haarlem, then Amsterdam. Within a day, buy orders vanished across the Netherlands. Those who had hoarded bulbs based on the "market price" found no one to take them off their hands. The price crashed, dropping over 90% within a week.

In hindsight, the "10,000 guilders" for that *Semper Augustus* wasn't the judgment of a market, but the judgment of a room. Amplified by the printing press, the room's judgment became a national perception.

Eighty-three years later, 1720, London.

The South Sea Company's stock rose from £128 at the start of the year to £1,050 in June. Founded in 1711, this trading company held a monopoly charter for British trade with South America, but its actual trade profits were very thin. The real driver of the stock price surge was a complex debt-for-equity scheme. The company proposed taking over government debt and converting it into company stock, then sustaining the cycle by constantly driving the stock price higher.

Newton sold his South Sea Company shares when they hit £300, making a profit of £7,000. But the price continued to soar. In July, Newton bought back in, this time buying at the top. After the crash that autumn, his total loss reached £20,000, equivalent to about ten years' salary as Master of the Royal Mint.

Newton probably never gave much thought to how the "£1,050" he was referencing was generated.

1720 had no electronic trading systems, no central counterparty clearing. Buying or selling South Sea Company shares required going to the company's office in London to transfer ownership, or finding a broker in one of the coffeehouses on Exchange Alley. Daily transactions might have numbered in the dozens to hundreds, involving perhaps a hundred direct counterparties.

These prices were recorded on the price lists at Jonathan's Coffee House. When newspapers reprinted these lists, they didn't add a footnote saying "12 trades today, total turnover about £8,000." Readers across England saw only one number: "South Sea Company: £1,050."

When the panic selling started in late July, the price born from the limited game of a few hundred people was instantly shattered. No market maker was obligated to step in. No circuit breakers, no central bank intervention. By December, the stock price had fallen back to £124, almost back to where it started the year.

Now leap forward two hundred and sixty years. Tokyo, late 1980s.

"The land value of the Imperial Palace is worth more than the entire state of California." This statement was widely cited by global media in 1989. Based on estimates at the time, the total value of the 2.3 square kilometers of land occupied by the palace, extrapolated from surrounding land prices, was about $850 billion. The assessed value of all land in California was about $500 billion. But this estimate only referenced the unit prices from a handful of actual transactions in the Ginza and Marunouchi areas.

Japan's land market had a unique structural characteristic: extremely low turnover. Japanese landowners treated real estate as a family asset passed down through generations, not for trading or arbitrage. At the peak of the bubble in 1989, the total number of annual land transactions in Tokyo's core business districts was very limited. The occasional plot that came to market was usually due to owner bankruptcy or family inheritance disputes. A large pool of well-funded, eager buyers competed for a very scarce supply.

The prices generated under this extreme supply-demand imbalance were extrapolated by real estate appraisal agencies as the "fair value" for all land in the area. Their logic: if this small plot is worth 20 million yen per square meter, then every adjacent plot must be worth the same.

In 1990, the Bank of Japan raised interest rates repeatedly, and banks tightened lending standards. When companies were forced to sell real estate to repay loans, the true liquidity test began. Sell orders flooded in; buy orders were thin. Actual liquidation prices were 50% to 80% lower than the so-called market valuations.

Japan's national land price index then fell continuously for twenty-six years, not seeing its first modest recovery until 2016.

The tavern in Haarlem, the coffeehouse in London, the real estate appraisal office in Tokyo, the Jupiter DEX on Solana. Four scenarios spanning nearly four centuries share the same narrative structure:

A very small number of participants generate an extreme price in a very thin market → Media amplifies it into broad consensus → A wider public makes decisions based on it → When liquidity is truly tested, the price reverts.

The media evolved—pamphlets, newspapers, telegraphs, television, WeChat public accounts—but that core flaw was never fixed. When prices are transmitted, their conditions of birth are systematically omitted.

Why?

Complex Stories Are Always Compressed into an Easy-to-Spread Number

Business reporting has an inherent problem: the real world is too complex, and the window for spreading information is too short.

To explain what's actually happening with a company, you often need to cover its funding structure, product progress, revenue quality, competitive landscape, equity rights, exit paths, and market sentiment. But headlines are just one line, and readers' attention spans are just a few seconds. So expressions like "valuation breaks a hundred billion," "market cap evaporates a trillion," "unicorn born," "super platform rises" become an easily chosen form of compression. It's a narrow gateway that complex business information must pass through to enter public discourse.

Writers are also in this gateway. We all know that explaining the birth conditions of a valuation is much harder than writing a punchy headline. The former requires patience, space, and time readers are willing to spend. The latter needs only a bright enough number to make people immediately know "something is happening here." If a headline read, "Anthropic's implied valuation reaches $1.2 trillion via extrapolation of marginal prices of on-chain Pre-IPO synthetic assets in a low-volume market," it might be more accurate, but it would likely exhaust its communicative power before reaching the reader.

If it's written as "The New Global AI King is Born," things are different. It has drama, winners and losers, a throne, and the power transition that humans always love to see. Communication isn't a porter of facts; it's more like a juicer. Facts go in, and emotion comes out.

The second reason is market structure. The Chinese business information environment lacks a key player: the short seller.

In the US capital market, a high price disconnected from fundamentals doesn't stay safe for long. The business model of short-selling research firms like Muddy Waters, Citron Research, and Hindenburg Research is to identify targets whose prices far exceed what their liquidity or fundamentals can support, then publish reports publicly while shorting for profit.

They have strong economic incentives to show the public the birth conditions of a number. Muddy Waters' 2020 report shorting Luckin Coffee was 89 pages long, using 92 full-time and 1,418 part-time investigators who recorded over 11,000 hours of store footage at 981 stores nationwide, and meticulously checked 25,843 receipts. All this just to prove one thing: Luckin's reported daily cups per store was fake, with the real number being roughly half of what was claimed.

This level of adversarial research requires two prerequisites. First, a short-selling mechanism exists that allows profit from "price reversion." Second, legal protections exist to prevent the suppression of short-selling reports. Both exist in the US stock market. In mainland China's A-share market and primary markets, both are essentially absent.

The result is that no one can make money by questioning a valuation, so no one has the incentive to ask under what conditions that price was generated.

Short sellers are not destroyers. They are a corrective mechanism within the pricing system. The result of dismantling the corrective mechanism is that price deviations can widen without any resistance until they collapse under their own weight. And every day before the collapse, the market looks normal.

The consequences of these two forces combining are not lacking in Chinese business history.

In June 2015, shares of LeEco (Leshi Internet) hit their peak on the Shenzhen ChiNext Board, giving it a market cap of about ¥170 billion. Jia Yueting's "LeEco Ecosystem" spanned phones, TVs, cars, sports, and movies. The concept of "chemical reactions" between its seven sub-ecosystems led investors to believe the synergies shouldn't be valued individually, but rather priced based on the exponential growth of the entire ecosystem.

No one questioned how much capital was actually playing for that ¥170 billion market cap. LeEco's daily trading volume in 2015 was indeed decent. But over 70% of the shares underlying that ¥170 billion market cap were locked up, meaning the actual floating supply was much smaller than the total market cap suggested. The tradable shares that retail investors and small institutions could buy pushed the price up based on this limited float, and that price was then automatically multiplied by the total shares to produce "¥170 billion."

A big number is produced → It enters a ranking → It provides a sense of certainty → No one has the incentive or ability to question it → An even bigger number is produced.

Seen this way, Anthropic's "$1.2 trillion" isn't surprising; it's just the output of a system operating normally.

Anxiety, Anxiety

Let's read that $1.2 trillion from a different angle.

What kind of person buys a synthetic token with no legal guarantees at an implied valuation three times higher than the latest institutional round, in a market with a liquidity pool of just over a million dollars?

The answer is someone whose FOMO is strong enough to pay a premium for anxiety.

When Anthropic's Series G closed in February 2026, the valuation was $380 billion. Two months later, the implied valuation of the token on Jupiter was over three times that number.

Is this 3x premium due to an information advantage? Do the traders on Jupiter know more about Anthropic's business than GIC's due diligence team? Obviously not. This premium isn't buying cognitive difference; it's buying a form of psychological insurance, a hedge against the fear of "missing out."