More and more people are bearish on Anthropic
Anthropic's highly anticipated, history-making upcoming IPO has been seeing one "unexpected twist" after another.
The latest twist came on November 9, the day before Anthropic was originally scheduled to kick off its investor roadshow, when the company withdrew its IPO pricing filing. The document was initially set to cap its maximum market valuation at $1.441 trillion. Anthropic then updated the filing, setting the valuation range at $70 to $78 per share, which means the maximum market cap was further lowered to $1.404 trillion, with the minimum market cap of the range standing at only $1.26 trillion.
The fierce pricing game surrounding Anthropic's IPO has evolved into a winding serial drama. The valuation dropped from the initial rumored $2 trillion to the current $1.4 trillion in less than half a month.
Meanwhile, in a corner of the internet — the prediction market, a fierce long-short showdown over Anthropic's post-listing market capitalization performance is unfolding, with the bears coming out as overwhelming winners for now. Theoretically, one can net more than 10 times returns within a week by betting against Anthropic.
More and More People Are Betting Against Anthropic
Over the past more than half a month, a carnival has broken out on public betting platforms centered around Anthropic's IPO. On Polymarket, the world's largest prediction market, more than a dozen prediction markets related to Anthropic's IPO of varying sizes have emerged since October, with total wagers exceeding $37 million. This figure does not even include several other mainstream betting platforms such as Kalshi and trade.xyz.
The so-called "prediction market" is a concept that has become extremely popular in the cryptocurrency circle this year. Simply put, it allows users to trade predictions of real-world events: buying "Yes" means betting that an event will happen, while buying "No" means betting that it will not. The price of each prediction unit ranges from $0 to $1, directly representing the probability of the event happening as perceived by the market. The difference between it and ordinary sports betting platforms is that there is no bookmaker, only peer-to-peer betting between users.
Polymarket is a leading player in the prediction market. In the first half of this year, its average daily trading volume on the platform increased 4 times year-on-year. After its Series F financing at the end of October, the company's valuation reached $21 billion. Due to the large number of users on Polymarket and the fact that all bets are placed with real money, its prediction results are often regarded as a kind of vane on social platforms, especially for predictions of events such as the US general election and sports events like the World Cup.
Finance is traditionally not a popular field on the Polymarket platform, but this year more and more people have begun to "trade stocks" on Polymarket. The last time Polymarket drew attention in the secondary market was during SpaceX's landmark IPO in the first half of the year. At that time, the fierce debate over SpaceX's valuation was no less intense than the current one surrounding Anthropic. From early April to early May this year, the implied probability that SpaceX's market cap would exceed $2 trillion on its first trading day close on Polymarket rose all the way from 45% to 62%. In the end, SpaceX's market cap at close on the first day of listing indeed exceeded $2 trillion. Since then, many media outlets and analysts have begun to regard the betting results on prediction platforms as a reference for price discovery before major IPOs.
This time, Anthropic's IPO has drawn even more attention than SpaceX's. Reuters quoted a wealth manager as saying: "People's interest in participating in Anthropic's IPO predictions is just like queuing up to grab a newly released iPhone." Among all kinds of predictions (listing date, fundraising amount, etc.), the traditional "market cap at close on the first trading day" project is of course the one that retail investors are most enthusiastic about participating in.
However, on this project, Anthropic's performance in recent days has been quite worrying, with the predicted market cap shrinking rapidly. Let's take a look at the odds. As of November 10, the odds for betting that Anthropic's market cap at close on the first trading day will be lower than $1.25 trillion on Polymarket have dropped to 2.8x. Just a week ago, this odds was as high as 42x. At the same time, the odds for betting that Anthropic's market cap will be between $1.75 trillion and $2.0 trillion have risen from 2.5x at the end of October to 5.6x now.
To put it simply, a week ago, the betting market believed that the possibility of Anthropic's market cap at close on the first trading day being lower than $1.25 trillion was only about 2.4%, but now this figure has risen to about 36%.
This is no ordinary poll: everyone who makes a choice has placed a bet with real money. Moreover, the "prediction shares" on Polymarket can be traded in real time. That means, if you bought the prediction that "the first-day market cap will be lower than $1.25 trillion" a week ago and sell it today, you can theoretically get a return of about 15 times.
Although it has not been listed yet, Anthropic's stock price on Polymarket has already crashed in real terms. If Anthropic's market cap on the first trading day is indeed lower than $1.25 trillion, it means that the stock will break even if priced at the lower end of the range. This is a world of difference from the state where SpaceX's IPO was enthusiastically sought after by bettors on Polymarket in May.
Why Did Anthropic Withdraw Its Pricing Filing?
At a time when the prediction market has largely shifted to a bearish stance, the seemingly erroneous withdrawal of the offering filing has added another haze to Anthropic's IPO journey.
Reuters reported that Anthropic sent a document to investment banks on the evening of November 9 US time, showing that the prospectus plan was to issue approximately 867 million shares with a price range of $70 to $80 per share, and the fundraising amount would be about $693 billion calculated at the upper limit of the range. But less than two hours later, Anthropic sent another email, stating that the previously provided pricing information was inaccurate and that it would update the filing later.
Subsequently, Anthropic updated the filing, lowering both the pricing range and the fundraising amount. It finally confirmed that it would issue a total of 825 million Class A shares, with a valuation range of $70 to $78 per share, and the raised funds would drop to between $577.5 billion and $643.5 billion.
If we look at it from a conspiracy theory perspective, maybe Anthropic wanted to test the market reaction before deciding whether to adjust the terms, and the investors' feedback was poor, so it withdrew to redesign the offering structure. Of course, it is also possible that this is an overinterpretation: the world is indeed a makeshift setup, and this is just a simple internal communication error, where the wrong version of the document was sent.
Before October, when Anthropic's prospectus had not yet been released, the expected market cap announced was $2 trillion. In the past 10 days or so, this figure has shrunk sharply. In early November, media reports said that $1.5 trillion was regarded as the upper limit of Anthropic's pricing range.
Then came the stock market crash on November 5, when the Nasdaq Composite Index fell 5.56% in a single day, and the Nasdaq's market cap evaporated by about $2.5 trillion, marking its largest single-day market cap decline in history. Technology stocks, chip stocks and AI-related companies almost all collapsed. On the same day, The Wall Street Journal reported: "Some IPO investors are skeptical about whether Anthropic can achieve a $1.5 trillion valuation." There were also rumors that Anthropic might decide to postpone the offering.
At present, Anthropic seems to be very "receptive to feedback", choosing to significantly lower the offering price to continue pushing for the IPO, instead of "waiting for a more favorable market environment".
The Giant Ship of AI Is Swaying
The fierce pricing game surrounding Anthropic's IPO has clearly evolved into a general showdown between the bullish and bearish attitudes towards AI in the capital market. For the same prospectus, one faction's analysis believes it "exceeds expectations" and even "re-energizes the struggling AI industry"; the other faction focuses on the slowdown in Q3 growth, believing that the grand narrative of AI has "peaked".
As the most outstanding college student in the village, Anthropic carries too much on its shoulders.
For VCs in the primary market, Anthropic's IPO is of great significance. Due to its and OpenAI's unprecedented huge financing needs, many established VCs have had to make "decisions against their old conventions", and have to significantly expand the size of their funds in order to get on the bandwagon. After entering 2025, a large number of controversial SPVs were even issued to specifically invest in these two giants.
The current problem is that Silicon Valley VCs hold a large amount of paper gains, but over the past 10 years they have not returned much cash to LPs, so it is difficult for LPs to make additional contributions to larger funds in the future. If Anthropic is successfully listed, hundreds of billions of dollars of massive capital will flow back into the Silicon Valley VC ecosystem, and all these problems will no longer exist.
For the entire AI industry, Anthropic's IPO affects the whole body with one single move. On the one hand, its stock price will directly become the valuation anchor for other AI companies. On the other hand, Anthropic has a whopping $895.4 billion in future computing power and infrastructure expenditure commitments on its books, of which $447.5 billion will mature before 2030. If Anthropic's performance or cash flow falls short of expectations, the entire AI infra will need a revaluation.
Anthropic's products remain excellent. While all kinds of unexpected dramas are playing out during the IPO process, Anthropic released its new generation flagship model Fable 5.2 on November 9, and simultaneously launched Mythos 5.2 based on the same underlying model. Fable 5.2 continues to lead the world in performance, and while delivering stronger performance, it has also significantly reduced its price, even cheaper than OpenAI's secondary model Sol 6.
However, in stark contrast to the first half of the year, the capital market can hardly simply pay for "stronger models" now.
It is worth mentioning that SoftBank released its latest quarterly report (July-September) on November 10. There are two eye-catching pieces of information in the financial report: one is a net loss of 1.12 trillion yen for the single quarter; the other is that Musk was proven right — SoftBank actually defaulted on a batch of procurement orders for power generation equipment, which were originally intended to be used in a data center in Ohio, USA, and this data center is part of the "Stargate" project jointly built by OpenAI, Oracle and SoftBank.
From Anthropic to SoftBank, it would be too exaggerated to say that the bubble has burst, but there is no doubt that the giant ship of AI is swaying.
This article is from the WeChat Official Account "ChinaVenture", Author: Tao Huidong, Editor: Wang Qingwu, published with authorization from 36Kr.