Skip to content
Business

Anthropic is on the market, and the big model bubble is not far away

Author | Mangosteen

Produced | Zinc Industry

A large model company set a global IPO fundraising record, while on the other hand, Chinese and American technology stocks were falling.

On November 19th, U.S. time, Anthropic, the company behind Claude, landed on Nasdaq with the stock code “ANT”. The issue price was US$83, and the opening price was US$111.50. It once reached US$115 during the session and finally closed at US$97.17, up approximately 17.1% on the first day.

This was a listing worthy of being written into the history of the capital market: the public issuance of 980 million shares, totaling US$81.34 billion, setting a new global IPO fundraising record.

But it didn’t restore the market’s belief that AI stocks can only go up.

On the same day, the three major U.S. stock indexes all fell by more than 2%. In Hong Kong, Tencent also suffered a sell-off due to financial performance and pressure on AI investment. It once fell by more than 7% during the session on November 19.

What’s more subtle is that just as the market is recalculating the return on investment in AI, the pace of financing and listing has not stopped.

OpenAI is advancing preparations for listing, Altman has just reiterated its intention to go public, and DeepSeek’s second round of financing worth hundreds of billions has also entered the public eye in the past two days with the industrial and commercial changes.

Funds are still pouring into leading companies, but the public market is increasingly unwilling to pay a premium just for the “prospects of AI.”

This is the scene that deserves more vigilance when the bubble is approaching:

Technology and business are still growing, but capital’s pricing of the future may have gone too far.

01 The largest IPO in history failed to save the AI ​​market

Anthropic’s IPO has almost brought investors’ enthusiasm for leading AI companies to the forefront.

The final issuance price exceeded the previous guidance range of US$70 to US$78, and the number of shares issued increased from 825 million to 980 million shares. The expanded issuance and price increase showed that the company’s stock was still sought after.

In addition to the public offering, Anthropic also simultaneously arranged a private placement of US$26 billion, with the two transactions totaling approximately US$107.34 billion. What needs to be distinguished is that US$81.34 billion is the total amount of IPO issuance, which includes the sale of old shares, not all of which enters the company’s account. Private placement is another financing arrangement.

Such a huge transaction did not prevent the stock price from rising and falling on the first day. It retreated from the opening price of $111.50 to the closing price of $97.17.

This record-level IPO failed to reverse market sentiment, which has shown that investors’ attitudes towards AI are diverging.

In the past, ordinary investors mainly participated in AI competition through NVIDIA, cloud vendors and data center companies. The investment logic was relatively straightforward: model companies continued to increase investment, and infrastructure providers had the opportunity to obtain orders.

After Anthropic went public, the market could directly buy shares of cutting-edge model companies, and the question became more acute: How much money can be earned from customers by paying to chip and cloud computing manufacturers?

It will become a valuation benchmark for peers, and each subsequent revenue growth, cost change and profit fluctuation may affect how the market evaluates other large model companies.

The private market can wait for the next round of financing and use the new transaction price to prove the value of the previous round of investment. After entering the public market, this process becomes a daily purchase and sale.

Going public opens up new financing channels and puts issues that were previously wrapped up in long-term growth expectations before investors.

02 Tencent was sold, and the market began to calculate the consequences of AI

Compared with large-model startups that are still explaining their profit models, Tencent can better reflect changes in market attitudes.

It has mature businesses such as games and advertising, with profits and cash flow. Even so, once AI investment begins to significantly affect current returns, investors will still reprice.

Tencent’s third quarter results announced on November 18 showed that revenue was approximately 207.3 billion yuan, a year-on-year increase of 8%, non-IFRS operating profit was approximately 68.1 billion yuan, a year-on-year decrease of 6%, and the corresponding operating profit margin dropped from 38% to 33%.

If the contribution of new AI products is excluded, the operating profit of the same caliber will be approximately 83.3 billion yuan, a year-on-year increase of 8%, and the operating profit margin will be 40%, remaining stable year-on-year.

However, changes in cash flow are more intuitive.

In the third quarter, Tencent’s capital expenditure payment was approximately 65.1 billion yuan, a year-on-year increase of 225%, and the net free cash flow outflow was approximately 69.1 billion yuan. The company explained that operating cash flow includes large prepayments for computing power purchases, which are used for model upgrades, WorkBuddy reasoning, and other product and cloud service needs. If this part of the prepayments is excluded, free cash flow is still positive, approximately 16.9 billion yuan.

This shows that AI expansion will bring cash expenditures forward, and the returns need to be realized in the future.

After the financial report was released, Tencent’s stock price fell by more than 7% intraday on November 19. When profit margins, capital expenditures and cash flow became the focus, the market has actually changed the way it asks questions.

Previously, investors were worried that the company would not invest enough and would miss out on AI. Now, investors are also worried that the company would invest too much and the returns would come too slowly.

For AI, investors have begun to lower their optimistic pricing for future growth while increasing their requirements for current profits and cash flow.

If the fees that companies are willing to pay for AI cannot keep up with the funds that the industry chain has invested and committed to investing, the pressure will spread from a certain company to the entire chain.

Anthropic’s business is not without growth. The prospectus shows that revenue in the first half of this year was approximately US$16.3 billion and approximately US$17 billion in the third quarter. Tools such as Claude Code integrate models into workflows and are creating more specific reasons for paying.

However, high growth did not eliminate cost fluctuations. The company achieved an operating profit of approximately US$974 million in the second quarter, and recorded an operating loss of approximately US$567 million in the third quarter. The prospectus linked the third quarter changes to more expensive short-term computing power arrangements, non-cash expenses related to donation matching and other factors.

This is where valuation pressure comes in. Even if demand is real, companies still need to prove that growth can translate into sustained, predictable returns.

The fact that technology is effective does not mean that buying at any price is reasonable; the growth of income does not mean that there is no risk in overdrafting the valuation for many years in advance.

03 OpenAI and DeepSeek continue to take money

The public market has begun to calculate returns and has not stopped leading large model companies from financing.

OpenAI’s listing process has also exceeded the “intentional preparation” stage.

Reuters reported in June this year that OpenAI had confirmed that it had secretly submitted a U.S. IPO application, which means that the company has started the relevant process. However, secret filing does not mean that it has been approved for listing, nor does it mean that the issuance price and date have been determined.

In September, OpenAI CEO Sam Altman said the company would not go public in 2026, citing security concerns. On November 18, Altman reconfirmed at the New York Times DealBook Summit that the company planned to become a public company.

The latest statement continues the listing direction, but does not give a firm date.

Anthropic went public first, which provided a reference for OpenAI and also brought new comparative pressure. The two companies not only had to compare models and products, but also faced investors’ horizontal examination of revenue quality, computing power costs and profit paths.

Funding can help them continue to compete. The larger the scale of financing, the greater the returns that need to be realized in the future.

The same is true for the Chinese market.

From November 19th to 20th, the latest progress of DeepSeek’s second round of financing was disclosed by many media. This new change is that the previous news surrounding the scale of financing has begun to correspond to the industrial and commercial registration level.

According to Caixin reports, Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co., Ltd., the main body of DeepSeek, completed its industrial and commercial changes on November 17 and added a new shareholding platform, Hangzhou Cheng Yan, holding approximately 17.27% of the shares. The platform added 31 limited partners during the same period.

Although Liang Wenfeng has withdrawn from the list of direct partners of Hangzhou Cheng Yan, he still controls the platform by holding 99% of the shares of executive partner Hangzhou Cheng Ce.

In addition, relevant investment entities such as Tencent and CATL, as well as industrial capitals such as China Mobile, Midea, and Vivo, appear in the investment network.

It is reported that the scale of this round of financing is approximately 100 billion to 105.5 billion yuan. In the latest financing, Tencent’s role is particularly worth looking at together.

On the one hand, Tencent is under market scrutiny due to the pressure on profits and cash flow brought by its own AI expansion. On the other hand, it continues to participate in the competition of leading model companies through investment.

These two options can be established at the same time. Investors require current returns, and companies need to reserve positions for future competition. The question is how long the company can withstand the investment and how long the market is willing to be patient.

On one side, Anthropic is going public, OpenAI is preparing to go public, and DeepSeek has completed a large financing. On the other side, the stock market is beginning to kill AI expectations.

About Us · 關於我們