Anthropic's trillion-dollar IPO shockwave: Zhipu AI and MiniMax plunge, the market may re-evaluate the pricing logic of large models
In the early morning of November 20, Beijing time, Anthropic, the giant AI large model company, officially landed on the Nasdaq, raising approximately 813 billion US dollars through IPO. Coupled with the completed 260 billion US dollars private placement, the total fundraising scale reached 1.073 trillion US dollars, breaking the previous global IPO fundraising record held by SpaceX.
On the first day of listing, Anthropic's share price surged to a maximum of 115.10 US dollars, briefly surpassing Tesla and Meta. By the close of trading, Anthropic's total market capitalization stood at 1.27 trillion US dollars, up 17.07% from the issue price, and was only 21 million US dollars less than Tesla's market capitalization that day.
The listing of Anthropic did not simultaneously boost the entire AI sector. At the close of US stock trading on November 19, the Nasdaq Index fell 2.25%, the S&P 500 Index fell 2.02%, and the Dow Jones Index fell 2.19%. The Hong Kong market subsequently extended its weakness. On November 20, the Hang Seng Tech Index once fell by more than 4%. By the close, MINIMAX (00100.HK) fell 10.8%, and Zhipu AI (02513.HK) fell 7.17%.
Founded in 2021, Anthropic is the world's third listed AI large model company following Zhipu AI and MiniMax. Under the highlight, the market's valuation logic for large model enterprises will also be calibrated synchronously.
"The listing of Anthropic brings a business model that was previously vaguely priced in the primary market to the public market. In the future, capital will re-evaluate the sustainability and verifiability of the revenue of model companies," Li Qing, Executive Director of Frost & Sullivan China, told the Time Weekly reporter, "For model companies that are going to be listed next, their valuation logic and share price will be structurally affected."
Behind the High Revenue Growth
Every move of Anthropic can be regarded as the vane of the development of the global AI industry.
In 2021, seven core backbones of OpenAI chose to leave due to conceptual differences with founder Sam Altman on model commercialization and AI safety, and then founded Anthropic.
In May 2021, Anthropic completed its first round of financing, raising 124 million US dollars, with a post-investment valuation of 461 million US dollars. In September 2025, Anthropic completed the US$130 billion Series F financing, which was 271.43% higher than the previous round of financing, and the company's valuation reached 183 billion US dollars, hitting a new high in stages.
In more than five years since its establishment, Anthropic has completed several rounds of financing, with a total cumulative financing amount exceeding 1.134 trillion US dollars, and the company's pre-listing valuation reached 9.65 trillion US dollars. Along with the financing scale, the investors behind it have also attracted much attention. At first, the company's investors were only individual investors and venture capital institutions. In 2023, Google took the lead in investing 300 million US dollars. Shortly afterwards, Amazon announced an investment of 40 billion US dollars, and later Microsoft and NVIDIA also participated in the investment.
What supports Anthropic to win multiple rounds of large-scale financing is the product system with Claude as the core and its commercialization capability.
In 2023, Anthropic launched the Claude model, which maintains leading performance in mathematics, programming and other fields. A Semi Analysis report in August 2026 showed that Anthropic's annualized recurring revenue (ARR) has exceeded 440 billion US dollars, a nearly 5-fold increase from 90 billion US dollars at the end of 2025. The core driving force of growth is the explosion of enterprise-side demand and the strong performance of the programming agent Claude Code.
Krishna Rao, Chief Financial Officer of Anthropic, once said in an interview, "Claude is becoming more and more important to the way enterprises operate, and the company's financing also reflects the huge demand of customers."
According to the data from Anthropic's prospectus, in fiscal year 2025 (as of December 31, 2025), Anthropic's revenue was 4.59 billion US dollars, a year-on-year increase of 1088%; the company's operating loss expanded from 2.98 billion US dollars to 8.06 billion US dollars. For the whole year of 2025, Anthropic's expenditure on computing and infrastructure reached 7.33 billion US dollars, accounting for 58% of the total operating expenses of that year.
Although Anthropic has completed its listing, there is still room for improvement in the fundamental of the company's revenue.
Li Qing told the Time Weekly reporter that in the previous two or three years, the pricing premise of the capital market was that the demand for computing power continued to grow, the capital expenditure of cloud manufacturers expanded, the model capability continued to break through, and AI would become the next generation of infrastructure. The market was willing to pay a high premium for growth expectations and was relatively tolerant of short-term profitability.
With the listing of Anthropic, the business model and revenue system of large model enterprises are presented to the market. Investors may tend to ask more realistic questions. "How long can the huge capital expenditure continue to grow, can commercial revenue really support the current valuation, and are there financing and supply risks in infrastructure such as power and data centers?" Li Qing said.
The Market May Reassess the Large Model Business Model
At the same time as Anthropic's listing, domestic leading large model enterprises are also impacting the capital market. In addition to the already listed Zhipu AI and MiniMax, Moonshot AI and StepFun, which were once called the "Six Little Tigers of Large Models", have also spread related news from time to time. Recently, DeepSeek also completed the second round of financing at the level of 100 billion yuan.
A number of experts analyzed to the Time Weekly reporter in interviews that Anthropic's listing will have a structural impact on model companies to be listed in the future.
Senior market analyst Wang Jiyue told the Time Weekly reporter: "The listing of Anthropic may prompt the capital market to adjust its expectations for the revenue and cash flow of large model companies, and the adjusted expectations will not be as optimistic as before."
For the business model that was vaguely priced in the primary market before, Li Qing divided it into three layers: the first layer is model API and model services, which generates revenue through Token consumption, API calls and enterprise subscriptions; the second layer is professional workflow and AI Agent, which embeds model capabilities into specific business processes such as programming and research, so that enterprises no longer buy only model capabilities, but work results; the third layer is enterprise-level AI platforms, which enter the core IT systems of enterprises through office suites, permission management and security systems.
Anthropic follows a closed-source technology route. The closed-source route makes it easier to encapsulate model capabilities into chargeable enterprise-level products and form customer stickiness. The monetization of the open-source route relies more on ecological scale, deployment services and computing power collaboration, and revenue recognition is relatively scattered.
However, the two routes face the same set of public market indicators, namely large customer revenue expansion, renewal rate and net revenue retention rate, actual usage frequency of AI products, revenue per unit of computing power, gross profit margin and operating cash flow.
Lu Hong, founder of M&A Expert, believes that the listing of Anthropic will most likely prompt the capital market to re-evaluate the business model of large models. Specifically, the focus of capital assessment may shift from model benchmark scores to unit computing power return and gross profit margin, operating cash flow and customer retention rate.
The performance of Anthropic after listing may become a watershed in the valuation system of the AI industry. Li Qing told the Time Weekly reporter that if Anthropic achieves sustained revenue growth and improves profit margins to gain market recognition, it will have the opportunity to become the valuation anchor for AI foundational model enterprises, opening the listing window for latecomers and raising the valuation center.
"If the market finds that even industry-leading enterprises cannot support excessively high valuations, valuation inversion in the primary market, downward adjustment of issuance pricing and break after listing will become more common risks for subsequent model companies," he further pointed out. For enterprises preparing for listing, share price performance will rely more on verifiable financial quality in prospectus documents rather than track narratives.
According to media reports, Moonshot AI is advancing pre-listing equity restructuring. For model companies that are rushing to list, industry competition will become more intense. Lu Hong believes that Moonshot AI follows the C-end + API route, and capital will directly benchmark its revenue growth rate and gross profit model with Anthropic. StepFun focuses on end-side and hardware embedding, and its valuation logic may be more biased towards the scale of scenario implementation.
He analyzed that if there is a valuation connection problem in Moonshot AI's equity restructuring, it may affect the listing schedule. Among StepFun's shareholders, industrial capital and state-owned capital coexist, and it is necessary to clarify related transaction and independence issues under the rules of the Hong Kong Stock Exchange.
On the other hand, under the superposition of multiple events such as Anthropic's listing and the US mid-term elections, the uncertainty of AI supervision has increased.
Li Qing told the Time Weekly reporter that in the short term, enterprises with real revenue, customer retention and operating cash flow will get a premium, and companies that only rely on model benchmark scores and financing narratives will face valuation correction.
In the medium and long term, the industrial logic of AI as infrastructure has not reversed. The market has only shifted from expectation-driven valuation expansion to performance fulfillment and capital return-driven valuation reconstruction, putting forward higher requirements for enterprise profitability, unit computing power output and return on capital.
"In the future, a large number of small companies that cannot afford computing power will go bankrupt, and most of the startup platforms related to AI applications will not survive," Lu Hong said.
This article is from the WeChat official account "Time Weekly" (ID: timeweekly), author: Liu Juan, published with authorization from 36Kr.