The Zhitong Finance App learned that some media reported, citing information revealed by people familiar with the matter, that Anthropic PBC, a global AI model/ AI application leader, expects its initial public offering (IPO) in the US stock market to match or surpass the record IPO scale when SpaceX listed in the US stock market in June. This latest trend on the AI industry chain can be described as once again showing that institutional and retail investors hoping to profit from the unprecedented AI investment boom are extremely strong.
According to people familiar with the matter, the developer of this Claude AI application platform is carrying out related valuations and IPO fund-raising estimates, and management is preparing to publicly submit the necessary application documents for a potential large-scale IPO as soon as the end of this month. They also said that the recent investor briefing hosted by Anthropic Chief Financial Officer Krishna Rao deliberately avoided valuation issues. It is worth noting that just in May of this year, Anthropic completed a round of financing worth about 65 billion US dollars, with a valuation close to trillion US dollars (about 965 billion US dollars), surpassing competitor OpenAI's valuation of 852 billion US dollars in March this year.
According to statistics compiled by the agency, SpaceX, a rocket and satellite company owned by Tesla CEO and the world's richest man Elon Musk, initially successfully raised 75 billion US dollars, making it the largest initial stock offering in history, with a market value of 1.77 trillion US dollars when it landed in the US stock market. After including the so-called “over-allotment option,” the final fundraising scale increased to an unprecedented $86.2 billion; if stocks rise in the early stages of listing, this option is usually exercised.
People familiar with the matter said that related discussions are still ongoing, and details such as the scale of Anthropic's current IPO may still change. Since the relevant information has not been made public, they requested anonymity. A representative from Anthropic failed to immediately respond to a request for comment.
Claude with a market capitalization of nearly a trillion dollars climbs to the top of capital
The scale of Anthropic's more ambitious IPO target than SpaceX is enough to reflect that the hottest leaders in the artificial intelligence industry are reshaping the technology investment landscape. The company, which has only been in existence for five years, raised 65 billion US dollars at a valuation of 965 billion US dollars in May this year, surpassing the $852 billion valuation reached by rival OpenAI when it raised 122 billion US dollars in March this year.
Although Anthropic's adjusted operating profit for the second quarter improved faster than market expectations, according to a content document seen by the agency, the company's net loss in 2025 was close to US$42 billion, which is about five times the net loss of about US$8.3 billion the previous year.
The good news is that demand for artificial intelligence-related applications is growing rapidly, as evidenced by the sharp rise in Anthropic's revenue statistics. According to media reports citing information revealed by people familiar with the matter, the company's initial revenue for the second quarter exceeded 11.5 billion US dollars, while the same period in 2025 was only about 787 million US dollars; by the end of July, its annualized revenue run rate (run rate) had reached an astonishing 65 billion US dollars. The operating rate of revenue is an indicator for estimating annual revenue based on short-term revenue.
Anthropic's underlying business is no longer a single chatbot, but a full-stack business system of “leading-edge big model+API+Agent Harness (Agent Harness) +enterprise workflow+multi-cloud distribution”: Claude provides core intelligence, Claude Code cuts into the high-frequency, quantifiable return scenario of software development. Model Context Protocol (MCP) and Agent SDK connect enterprise data and tools, AI intelligence super platform Claude Cowork then extended the same intelligence capabilities to documentation, compliance, customer service, and knowledge work, and covered large-scale enterprise procurement channels through Amazon Bedrock, Google Vertex AI, Microsoft Azure, and direct APIs.
The reason why Claude Cowork, an AI intelligence platform closely centered around AI programming applications, took the lead in generating strong compounding because whether the code can be compiled, whether the test passes, and whether the tool call is successful can all be automatically verified; Anthropic also transforms the actual failure trajectory into evaluation sets (evals) and next-generation model training goals, forming a closed loop of collaborative design of “user use — product feedback — model enhancement — experience improvement — usage growth”.
The demand for large models has moved from one-time question-and-answer to a multi-step workflow where AI agents continuously complete planning, search, code execution, tool call, and result verification: each task consumes tokens exponentially, while depositing actual usage data and evaluation feedback, driving another increase in model capability, task success rate, customer stickiness, and revenue, forming a closed loop of “more calls — stronger models — higher automation rate — more paid calls”.
Claude Cowork and its plug-ins for law, sales, marketing, and data analysis are the epitome of this paradigm — it elevates Claude from a chat portal to the “executive layer” of enterprise work, which has raised market concerns about shrinking traditional software seats, commercialization of functions, and weakening pricing power. Related releases have triggered the software and data services sector to evaporate about US$285 billion in a single day, and Goldman Sachs's US software stock basket falls 6% in a single day; although this shock does not mean that traditional SaaS will be replaced immediately, it marks that global AI competition has moved from “who has the strongest” The “model” has entered a new phase of “who can let agents take over most workflows and turn tokens into cash flow”.
Some people familiar with the matter said that Anthropic and its rivals are dealing with the extremely high costs required to build more advanced artificial intelligence super application systems. Training so-called cutting-edge models requires huge AI computing infrastructure resources — in one of the agreements Anthropic and data center operators alone, the company agreed to deploy AI computing resources on a large scale through SpaceX, and the contract value could reach tens of billions of dollars within the next three years.
According to media reports, Anthropic is expected to land on the open market before OpenAI, which is currently considering listing in 2027. Both companies have secretly submitted separate listing applications.
People familiar with the matter previously revealed that before publicly submitting listing documents, Anthropic is about to finalize a revolving credit line, the size of which will exceed the initial target of about $10 billion.
People familiar with the matter said that Anthropic is cooperating with Morgan Stanley, Goldman Sachs Group and J.P. Morgan Chase to advance the IPO, and other major commercial banks, including Wall Street financial giant Citi, may also join the underwriting team.
A person familiar with the matter said that Anthropic is considering adopting so-called super-voting shares to give CEO Dario Amoudi, who holds about 2% of the company's shares, and other co-founders stronger actual control over the company. The Information was the first to report on super-voting shares before.
According to data compiled by the agency, if the overall scale of Anthropic's initial stock offering surpasses SpaceX, it will easily push 2026 to become the best year for a record US IPO to raise capital. As of August 19, newly listed US companies have raised a total of US$160.6 billion since this year, which is lower than the record of US$195.2 billion set in 2021.
Anthropic's listing will further strengthen the global stock market IPO market, which has already achieved excellent results this year. Two of the largest public transactions with listed companies in history have been launched this year. One of them is SpaceX, an American supertech giant founded by Musk that focuses on “AI+ space exploration,” and finally landed a major US stock market with a record IPO. The other is that SK Hynix, a Korean memory chip manufacturing giant, raised 26.5 billion US dollars through the initial issuance of American depository certificates.
Seven-month revenue jumped 7.2 times: Anthropic challenges SpaceX's financing throne with agent compound interest
Anthropic's financial curve provides rare verifiability for Wall Street's extremely optimistic profit path and record IPO expectations for the company. Anthropic's annualized revenue operating rate rose from about US$9 billion at the end of 2025 to US$47 billion in May 2026 and over US$65 billion at the end of July. The revenue scale reached about 7.2 times, an increase of about 622% within seven months; initial revenue for the second quarter exceeded US$11.5 billion, about 14.6 times that of US$787 million in the same period in 2025, and reached about 2.4 times that of US$4.73 billion in the first quarter, and achieved a correction in adjusted operating profit for the first time.
The financial fundamental-level operating data described above all mean that the revenue scale and corporate seat revenue related to AI inference in the AI inference era began to grow faster than R&D, personnel, and unit computing power costs; model compression, caching, batch processing, different model routing, and custom chip deployment continued to reduce the cost per token, while high-paying enterprise workloads also increased revenue at the computing power level per unit. However, $65 billion is only an annualized operating rate (run rate) extrapolated from the revenue rate at the end of July; it is not a full-year revenue that has already been achieved; different companies may also have different levels of confirmation of total or net revenue from cloud channels.
Anthropic is likely to equal or surpass SpaceX's initial IPO of $75 billion, or even the final size of $86.2 billion after overallotment. It is essentially the result of a combination of “certainty in performance growth, scarcity premiums in the artificial intelligence industry, and strong capital demand”: the company just raised $65 billion at a valuation of $965 billion in May, more than double the valuation of $380 billion in February, and provided the open market with extremely scarce pure cutting-edge AI model investment targets prior to OpenAI. Wall Street giants such as Morgan Stanley, Goldman Sachs, and J.P. Morgan Chase can take advantage of the structural AI allocation requirements of sovereign funds, pensions, and large technology funds to undertake huge distributions.
At the same time, the IPO is not only a market reward for Claude, but also to raise capital for training clusters, long-term reasoning capacity, talent, and potential mergers and acquisitions: its pre-listing revolving credit line is expected to exceed $10 billion, further indicating that the company is building an “equity+credit+long-term computing power contract” financing stack ahead of schedule. Roughly calculated at a valuation of US$965 billion, US$75 billion to US$86.2 billion is equivalent to 7.8% to 8.9% of the current valuation; the actual circulation ratio and share dilution still depend on the IPO valuation and the composition of new and old shares.
Anthropic has the clearest cutting-edge model commercialization and profit path so far, but it cannot simply be valued by mature software companies after listing: if Claude Code and Cowork continue to expand and unit token costs continue to decline, huge IPOs may become the capital starting point for the next round of enterprise AI platform expansion; if model price wars, computing power costs, customer multi-model procurement, or revenue recognition measures weaken the quality of growth, the valuation based on 2028 forward revenue data will also be drastically compressed.
Earlier, according to media reports, two people familiar with Anthropic's financial situation said that Anthropic's annual revenue is expected to reach 190 billion to 200 billion US dollars by 2028; therefore, what Wall Street is really betting on is not Anthropic's current profit, but whether it can extend the “model - product - feedback - training” compound interest to the entire enterprise-side AI agent market. Based on the current $2 trillion valuation and $200 billion revenue outlook, which is currently hotly debated in the market, it is about 10 times the 2028 sales volume; in contrast, the $965 billion valuation corresponds to about 14.8 times the current operating rate, indicating that the implementation of growth can absorb part of the valuation multiples.