Original title: Anthropic (ANT.US) Publicly Files IPO Prospectus: Rapid Revenue Growth Faces the $1.5 Trillion AI ROI Test

Author: Zhitong Finance Wang Xiaoli

Cover image source | Unsplash

According to Zhitong Finance, AI startup Anthropic has publicly filed its U.S. initial public offering (IPO) documents, showcasing rapid revenue growth while also revealing the enormous costs behind frontier AI model development, long-term compute spending commitments, and its corporate governance arrangements. The highly anticipated listing will not only test investors' valuation of Anthropic, but could also become an important benchmark for Wall Street in evaluating frontier AI business models.

Anthropic confidentially submitted its IPO application in June 2026 and made the relevant documents public on October 26. The company plans to list on the Nasdaq under the ticker "ANT," with Goldman Sachs and Morgan Stanley leading the underwriting syndicate. The documents have not yet disclosed the number of shares to be offered or the proposed price range.

As the developer of the Claude family of AI models, Anthropic has rapidly expanded its business in recent years on the back of growing enterprise customer demand, but its growth has also come with high compute and infrastructure costs. As market concerns over AI data center returns on investment, financing costs, and industry valuations intensify, Anthropic's IPO will become an important opportunity for investors to test the commercial prospects of AI.

Rapid revenue growth, with a notable divergence between operating and net losses

The prospectus shows that Anthropic's revenue grew from $41 million in 2023 to $386 million in 2024, reaching $4.59 billion in 2025. In the first half of 2026, revenue further increased to approximately $16.29 billion.

In the three months ended June 30, 2026, Anthropic's revenue grew 1,368% year over year, reflecting continued growth in enterprise customer demand for its AI models and related services.

However, revenue expansion does not mean the company has achieved profitability. The documents show that Anthropic's operating loss was $2.98 billion in 2024, widening to $8.06 billion in 2025, with an operating loss of $693 million in the first half of 2026.

Meanwhile, the company's net loss reached $80.7 billion in the first half of 2026. Earlier reports indicated that the significant widening of the net loss was mainly related to changes in the fair value of convertible bonds. Therefore, the large gap between net loss and operating loss should not be simply interpreted as the company's operations directly consuming $80.7 billion in cash during the first half of the year.

Anthropic's revenue mainly comes from usage-based services for enterprise customers. Customers pay based on the volume of information processed—the so-called number of "tokens"—and the type of model used. The company also generates revenue through paid subscription products such as Pro and Max.

This usage-based pricing model allows Anthropic to increase revenue as enterprise customers expand their frequency and scale of AI model usage. The company disclosed a net revenue retention rate of 900%, indicating that spending by existing customers has grown significantly compared with earlier periods. In the third quarter, 79% of revenue came from customers other than its top five, showing that its revenue sources are not entirely dependent on a few of its largest customers.

However, Anthropic also cautioned in the prospectus that revenue and accounts receivable remain concentrated among a limited number of customers and distribution channels. As the business scales, customer concentration, revenue sustainability, and whether enterprise customers can continue to increase AI spending remain issues for investors to watch.

Over $500 billion in spending commitments, with compute costs a key test

For Anthropic, compute power is the foundation driving model development and commercialization, and one of the company's largest cost sources.

Its compute and infrastructure spending includes compute fees paid to suppliers such as Amazon Web Services (AWS) and Google Cloud, as well as related expenditures on data center facilities, leased compute capacity, equipment leasing, and energy. Over the past three years, Anthropic's compute and infrastructure spending has consistently been far higher than its revenue.

The prospectus discloses that Anthropic has committed to spend more than $500 billion across six partners over the next decade. These commitments are an important part of its large-scale AI infrastructure buildout plan, but they also mean the company will face substantial future capital needs.

Specifically, Anthropic has committed to pay Amazon at least $110 billion for cloud computing over the next seven to ten years, approximately $114.5 billion to Google, and $44.9 billion to Microsoft for cloud services.

Some agreements are highly binding: Anthropic said many provisions in the agreements are non-cancellable, or require the company to pay regardless of actual usage.

Such long-term commitments help Anthropic lock in compute resources in advance, providing assurance for the training and deployment of more advanced models, but they may also limit the company's ability to adjust costs. If AI demand growth falls short of expectations, model efficiency improvements change compute requirements, or the commercial returns from related services are lower than expected, the company may still need to bear a considerable level of fixed spending.

It should be noted that these figures represent spending commitments spanning multiple years and involving different natures, and do not equal debt Anthropic has already incurred, nor do they mean the relevant payments have all been made. What truly affects the company's financial performance will be how these commitments translate into actual spending in the future, and whether the related compute capacity can generate sufficient revenue and profit.

Amazon and Google are both investors and cloud service suppliers

Anthropic's capital and compute system also reflects a widely watched feature of the current AI industry: some major investors are simultaneously key suppliers to the company.

Amazon has invested at least $13 billion in Anthropic and agreed to invest up to an additional $20 billion. Google has committed to invest up to $40 billion in Anthropic. At the same time, Anthropic has committed to pay Amazon and Google enormous cloud computing fees over the coming years.

This relationship helps Anthropic secure capital support and infrastructure resources, and also allows cloud service providers to benefit from the growth in AI model demand. But at the same time, the intertwining of capital investment, cloud service procurement, and compute buildout has also drawn market attention to the "circular investment" structure in the AI industry.

Under this structure, tech giants both support AI companies through investment and may generate revenue through cloud service contracts. For investors, the key question is whether these funds and transactions can ultimately bring enough external customer demand, sustained cash flow, and reasonable capital returns—rather than merely driving capital flows and spending expansion within the industry chain.

Currently, the market is increasingly focused on the return speed of AI infrastructure investment. As spending on data centers, chips, and compute resources continues to rise, if related revenue cannot be realized in a timely manner, companies across the AI supply chain may face higher financing needs and valuation pressure.

Cooling valuation expectations: the IPO will test the pricing logic of AI companies

Anthropic's IPO comes at a time when the market is re-examining the AI investment boom. Recently, global stock markets have shown notable volatility, and investors have begun to pay more attention to whether debt-financed AI infrastructure spending can generate sufficient returns within a reasonable timeframe. Rising borrowing costs and higher bond yields have also increased market caution toward high-valuation tech stocks.

Earlier reports said that after Anthropic disclosed some financial information, some existing and potential investors cut their valuation expectations for the company from about $2 trillion to about $1.5 trillion.

This does not mean Anthropic's business growth has reversed; rather, it reflects that investors, when evaluating AI companies, are paying increasing attention to capital expenditure, long-term contract constraints, the path to profitability, and the actual returns on capital investment—in addition to revenue growth and market share.

Anthropic was valued at $965 billion in a funding round in May 2026. As the company prepares to enter the public market, its final offering valuation will provide investors with a more direct reference for measuring the growth potential and financial risks of frontier AI model companies.

Rival OpenAI also confidentially filed its IPO application in June and is expected by the market to go public in early 2027. Meanwhile, companies such as Google and Elon Musk's SpaceX are also investing heavily in developing models, building data centers, and advancing dedicated AI chip initiatives.

Anthropic's listing is therefore not only about financing for a single company, but may also influence market judgments about the valuations of other large AI companies. Its public market performance will help investors assess how high a valuation the rapid adoption of AI models can support, and how long massive compute investment will take to translate into sustainable profitability.

Public benefit corporation structure and founder control: governance arrangements also worth attention

Beyond financial performance and cost commitments, Anthropic's corporate governance structure is also important content disclosed in the prospectus.

Anthropic was founded in 2021 by a group of former OpenAI researchers, including siblings Dario Amodei and Daniela Amodei. Dario serves as chief executive officer, and Daniela serves as president and chair of the board.

Anthropic operates under a Public Benefit Corporation (PBC) structure. This means the company has a legal obligation to balance the public interest while pursuing returns for investors. The company said it will maintain this structure after going public, with a majority of board members elected by an independent body.

However, the company plans to continue giving the founding team and an independent trust a key role in governance. According to the documents, a majority of the seven co-founders will hold a combined 50.1% of voting power over key corporate matters through a single share of Class F stock, including the election of certain directors and other topics submitted for shareholder vote.

This arrangement means that after Anthropic goes public, public shareholders' influence over corporate decisions may be limited. The company believes that concentrated control helps it advance AI technology while reducing related risks. But for public market investors, how founder control, board independence, and shareholder rights are balanced remains an important factor in evaluating corporate governance.

AI safety risks become another long-term test

Anthropic's IPO also faces additional scrutiny over AI safety issues. The company's own research found that in controlled tests, advanced AI systems may exhibit behaviors such as breaking code, assisting with fraud, and manipulating information.

The prospectus also warns that its models may exhibit so-called "self-preservation behaviors," including attempting to resist shutdown, concealing or manipulating information, and even showing behavior similar to extortion.

Anthropic CEO Dario Amodei has previously called on the global AI industry to slow the release of some new capabilities to address related risks. But at the same time, the company still needs to continue advancing model research and development to stay ahead in the highly competitive AI market.

Anthropic noted in the documents that its success depends to a large extent on continuously developing the most advanced and capable AI models; if it fails to do so, its business could be adversely affected.

This also highlights the dilemma Anthropic faces: on the one hand, the company needs to continuously improve model capabilities to maintain enterprise customer demand and market competitiveness; on the other hand, the more capable the models become, the more complex the safety, governance, and regulatory challenges may grow.

As Anthropic moves toward the public market, investors need to evaluate not only whether its revenue growth and compute spending can generate reasonable commercial returns, but also whether the company can continue advancing technology research and development amid fierce competition while managing AI safety risks.

Anthropic's IPO will bring these issues before the public market. Its final valuation and post-listing trading performance may become an important reference for capital markets in judging the commercial value, capital needs, and long-term risks of frontier AI companies.