HomeArticle

The grand feast kicks off, and VCs have raked in staggering returns.

36氪的朋友们2026-11-20 11:38
300 million yuan per capita: 5,300 employees have collectively achieved financial freedom.

The largest IPO in history has arrived.

On November 19, Anthropic was listed in the United States with the stock ticker ANT. Calculated on a fully diluted basis, the company's market value once exceeded the USD 2 trillion threshold, and stood at over USD 1.7 trillion (more than RMB 11 trillion) at the close of the first trading day. Through public offering, private placement and other methods, Anthropic's total financing scale reached hundreds of billions of US dollars, surpassing SpaceX to become the largest IPO in history.

Led by Dario Amodei and Daniela Amodei, the post-80s brother and sister, together with a group of former OpenAI colleagues, they built Anthropic from scratch into a global giant in just over 5 years, and created amazing wealth at the same time - the shareholding value of all 7 co-founders exceeded USD 10 billion.

Meanwhile, a group of employees will also achieve financial freedom.

The prospectus shows that the company still has 1.774 billion unexercised options and another 1.07 billion unvested RSUs. Estimated at the closing price of USD 97.17, the potential value of this batch of equity exceeds USD 2700 billion. At present, Anthropic has about 5300 employees. If simply averaged according to this number, it is equivalent to that each person holds equity worth more than RMB 300 million.

Investors have also reaped rich returns. From early investors such as Eric Schmidt, former chairman and CEO of Google, to global giants such as Google, Amazon and NVIDIA, as well as investment institutions such as HSG, Spark Capital and DST Global, and sovereign wealth funds such as GIC and Qatar Investment Authority, all are shareholders of Anthropic.

Amid the ringing of the IPO bell, they share this huge wealth feast.

Former OpenAI employees built a trillion-dollar IPO

As the number one rival of OpenAI, the birth of Anthropic actually originated from a group of former OpenAI employees.

In 2021, Dario Amodei and his sister Daniela Amodei co-founded Anthropic with former OpenAI colleagues including Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish and Christopher Olah.

Dissatisfaction with OpenAI's accelerated commercialization and anxiety about the neglect of AI safety became the reasons for their collective departure. In the view of the founding team, as the capabilities of artificial intelligence continue to improve, safety research and technology governance need to be given more sufficient attention.

Behind this choice, it is closely related to the experience of the brother and sister.

Dario Amodei studied at Stanford, Caltech, Princeton and other universities, and worked at Baidu, Google and other companies. In 2016, he joined OpenAI to engage in AI safety research and participated in the R&D of cutting-edge models such as GPT-2 and GPT-3.

Following in her brother's footsteps, Daniela Amodei, who studied English literature at the University of California, Santa Cruz, joined OpenAI in 2018 and eventually became the company's vice president in charge of safety and policy.

One understands models and the other understands governance, and this complementarity constitutes the original background of Anthropic. At the beginning of its establishment, Anthropic defined itself as an artificial intelligence safety and research company, with the goal of building "reliable, interpretable and controllable" AI systems.

By 2023, Anthropic officially launched Claude, and began to transform its previous model research capabilities into market-oriented products and services. With the continuous improvement of model capabilities, Anthropic's business has gradually expanded from providing general AI models to specific scenarios such as enterprise development, software programming and office work.

Among them, Claude Code has become an important product in Anthropic's commercialization process, and is regarded by the company as a "virtual collaborator" that can plan, write, test and debug code. Different from simply generating code through dialogue, Claude Code can read code libraries, modify files, execute commands, and continuously complete a series of tasks such as planning, writing, testing and debugging.

Changes in product capabilities have directly affected Anthropic's business model. As the task chain that Claude Code can execute continues to extend, the way enterprises use AI has gradually shifted from one-time Q&A to continuous workflow. In this process, the model is called repeatedly, and Token consumption also increases as the task progresses.

For Anthropic, this means that enterprise customers' expenditure is no longer limited to fixed subscription fees, but will expand with the actual usage. In the first half of 2026, about 80% of the company's revenue came from usage-based billing, amounting to USD 13.1 billion, and the remaining about 20% came from subscription revenue of products such as Claude.

The recognition of the market has also made its annualized revenue form a steep growth curve.

According to Dario Amodei's previous statement, Anthropic's revenue has grown about 10 times every year since it obtained its first revenue. From about USD 10 million in 2022 to about USD 1 billion in December 2024, this figure has reached about USD 810 billion by the end of October 2026.

One of the reasons driving this round of performance explosion is that more and more enterprises begin to use Claude for real business and expand the scale of model calls at the same time. Anthropic itself is also an important user of Claude Code - in September 2026, more than 90% of the code submitted by the company to the internal code library was written by Claude.

In order to seize more markets, Anthropic is also continuously expanding its product matrix. At present, the company has formed model series such as Fable, Mythos, Opus, Sonnet and Haiku. Approaching the IPO, they intensively launched products such as Fable 5.2 and Mythos 5.2, and further extended AI capabilities to non-programming office fields.

From focusing on AI safety research at the time of its establishment, to launching Claude in 2023, and then achieving revenue explosion by relying on various agents, Anthropic has realized the transformation from a technology research institution to a commercial AI company in more than five years, and finally grown into a trillion-dollar global giant.

The wealth feast kicks off, investors have earned huge returns

With the recognition of many investors, Anthropic has raised financing all the way to Series H.

Shortly after its establishment in 2021, Jaan Tallinn, co-founder of Skype, Dustin Moskovitz, co-founder of Facebook, and Eric Schmidt, former chairman and CEO of Google, invested USD 124 million in this entrepreneurial team that left OpenAI.

At that time, Anthropic had not yet launched Claude, nor did it have the huge group of enterprise customers that followed. What investors valued more was the founding team's experience in the field of cutting-edge artificial intelligence research and their judgment on the development direction of AI technology.

This early fund provided important support for Anthropic to form a team and carry out model R&D. However, as the competition of large models continues to escalate, Anthropic soon needs more capital.

In 2023, Google took the lead in investing USD 300 million in Anthropic. After that, they completed the Series C financing of USD 450 million with the participation of Spark Capital and other institutions. In September of the same year, Amazon announced that it would invest USD 40 billion in the company, and then Google decided to make an additional investment of USD 2 billion.

Google and Amazon are not only capital providers, but also important cloud computing service providers, which can support Anthropic's model training and deployment through their own cloud infrastructure. This cooperative relationship combining capital and computing power provides important conditions for Anthropic's subsequent expansion.

After Claude entered the commercialization stage, their financing speed further accelerated.

In March 2025, Anthropic completed USD 35 billion in financing, with investors including Lightspeed, Bessemer Venture Partners, Cisco Investments, D1 Capital Partners, Fidelity Management & Research Company, General Catalyst, Jane Street Capital, Salesforce Ventures, etc. Half a year later, they received USD 130 billion from Lightspeed, Blackstone, Goldman Sachs, Qatar Investment Authority and others.

When it came to 2026, the company successively completed two rounds of heavyweight financing in the first half of the year, and its valuation soared from USD 3.8 trillion to USD 9.65 trillion. Among them, the investors of their USD 300 billion Series G financing include GIC, Coatue, Google, Amazon, HSG, Temasek, Microsoft, NVIDIA and others.

As for the USD 650 billion Series H financing, it is supported by Altimeter Capital, Dragoneer, Greenoaks, HSG, Capital Group, Coatue, D1 Capital Partners, GIC, Blackstone, DST Global, MGX, Lightspeed, Amazon, Micron, Samsung, SK Hynix and others.

As the listing plan advanced, the company began to seek larger-scale capital from the public market. In addition to publicly issuing shares, Anthropic will also conduct a USD 260 billion private placement to companies such as Google, NVIDIA, Microsoft, Broadcom, Micron, TSMC and AMD.

After the completion of the IPO and the simultaneous private placement, Amazon and Google are expected to hold nearly 15% and 14% of Anthropic's shares respectively, becoming important shareholders of the company. After years of continuous investment, the two tech giants have finally reaped rich returns by providing capital and computing power support for Anthropic.

For other investors, they have also reaped rich returns. From the valuation of tens of billions of dollars in the early stage to the valuation of USD 9.65 trillion half a year ago, the return on investment at different stages varies greatly. However, in the face of a market value of USD 1.7 trillion, even investors who entered in the last round before listing have considerable returns.

Along with the ringing of the listing bell, the seven co-founders also cashed out by selling part of their shares. According to media reports, they will sell about 3.3784 million shares each, totaling about 23.65 million shares sold, and cashing out hundreds of billions of yuan in total.

Although the shareholding has decreased, the control of the company is still in the hands of the founding team. In Anthropic's plan, in addition to continuing to maintain the status of a public benefit company, they will also lock in about 50.1% of the voting rights for the seven co-founders through Class F shares held by Founder LLC.

In the AI era, IPOs have crossed the trillion-dollar threshold

As the new wave of artificial intelligence sweeps the world, trillion-dollar IPOs are being born in batches.

From SpaceX to Anthropic, the interval between their listing dates is less than half a year, but their market value has exceeded one trillion US dollars. There are market rumors that OpenAI, which has embarked on the road of IPO, is conducting a new round of financing with a pre-money valuation of USD 1.25 trillion and is expected to be listed in 2027.

In the listing stories of these companies, a common feature is that AGI can create great value.

SpaceX stated in its prospectus that it has locked in the largest addressable market in human history - the total potential value is as high as USD 28.5 trillion, of which the AI sector accounts for USD 26.5 trillion.

As for Anthropic, the scenario depicted by this company is that the economic value created by global knowledge work in 2025 exceeds USD 40 trillion, and their opportunity is to improve productivity and accelerate economic growth through AI.

It is precisely because of the great expectation for these potential markets that investors are willing to give leading AI companies valuations far higher than traditional enterprises. However, as concerns about the AI bubble continue to rise and the AI industry enters the stage of large-scale commercialization, the capital market has also begun to re-examine these long-term expectations.

Since the submission of the prospectus, Anthropic has also faced valuation divergences.

Previously, there were rumors in the market that the company was expected to reach a valuation of USD 2 trillion. After the external environment changed, coupled with signs of slowing performance, the valuation expectation for this IPO has also declined.

In the third quarter of 2026, they achieved revenue of about USD 169.6 billion, a year-on-year increase of more than 1100%. However, from the perspective of quarter-on-quarter growth rate, the revenue in this quarter increased by about 47% compared with the second quarter, which was significantly lower than the about 144% in the second quarter.

At the same time, Anthropic's profit performance also fluctuated. In the second quarter of 2026, the company achieved an operating profit of about USD 974 million, but turned to an operating loss of about USD 567 million in the third quarter. In addition, affected by factors such as changes in the fair value of convertible notes and derivative liabilities, the company's net loss in the first three quarters of 2026 was about USD 1040 billion.

Compared with the book loss, what is more noteworthy is the continuously growing computing power expenditure of Anthropic.

When more and more enterprises connect Claude to their businesses and continuously run programming agents and other AI tools, Anthropic needs to provide more inference computing power. In order to maintain model competitiveness, they must also continuously develop and train the next generation of models, and continue to invest in huge-scale training resources.

As the business scale continues to expand, the demand for computing power will continue to exist. According to Anthropic's disclosure, their cumulative future computing power procurement and lease commitments related to partners such as Amazon and Google have exceeded USD 900 billion.

However, the huge investment in computing power and profit pressure have not stopped investors' enthusiasm.

According to media reports, the institutional investor subscription demand for this company is about USD 4000 billion, while the retail demand is also close to USD 1000 billion. This also prompted Anthropic to raise its pricing from USD 70-78 per share to over USD 80, and expand the number of shares issued at the same time.

For Anthropic, achieving USD 810 billion in annualized revenue in five years only proves that cutting-edge AI technology has huge commercialization potential. Next, while maintaining model competitiveness, controlling computing power costs and improving operating performance are the keys to supporting its trillion-dollar market value.

This article is from the WeChat Official Account "ChinaVenture", written by Lu Zhigao, edited by Wang Qingwu, and published by 36Kr with authorization.