Chart - Anthropic's Initial Public Offering: An In-depth Analysis of the Rapid Growth and Rising Costs of this AI Leading Company

Update charts and text, including financial data for the first half of 2026 on revenue and operating loss

Deborah Mary Sophia

Reuters, October 26 - Anthropic on Monday released its US Initial Public Offering (IPO) filing (link), paving the way for one of the most sensational debuts on the market. As the market grows increasingly concerned about the company's massive debt-driven spending, this IPO will test whether investors are willing to pay for the AI giant's outlays.

The company had previously filed a confidential statement in June (link). In recent weeks, Reuters' exclusive reports have highlighted its ambitious AI aspirations (link), surging costs (link), and concerns about so-called “existential risks to humanity (link)” posed by advanced AI.

Below are some details regarding Anthropic’s financial status, business structure, and its position in the global artificial intelligence race:

Background and Financial Status

Anthropic, founded in 2021 by former OpenAI researchers including siblings Dario Amodei and Daniela Amodei, is the start-up behind the “Claude” series of AI models. The Claude models are divided into four levels: Fable, Opus, Sonnet, and Haiku, supporting Anthropic’s chatbot, programming agents, and enterprise-focused Claude Cowork AI agents.

Anthropic initially focused on training models for programming tasks, propelling the company to the forefront of the AI race and helping to attract a steady stream of enterprise customers.

Most of its revenue comes from enterprise usage-based models, measured in information-processing “tokens” and referred to by Anthropic as “usage-based billing” revenue. Additionally, the company generates income through paid customer subscription packages like “Pro” and “Max.”

This helped Anthropic’s revenue grow from $386 million a year earlier and $41 million in 2023, to $4.59 billion in 2025. In the six months ending June 2026, Anthropic’s revenue reached $16.29 billion, a substantial increase from $1.16 billion during the same period in 2025.

Over the past years, the company’s operating losses have continued to widen. But in the first half of 2026, the operating loss narrowed to $693 million, down from $2.14 billion in the first half of 2025, as a fourteen-fold revenue increase offset a fivefold rise in total expenses.

https://www.reuters.com/graphics/ANTHROPIC-IPO/lgvdbeookvo/chart.png

Costs, Computing Needs, and Circular AI Investment

Computing power is Anthropic’s largest expense.

The company’s costs include computing fees paid to cloud service providers such as Amazon Web Services (AWS) and Google Cloud, costs related to data center facilities and leased data center capacity, equipment leasing fees, and electricity expenses.

Over the past three years, Anthropic’s computing and infrastructure spending has far outpaced its revenue.

https://www.reuters.com/graphics/ANTHROPIC-IPO/zdpxeglrbvx/chart.png

The company has committed to pay over $600 billion to about eight partners in the next decade (link)—one of the largest AI infrastructure undertakings ever. Anthropic states that many sections of its agreements are non-cancellable or must be paid regardless of actual usage.

Over the next seven to ten years, the company is committed to paying Amazon AMZN.O at least $110 billion, Google GOOGL.O about $114.5 billion, and Microsoft $44.9 billion for cloud computing services.

https://www.reuters.com/graphics/ANTHROPIC-IPO/zjvqlmxwopx/chart.png

But some of these companies are also its largest backers. Both Amazon and Google are major shareholders in Anthropic, illustrating again how circular financing structures dominate capital flows in the AI industry.

Amazon has invested at least $13 billion in Anthropic and agreed to reinvest up to $20 billion more. Google has pledged up to $40 billion in investments to Anthropic.

https://www.reuters.com/graphics/ANTHROPIC-IPO/zdvxeglqwpx/chart.png

Governance and Structure

Anthropic is a Public Benefit Corporation (PBC), meaning it is legally obliged to both achieve profits for investors and develop AI that benefits humanity. Most of its board members are elected by an independent institution.

The company is jointly led by CEO Dario Amodei and his sister Daniela—who also serves as president and chair of the board—and will continue to operate as a public benefit corporation after listing.

The founders and an independent trust are expected to retain control, a structure that may limit investor oversight and influence from public shareholders. The company maintains that to reduce the risks of artificial intelligence, it is necessary to concentrate control and capital in the hands of a few.

Under this plan, the majority of the seven co-founders will decide the voting of a single Class F share (which carries 50.1% of total voting power) on key corporate matters, including electing some directors and other matters put forward to investors.

Competition

Anthropic’s most direct rival is OpenAI, the developer of ChatGPT, whose launch in 2022 triggered a global race to adopt and advance artificial intelligence technologies.

Following Anthropic’s lead in the enterprise market, OpenAI has shifted its focus further towards developer- and enterprise-oriented products.

https://www.reuters.com/graphics/ANTHROPIC-IPO/dwvkmnkdrvm/chart.png

Anthropic is also in competition with Elon Musk’s SpaceX, Google under Alphabet, and Meta META.O. These companies are each investing tens of billions of dollars on model development, data centers, and specialized AI chips, fighting for supremacy in the generative AI sector.

Ramp data (reflecting a portion of enterprise AI spending) shows that for the week ending October 11, Anthropic accounted for 60.1% of AI token expenditures.

https://www.reuters.com/graphics/ANTHROPIC-IPO/akpelroadvr/chart.png

(For the convenience of non-English speakers, Reuters automatically translates its reports into several other languages. As automated translation may contain errors or lack required context, Reuters does not guarantee the accuracy of automated translation, which is offered only to aid readers. Reuters assumes no responsibility for any damage or loss arising from the use of automated translation features.)