Anthropic IPO Discloses Revenue Structure, Nearly Half of Sales Rely on Amazon and Google, Subscriptions Account for Less Than 20%

TradingKey
Yesterday

TradingKey - Anthropic's IPO prospectus further reveals the AI company's revenue structure and the extent of its reliance on Big Tech companies.

According to Reuters, Anthropic generated nearly $4.6 billion in revenue in 2025, up about 12-fold year-over-year. Of this, approximately $2.16 billion was completed through Amazon (AMZN) and Google (GOOGL) cloud marketplaces, accounting for 47% of its full-year revenue. This proportion was only 11% in 2023 and rose to 32% in 2024, indicating that as its Claude business expands, the company's reliance on cloud platform channels is also deepening.

In terms of pricing models, approximately $3.8 billion of Anthropic's revenue came from customers' actual usage of the Claude model, while subscription revenue was about $789 million, accounting for less than 20% of total revenue. The company expects usage-based pricing to remain the primary component of its future revenue.

This demonstrates a clear difference between Anthropic's business model and that of traditional software companies. Typical software companies usually rely on annual subscription contracts to generate relatively stable recurring revenue, whereas Anthropic's revenue depends more on the frequency and compute volume of customers' usage of Claude.

Nearly Half of Sales Completed Through Amazon and Google

In 2025, Anthropic generated approximately $2.16 billion in sales through Amazon and Google cloud marketplaces, accounting for 47% of its full-year revenue—a significant increase from 32% in 2024 and 11% in 2023. As the company scaled up, its reliance on the two major platforms instead continued to rise.

In return, Anthropic paid approximately $351 million in channel fees to the relevant cloud platforms. Calculated against sales, this equates to cloud service providers charging roughly $0.16 for every $1 in revenue they helped the company generate.

For Anthropic, this model significantly lowers the difficulty of enterprise customer acquisition. Many large clients are already using AWS, Google Cloud, or Microsoft Azure, and purchasing Claude services directly through cloud marketplaces reduces contract, billing, and technical deployment processes. In its prospectus materials, the company stated that these platforms possess sales networks that are difficult for a single company to replicate independently, helping Claude enter the global enterprise market more quickly.

The issue is that Amazon and Google are not ordinary distributors. They both invest in Anthropic and sell it the computing resources needed for training and inference, while also maintaining their own AI models and cloud computing businesses. This means they can access some of Anthropic's pricing and commercial terms and could potentially influence compute allocation, product promotion, and customer relationships.

Anthropic also acknowledged in its prospectus that reliance on a small number of partners and suppliers could trigger conflicts of interest and adversely affect the company's access to computing resources. Put simply, the platforms supporting Claude's expansion also control the channels and infrastructure required for the company's future growth.

These partnerships also add to the difficulty of financial comparisons. Anthropic recognizes the total contract value of customers purchasing Claude services through cloud marketplaces as revenue, while categorizing platform commissions under sales, marketing, and partnership expenses. OpenAI previously contended that this gross recognition approach inflates Anthropic's disclosed revenue figures. Anthropic stated that because it is responsible for pricing and service provision, it acts as the primary obligor in the transactions, rendering the accounting treatment compliant with established rules.

Customer Concentration and Massive Compute Commitments Amplify Operational Risks

Anthropic's customer base is similarly concentrated. In 2025, two undisclosed customers contributed approximately 12% of revenue each, together accounting for nearly a quarter of the company's total revenue. More notably, many large customers have not signed long-term contracts and can reduce or stop purchases based on their own needs.

In terms of its revenue model, approximately $3.8 billion of Anthropic's revenue came from usage-based pricing, while subscription business contributed about $789 million. The company expects that most of its future revenue will continue to come from customers' actual usage volume of the Claude model.

Anthropic's revenue structure reflects both the growth advantages and potential risks of frontier AI enterprises. On one hand, usage-based pricing enables the company to benefit directly from the growth in enterprise AI demand, while cloud platforms help Claude rapidly expand its market reach. On the other hand, the company's revenue is relatively sensitive to customer usage and a small number of channel partners, rendering its stability inferior to traditional software companies that rely primarily on long-term subscription contracts.

Meanwhile, to support model training and commercial deployment, Anthropic also needs to continuously invest heavily in computing power. As of the end of 2025, the company's non-cancelable hosting and computing resource commitments reached $54.6 billion. Although securing computing power in advance supports business expansion, it could also increase the company's fixed costs and cash flow pressure if future demand growth falls short of expectations.

Find out more

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Most Discussed

  1. 1
     
     
     
     
  2. 2
     
     
     
     
  3. 3
     
     
     
     
  4. 4
     
     
     
     
  5. 5
     
     
     
     
  6. 6
     
     
     
     
  7. 7
     
     
     
     
  8. 8
     
     
     
     
  9. 9
     
     
     
     
  10. 10