Will Amazon.com, Inc. (AMZN.US), Alphabet Inc. Class C (GOOGL.US), and Microsoft Corporation (MSFT.US) emerge as the biggest winners? Anthropic could potentially pay over 80 billion US dollars in cloud fees by 2029.

According to media reports, artificial intelligence startup Anthropic is expected to pay at least $80 billion to Amazon.com, Inc. (AMZN.US), Alphabet Inc. Class C (GOOG.US, GOOGL.US), and Microsoft Corporation (MSFT.US) by 2029 to run its Claude artificial intelligence model on the above-mentioned cloud computing platforms. However, this is not the only way the three major cloud providers will profit from Anthropic. Reports suggest that according to Anthropic's latest financial forecasts, these cloud service providers will also receive a share of the revenue from Anthropic's AI sales. If their enterprise customers purchase Anthropic's AI products through the cloud platform, the cloud providers will receive a certain percentage of revenue sharing. This revenue source is rapidly growing, with the amount paid to cloud service providers for AI sales reaching only about $1.3 million in 2024, but expected to jump to around $360 million in 2025, and expand to $1.9 billion in 2026 and $6.4 billion in 2027. This revenue-sharing mechanism is seen as an important incentive for Anthropic's cloud partners. For example, Microsoft Corporation has encouraged salespeople in its Azure cloud business to promote Anthropic's models and include related sales in performance evaluations. Analysts believe that such revenue sharing expenses account for about one-tenth of Anthropic's predicted total revenue in the coming years, significantly impacting the company's finances. Additionally, in the business of reselling AI through Amazon.com, Inc.'s cloud technology (AWS), Anthropic is required to give up about 50% of its gross profit to Amazon.com, Inc. As cloud computing power procurement scales change, this ratio may be adjusted in the future. Anthropic's management believes that by partnering with the three major cloud providers, it has a competitive advantage compared to OpenAI in reaching enterprise customers. In terms of computing power input, Anthropic estimates that by 2029, expenditures related to model training alone could reach as high as $100 billion, highlighting the increasingly high costs of cloud computing and chip costs behind generative AI.