Anthropic, the operator of the Claude AI model, has renegotiated its contract with early investor Amazon to raise its AI usage fees, according to reports. In response, Amazon has drawn up alternatives that leverage OpenAI and its own AI model "Nova" to cut costs, but observers say AI operators are gaining greater leverage.
The Information, a technology news outlet, reported Saturday, citing multiple sources, that Anthropic renegotiated part of its contract with Amazon earlier this year, changing the terms so that fees for its AI models would increase. The contract takes effect next year.
Anthropic changed the fee basis from the compute hours Amazon used for the AI models to "tokens," the units of information processed by AI. Sources believe the change could significantly increase Amazon's AI costs.
The relationship between Amazon and Anthropic dates back to 2023, when Amazon invested $4 billion (about 6 trillion won). At the time, Anthropic chose Amazon Web Services (AWS) as its primary cloud provider and agreed to use Amazon's AI chips "Trainium" and "Inferentia." In turn, AWS provided Anthropic with computing resources for AI model training while also connecting it with large enterprise customers, completing a solid cooperative framework. Amazon uses Claude models across its enterprise and consumer AI services, including the shopping assistant "Alexa for Shopping."
However, the two appear to be drifting apart, as Amazon executives' concerns over deepening over-reliance clash with political interests. Amazon has also struggled to secure Claude-level performance in developing its own large language model (LLM). AWS executives are reportedly concerned that if their AI services rely too heavily on Claude rather than Nova, it could give the market the impression that "Amazon products are essentially just wrappers around Anthropic technology."
Amazon CEO Andy Jassy also triggered the U.S. government's export ban on Anthropic's latest AI models, Mythos and Fable. He reportedly expressed concern to White House officials about the possibility of circumventing the safeguards of Anthropic's latest AI models.
In response, Amazon is strengthening cooperation with Anthropic's competitor OpenAI. Amazon announced earlier this year that it would invest up to $50 billion (77 trillion won) in OpenAI. The Information reported that Amazon is planning to make additional use of OpenAI in response to Anthropic's fee increase. To prepare for the possibility that Anthropic's AI model prices could rise further, some Amazon engineers are also reportedly working on distilling Claude into smaller AI models (distillation, a machine learning technique that creates a fast, small "student" model based on a high-performance model).
The relationship with Amazon also comes at a high cost for Anthropic. When Claude models are sold through AWS, Anthropic reportedly pays cloud usage fees and, in addition, hands over half of the remaining total profit to Amazon.
