Anthropic seeks better terms from Amazon as AI economics shift in its favour

Anthropic's effort to renegotiate some key commercial terms with Amazon is highlighting a reality that many companies are facing: that building and launching the most advanced artificial intelligence models is proving to be pricier than many companies expected.

The AI startup is best known for its Claude family of models but has morphed from a possible OpenAI rival to one of the hottest AI providers on the block. That shift appears to have given Anthropic more leverage in its talks with Amazon, one of its original and biggest backers. Reports suggest the renegotiated terms could see Amazon pay more for Anthropic's technology than it would have under the original partnership deal.

The two companies have a long-standing relationship. The partnership expanded in April 2026 when Amazon announced a fresh $5 billion investment and the potential for billions more in future funding after already pouring billions into Anthropic since 2023. The companies also struck a massive long-term infrastructure deal, in which Anthropic will spend more than $100 billion on AWS technologies over the next decade.

The negotiations mentioned are focused on the shifting value of Anthropic's technology. When Amazon invested, Claude was still a fledgling product in a crowded AI field. Today, Claude is considered among the best frontier AI models, competing directly with products from OpenAI and Google. Anthropic is also apparently rethinking previous pricing and access terms due to this additional commercial relevance.

Also read: Anthropic restores Fable AI model after US lifts export restrictions

The development also points to a bigger problem that the AI industry is encountering. Training and inference of frontier models require enormous amounts of computing power, introducing infrastructure costs that continue to grow as adoption grows. More and more companies are shifting toward token-based pricing models that bill customers based on consumption, knowing the hefty costs of operating large-scale AI workloads.

As Amazon increasingly leans on Anthropic's models for its products, reports suggest it's become more sensitive to those costs internally. Other services, such as coding assistants and AI-powered consumer tools, are said to be heavily reliant on Claude, meaning that any increase in the cost of model access would be felt acutely there. Amazon is reportedly exploring other model providers and broadening its AI strategy as it prepares for higher AI costs.

The shifting dynamics also reflect how power in the AI ecosystem is changing. Cloud providers and investors seemed to be on the stronger side initially given the scarcity of computing infrastructure. Leading AI labs like Anthropic are building better products and larger customer bases, and they are gaining leverage of their own, which gives them the ability to negotiate more favourable commercial arrangements.

Tensions have been reported, but there is little evidence of the strategic partnership weakening. Anthropic continues to rely on AWS as its key cloud and training vendor and is increasing its use of Amazon's custom Trainium chips. More than 100,000 customers are already using Claude models via Amazon Bedrock, underscoring the importance of the tie-up to both companies.