Amazon is negotiating a second billion-dollar investment in AI company Anthropic. This time, however, Amazon is setting conditions: Anthropic should increase its use of Amazon's AI chips instead of Nvidia's.
Sources familiar with the negotiations told The Information that Amazon is in talks with OpenAI competitor Anthropic about a second major investment worth billions. The deal's structure would mirror the first $4 billion investment from last year.
According to The Information, Amazon ties the investment to one key condition: Anthropic must commit to using servers with Amazon's own Trainium chips more extensively for training its AI models, rather than the previously preferred Nvidia chips.
A person involved in the talks indicates that the total amount of Amazon's investment could directly depend on how many Amazon chips Anthropic agrees to use. The current status of negotiations remains unclear.
Technical challenges complicate the deal
The proposed switch to Amazon chips could present technical challenges for Anthropic: The software required for Amazon's Trainium chips is less mature than Nvidia's established CUDA platform.
Additionally, such a move could tie Anthropic to Amazon's ecosystem, as Amazon does not make its hardware available for other providers' facilities. This could limit Anthropic's options to use other cloud providers or operate its own data centers.
For Amazon, having Anthropic use its own chips makes strategic sense: It would reduce the cloud giant's dependence on expensive Nvidia hardware. Under the first deal, Anthropic had agreed to use some Trainium servers, but primarily relied on Nvidia hardware in Amazon's data centers.