Speaking at the company’s annual AWS re:Invent 2025 event in Las Vegas, CEO Matt Garman said Trainium is already a multibillion-dollar business, with 1 million Trainium chips deployed to date.
Trainium is AWS’ in-house chip used for both training and inference, positioning the company to compete with Nvidia, the market leader in AI training.
Customers using Trainium 3 include Anthropic, Karakuri, Metagenomics, Neto.ai, Ricoh, Splashmusic, and more. According to AWS, Trainium can reduce training and inference costs by up to 50%.
Additionally, the company is ramping up its hiring of AI agents. Garman said the emergence of agents represents an important turning point and that the true value of AI is still waiting to be unlocked. He added that reaching the age of agents requires rethinking all processes, which requires powerful AI at the lowest possible cost.
AWS’s AI chip efforts are a step in this direction. The company says it is also partnering with Nvidia to expand its computing portfolio with P6e-GB300 UltraServers powered by Nvidia’s cutting-edge GPU architecture on the AWS cloud.
The company also expanded Nova, AWS’s suite of AI models, with four new models focused on inference, multimodal processing, conversational AI, and code generation.
(The reporter is attending AWS re:Invent 2025 in Las Vegas at the invitation of AWS)
