Amazon suspends development of Inferentia AI inference chip, focusing on Trainium training chip
Amazon has stopped developing Inferentia artificial intelligence (AI) chips and instead focused on Trainium chips for training AI models. The company believes that this field provides a better way to compete with AI chip leader Nvidia.
Since announcing its entry into the AI chip field in 2018, Amazon AWS has developed Inferentia chips customized for inference and prediction, as well as Trainium chips for AI model training.
Rahul Kulkarni, Director of Computing at Amazon AWS, said, "The two product lines will merge. We will focus on Trainium to provide inference and training performance
Inferentia focuses on performing fewer calculations at lower costs. As generative AI becomes increasingly advanced, both training and inference require greater computational resources. Therefore, the benefits of using different chips are gradually disappearing.
Trainium has a larger memory capacity and supports a wider range of data formats. It also includes mechanisms for fast computation and communication when processing large amounts of data using multiple servers simultaneously.
At the annual technology event held in Las Vegas, Amazon announced that the Trainium2 chip has been launched. The company also announced plans to release the Trainium3 chip in the second half of 2025. Trainium3 will be manufactured using advanced 3nm technology, and its computational performance will be doubled compared to Trainium2.
Amazon's goal is to break Nvidia's dominant position in the AI chip market, which currently holds approximately 90% of the global AI chip market share. Amazon has been vigorously promoting Trainium2 to cloud customers, emphasizing that the operating cost of the chip is lower than that of Nvidia products.
Apple has stated that it will use the Trainium2 chip for its own AI development. Anthropic, a US AI startup, has received an $8 billion investment from Amazon and is acquiring hundreds of thousands of Trainium2 chips to assist in the development of its generative AI models.
Rahul Kulkarni, Director of Computing at Amazon AWS, said, "The two product lines will merge. We will focus on Trainium to provide inference and training performance
Inferentia focuses on performing fewer calculations at lower costs. As generative AI becomes increasingly advanced, both training and inference require greater computational resources. Therefore, the benefits of using different chips are gradually disappearing.
Trainium has a larger memory capacity and supports a wider range of data formats. It also includes mechanisms for fast computation and communication when processing large amounts of data using multiple servers simultaneously.
At the annual technology event held in Las Vegas, Amazon announced that the Trainium2 chip has been launched. The company also announced plans to release the Trainium3 chip in the second half of 2025. Trainium3 will be manufactured using advanced 3nm technology, and its computational performance will be doubled compared to Trainium2.
Amazon's goal is to break Nvidia's dominant position in the AI chip market, which currently holds approximately 90% of the global AI chip market share. Amazon has been vigorously promoting Trainium2 to cloud customers, emphasizing that the operating cost of the chip is lower than that of Nvidia products.
Apple has stated that it will use the Trainium2 chip for its own AI development. Anthropic, a US AI startup, has received an $8 billion investment from Amazon and is acquiring hundreds of thousands of Trainium2 chips to assist in the development of its generative AI models.