Amazon.com has announced the development of Trainium2, the second generation of its artificial intelligence (AI) chip designed specifically for training AI systems in its cloud computing service, Amazon Web Services (AWS). The Trainium2 chip boasts a fourfold increase in speed compared to its predecessor while demonstrating twice the energy efficiency. This move comes in response to the escalating competition between Amazon and Microsoft to dominate the AI chip market, with Microsoft recently introducing its own AI chip named Maia.

AWS Chief Executive Adam Selipsky unveiled Trainium2 at a conference in Las Vegas, highlighting its enhanced performance capabilities. The chip is expected to compete against AI chips from Google, notably Google’s Tensor Processing Unit (TPU), which has been available to cloud computing customers since 2018. Amazon’s Trainium2 aims to meet the rising demand for computing power in developing technologies such as large language models, essential for services similar to ChatGPT.

The announcement also included details about Graviton4, AWS’s fourth custom central processor chip, which is reported to be 30% faster than its predecessor. Both Amazon and Microsoft have opted for technology from Arm Ltd in their chip designs, marking a departure from traditional chip providers like Intel and Advanced Micro Devices in the cloud computing sector.

The Trainium2 chip is set to be available on AWS starting next year, emphasizing Amazon’s commitment to offering advanced solutions for AI training. This development aligns with the broader trend in the tech industry of companies investing in custom chips to address the growing demand for specialized computing power in various applications.