Recently, at the re:Invent 2025 conference, Amazon Web Services (AWS), Amazon's cloud computing division, introduced its latest self-developed AI chip, Trainium 3. This new chip is crafted using a cutting-edge 3-nanometer manufacturing process. In terms of performance, it boasts a computational prowess that is 4.4 times greater than its previous iteration. Moreover, its memory capacity has seen a fourfold increase, and it has achieved a 40% boost in energy efficiency. These enhancements collectively translate to significant cost reductions for enterprises engaged in AI model training and inference processes.

In addition to the chip launch, AWS also unveiled its AI Factory service. Furthermore, the company has decided to deepen its collaborative efforts with NVIDIA. Through this strengthened partnership, both entities aim to expedite the deployment of AI infrastructure. This move is strategically designed to cater to the data sovereignty requirements of enterprises and governmental bodies, ensuring compliance and security in the rapidly evolving AI landscape.