On October 19, Amazon’s stock rose about 5%, making it one of the standout large-cap tech names of the day. Morgan Stanley put Amazon on its focus list, citing AWS’s growth potential and the company’s positioning in AI infrastructure.
At the same time, Amazon(AMZN 0.00%↑) launched TorchNeuron, which aims to let developers run existing PyTorch programs on its in-house Trainium chips with fewer code changes.
Over the past few years, the main investment theme in the AI industry has centered on GPUs, advanced process nodes, and high-speed interconnects, with Nvidia building a significant advantage through its CUDA software ecosystem. But as large-model training and inference move into large-scale deployment, enterprises buying compute are paying more attention to total cost, supply stability, and deployment efficiency.
For cloud providers, in-house chips offer more room for resource scheduling, while software compatibility determines whether those chips can actually enter customers’ production environments.
