Amazon announced a new service called AI Factories that lets large enterprises and governments run AWS AI workloads inside their own data centers. The offering combines Amazon’s cloud software stack with Nvidia hardware, allowing customers to choose between Nvidia’s latest Blackwell GPUs or Amazon’s Trainium3 chips. AI Factories are positioned as a solution for data‑sovereignty concerns, giving organizations full control over data and hardware while still accessing AWS services such as Bedrock and SageMaker. Similar private‑cloud moves are also being pursued by Microsoft. Leia mais →

Amazon has introduced a new AI-powered feature called Help Me Decide, aimed at guiding indecisive shoppers toward a purchase. The tool appears on product pages in the U.S. when a user has been browsing without committing, and it draws on browsing history, search activity, and prior purchases to suggest a primary product, an upgrade option, and a budget-friendly alternative. Help Me Decide relies on Amazon's Bedrock and SageMaker machine‑learning platforms along with OpenSearch to compile recommendations and incorporate customer reviews. The feature is currently live in the Amazon app and mobile browsers. Leia mais →

Amazon has added a new AI-powered tool called “Help me decide” to its shopping experience. The feature analyzes a shopper’s searches, browsing and purchase history to surface product recommendations and explain why each item fits the user’s needs. Built on Amazon’s own large language models and services such as Bedrock, OpenSearch and SageMaker, the tool appears after a user has viewed multiple similar listings and is currently available in the United States on the Amazon Shopping app and website. It follows a series of earlier AI experiments, including the Rufus assistant, AI‑driven shopping guides, audio summaries and the Lens Live visual search tool. Leia mais →

OpenAI has released two open-weight large language models, gpt-oss-120B and gpt-oss-20B, and made them available through Amazon Web Services’ Bedrock and SageMaker platforms. The models run on modest hardware, with the larger model operating on a single 80 GB GPU and the smaller model requiring only 16 GB of memory. Licensed under Apache 2.0, they aim to lower barriers for developers and support high‑security or resource‑constrained environments. Integrated with Amazon Bedrock AgentCore, the models support AI agents, a 128K context window, and popular developer tools. Leia mais →