Amazon Com Stock Price, News & Analysis

Amazon (NASDAQ: AMZN) announced new AI customization features at AWS re:Invent on December 3, 2025 in Las Vegas. The update adds Reinforcement Fine Tuning (RFT) to Amazon Bedrock and serverless model customization with reinforcement learning to Amazon SageMaker AI.

The company says Bedrock RFT makes it easier to tailor models to unique cases and improve accuracy, while SageMaker AI shortens advanced customization workflows from months to days, reducing costs and accelerating AI development.

Supabase and AWS (AMZN) announced new Amazon S3-based storage products and a one-click ETL at AWS re:Invent on December 3, 2025 to help developers scale prototypes into production.

The launch adds Supabase Analytics Buckets (Apache Iceberg on Amazon S3 Tables) and Supabase Vector Buckets for large vector datasets, plus Supabase ETL to replicate Postgres data into analytics and vector storage in near real-time. The platform serves 5 million developers, has launched >10 million databases, runs across 17 AWS Regions, and operates on AWS Graviton processors to improve performance and cost-efficiency.

Amazon (NASDAQ: AMZN) announced enhancements to Amazon Bedrock AgentCore at AWS re:Invent on December 2, 2025.

Key features: Policy enforces real-time deterministic controls that block unauthorized agent actions outside agent code; AgentCore Evaluations enables continuous behavior-based quality inspection; AgentCore Memory adds episodic memory to help agents learn from experience and improve decision-making. The release lists enterprise users that trust AgentCore to accelerate agent deployment into production.

Amazon (NASDAQ: AMZN) announced general availability of Trainium3 UltraServers, powered by the new Trainium3 3nm AI chip, at AWS re:Invent on December 2, 2025. Trainium3 UltraServers deliver up to 4.4x more compute performance, 4x greater energy efficiency, and nearly 4x more memory bandwidth versus Trainium2 UltraServers.

Trn3 UltraServers scale to 144 Trainium3 chips and offer up to 362 FP8 PFLOPs with 4x lower latency for training larger models and serving inference. AWS cited customer cost reductions up to 50% on Trainium and specific examples of 4x faster inference at half the GPU cost for real-time generative video. Amazon Bedrock is already running production workloads on Trainium3.

Amazon Web Services (NASDAQ: AMZN) announced AWS AI Factories at AWS re:Invent on December 2, 2025. AWS AI Factories provide dedicated AI infrastructure that integrates the latest NVIDIA accelerated computing, Trainium chips, AWS AI services, and high-speed, low-latency networking.

The offering enables customers to use existing data center space, power, and network connectivity while AWS manages deployment and operations. AWS positions the solution to accelerate AI application development, support data sovereignty and regulatory requirements, and shorten deployment timelines for enterprises and public sector organizations.

Amazon Web Services (NASDAQ: AMZN) unveiled Frontier Agents at AWS re:Invent on December 2, 2025, introducing three autonomous AI agents: Kiro autonomous agent, AWS Security Agent, and AWS DevOps Agent.

Frontier Agents are described as a new class of AI agents that are autonomous, scalable, and able to operate for hours or days without constant human intervention. Kiro acts as a virtual developer that maintains context and learns over time. AWS Security Agent functions as a virtual security engineer for design reviews, code review, and penetration testing. AWS DevOps Agent helps resolve and prevent incidents and improve application reliability and performance.

Amazon (AMZN) expanded its Nova AI portfolio on December 2, 2025 with four new Frontier Nova models, a hosted training service called Nova Forge to let organizations build custom Nova variants by infusing proprietary data early in training, and Nova Act for building browser-based agents.

Key metrics: Nova Act reported 90% reliability on browser UI automation workflows for early customers. Amazon also described Nova 2 models as offering industry-leading price-performance across reasoning, multimodal, conversational AI, code generation, and agentic tasks.

WRITER announced a native integration with Amazon Bedrock and launched an agent supervision and orchestration suite to add model flexibility, governance, and security to its enterprise agent platform effective December 2, 2025. Customers can deploy Bedrock-hosted models or WRITER’s Palmyra models into prebuilt or custom agents with no additional setup, using WRITER’s no-code and pro-code tools under a unified observability and guardrails framework. The supervision suite adds centralized agent approval, continuous production supervision, user/agent behavior visibility, and interoperability with existing observability and security tools.

WRITER cited enterprise customers including Vanguard, Qualcomm, and AstraZeneca as users of the platform.

Amazon (NASDAQ: AMZN) announced at AWS re:Invent that Sony is adopting AWS AI services to scale enterprise AI and build a Sony Engagement Platform connecting fans and creators.

Sony uses Amazon Bedrock AgentCore and Nova Forge, processes 150,000 daily inference requests with an expected 300-fold increase, and supports 57,000 employees using AI agents. The Sony Data Ocean handles up to 760 TB from over 500 dataset types. The initiative targets faster content workflows and deeper fan engagement across Sony's businesses.