AWS re:Invent 2025: key announcements for businesses

Agentic AI at the heart of re:Invent 2025

AWS re:Invent 2025 marked a turning point in AWS’s strategy. For the first time, AI is no longer presented as a set of assistance tools but as autonomous agents capable of working independently for hours. The announcements cover hardware (Graviton5, Trainium3), models (Nova 2), agents (DevOps Agent, Security Agent) and development tools (Kiro). Here are the announcements that matter for businesses.

Graviton5: AWS’s most powerful processor

Graviton5 is the fifth generation of the ARM processor designed by AWS. With 192 cores per chip, a cache 5 times larger and performance 25% higher than Graviton4, it offers the best price-performance ratio for general-purpose workloads. The new EC2 M9g instances based on Graviton5 are available in the major regions.

For businesses, Graviton5 means higher performance at the same cost or lower costs at the same performance. Migrating from Intel or AMD instances to Graviton is seamless for most Linux applications.

Amazon Nova 2: the new AI model family

Amazon Nova 2 replaces and surpasses the Titan family. Nova 2 Lite is optimized for speed and cost. Nova 2 Pro offers a good price-performance balance for complex tasks. Nova 2 Omni is multimodal (text, image, video). Nova Sonic handles voice conversations in real time with multilingual translation.

These models are available in Amazon Bedrock and offer a competitive alternative to Anthropic’s Claude and OpenAI’s GPT, with the advantage of native integration in the AWS ecosystem.

Frontier Agents: autonomous AI in production

Frontier Agents represent the most significant announcement for operations teams. The DevOps Agent investigates incidents autonomously by analyzing logs, metrics and recent deployments. It identifies the root cause and proposes a remediation. The Security Agent continuously monitors vulnerabilities and applies fixes proactively.

These agents are not prototypes. They are designed to run in production with built-in guardrails. Humans stay in the loop to approve critical actions, but investigation and analysis are fully automated.

Kiro: the agentic IDE

Kiro is AWS’s new IDE designed for AI-assisted development. It supports spec-driven workflows where AI generates code from specifications. The Steering Files system lets you give agents context about your codebases and your conventions. Kiro CLI extends these capabilities to operations: script generation, infrastructure automation and incident response.

Amazon Q Developer: major evolutions

Amazon Q Developer, the AI assistant for developers, receives significant improvements. Code generation is more accurate and contextual. The transformation agent automatically modernizes Java and .NET applications. Integration with the AWS console lets you diagnose and resolve infrastructure problems in natural language.

What this means for businesses

re:Invent 2025 confirms that agentic AI is no longer experimental. The tools are production-ready and the use cases are concrete: reduced incident resolution time, faster development, automatic infrastructure optimization. Businesses that adopt these tools early gain a significant operational advantage.

LCMH supports businesses in adopting the latest AWS innovations.

To understand the AI-DLC approach, read our article on the AI-Driven Development Life Cycle.

Sources

1. AWS, Top Announcements of AWS re:Invent 2025. aws.amazon.com/blogs/aws/top-announcements-of-aws-reinvent-2025
2. About Amazon, AWS re:Invent 2025 News. aboutamazon.com/news/aws/aws-re-invent-2025-ai-news-updates
3. AWS, Amazon Nova. aws.amazon.com/ai/generative-ai/nova