The 2025 edition of re:Invent recently took place in Las Vegas. As anticipated, AI was a significant focus of the keynotes, but the community was more intrigued by announcements in the serverless space, including Lambda Managed Instances and Lambda Durable Functions. The conference marked the final keynote for Amazon CTO Werner Vogels after 14 years.
The Event
During his keynote, dominated by AI-related topics, Matt Garman, the AWS CEO, argued that we need to move past the conversation about whether AI is possible and enter the value-generation phase. The feedback from attendees and the community has been mixed, with many developers quipping, "It's clearly Day 2." Discussing their "dream announcement" on Reddit, a user wrote:
Dream would be that AWS lets up on the "hey guys we're not behind on AI" panic it's in and focuses on the core infrastructure that it's good at. Judging from the keynote so far it looks like that's wishful thinking.
Luc van Donkersgoed, principal engineer at PostNL and creator of the "AWS News feed," adds:
Non-AI topics get 10% of the keynote, but 70% of the attention on the AWS News Feed. Just sayin'.
As every year, while new services might have the spotlight, many significant and incremental announcements to core services that matter most to practitioners were made in the two weeks before the conference during the so-called Pre:Invent season, including regional availability for NAT gateway and a new life for CodeCommit.
Notably, Werner Vogels' keynote this year closed the conference, was significantly shorter than in past years, and was his final one at re:Invent. Vogels joked before the conference:
This year at AWS re:Invent I have the honour of a proper closing keynote: 3:30 PM on Thursday. Marketing calls it a "special keynote" but I would just call it the "last thing between you, me and the party".
During his one-hour session, Amazon's CTO discussed the dawn of the "renaissance developer":
As generative AI reshapes how we build software, a familiar trope has re-emerged, the narrative that developers will become obsolete. But if history has taught us anything, this is not the end of the developer, it's the dawn of something new, the renaissance developer.
The decision to leave the stage after 14 years, canceling the highlight of the conference for most practitioners, was the main discussion point after the event. Thiago de Faria comments:
A bittersweet moment to watch Werner Vogels keynote (...) Werner's keynote has been the highlight of AWS re:Invent every single year. Not only because of the amazing t-shirts, but because of the story telling and wisdom!
Below is a summary of this year's main announcements impacting computing, databases, storage, networking, security, machine learning, and development. InfoQ has covered or will cover the most significant ones as separate news pieces.
Compute
AWS introduced its new Graviton5 processors with the preview of m8g instances and Trainium3 UltraServers, which offer faster computing and lower costs for general workloads and AI training. It also launched new EC2 instances using 5th Gen AMD EPYC processors that support large memory requirements, such as chip design and database workloads.
For serverless deployments, AWS added Lambda Managed Instances and Lambda Durable Functions, giving developers ways to run Lambda functions on EC2 hardware and to build multi-step workflows that can pause and resume over long periods. Yan Cui, serverless expert and AWS Serverless Hero, comments:
Like London buses, we've waited years for true innovations to the Lambda platform and two came at the same time!
Steef-Jan Wiggers has covered both Durable Functions and Lambda Managed Instances on InfoQ.
Lambda durable functions. Source: AWS documentation
Amazon EKS now includes new capabilities that simplify Kubernetes operations by handling orchestration and cloud resource management.
Databases
In a long-requested update from the community, AWS is introducing Database Savings Plans. This new pricing model keeps database costs predictable while allowing flexibility across different services and deployment options. Corey Quinn, chief cloud economist at The Duckbill Group, writes:
This is better than it has any right to be, because it's not just for RDS! It supports DynamoDB, Keyspaces, DocumentDB ("Amazon Basics MongoDB"), ElastiCache(I TOLD you it was a database!! Note, this is for Valkey only; no Redis or Memcached), Aurora which both is and is not RDS depending upon how you squint at it, Neptune, Timestream (instance-based InfluxDB only), and for some godforsaken reason Database Migration Service.
More details are available in a separate news piece on InfoQ. The cloud provider also added new features to Amazon RDS for SQL Server and Oracle, and Amazon OpenSearch Service now offers GPU acceleration and auto-optimization.
Still, updates in the database space remained relatively limited and less significant compared to past years. Zongzhi Chen, manager of Alibaba Cloud RDS Team, writes:
AWS ReInvent only gave databases 1 minute and 31 seconds... Is the database industry really becoming a sunset industry now?
Franck Pachot, Developer Advocate at MongoDB and AWS Data Hero, acknowledges, "They cannot bring a new Aurora database every year ." Others argue instead that "when your most interesting announcement is database saving plans it was not a good year".
Storage
Amazon S3 Vectors is now generally available, offering faster performance, support for up to 2 billion vectors per index, 100ms query latencies, and broader regional availability.
FSx for NetApp ONTAP now integrates with S3, allowing users to access file system data directly through S3 and connect with analytics, ML, and generative AI services without moving data. Amazon S3 Tables introduces replication support and Intelligent-Tiering, enabling automatic cost optimization and table replication across regions and accounts. Additionally, S3 Storage Lens adds performance metrics, support for billions of prefixes, and integration with S3 Tables.
After many years, Amazon S3 is increasing the maximum object size from 5 TB to 50 TB.
Networking, Security, and Monitoring
Route 53 Global Resolver is now in preview and offers secure anycast DNS resolution, simplifying hybrid DNS management across public and private domains. Additionally, in a surprising move, AWS and Google Cloud have partnered to simplify multicloud networking, making it easier for organizations to manage and secure workloads across both clouds. Azure is expected to join in the near future. Quinn warns:
This is either transformative or a waste of everyone's time, and it's impossible to tell which because the one thing that matters most to settling that question is "what's the price." They aren't disclosing it yet, so at the moment it occupies a superposition of "excellent/crap."
Also in preview, the new AWS Security Agent provides AI-powered application security, including design reviews, code analysis, and contextual penetration testing.
GuardDuty includes Extended Threat Detection for EC2 and ECS, providing unified visibility across virtual machines and containers, and Security Hub is generally available with near-real-time analytics and risk prioritization. IAM Policy Autopilot, a new open-source tool, is designed to simplify IAM policy creation by analyzing code and generating valid permissions for AI coding assistants.
In preview, AWS DevOps Agent targets incident response and improves system reliability by analyzing data from CloudWatch, GitHub, ServiceNow, and other tools to identify root causes and coordinate incident response.
Source: AWS blog
CloudWatch now provides unified data management and analytics for operations, security, and compliance. The new feature is designed to reduce complexity and costs through automatic normalization, native analytics, and support for standards such as OCSF and Apache Iceberg. Simplified enablement of AWS CloudTrail events in CloudWatch was also announced.
Finally, the cloud provider announced new AWS Support plans and deprecated some existing ones. Matheus Guimaraes, senior developer advocate at AWS, explains:
Support plan capabilities will continuously evolve to add comprehensive visibility into your infrastructure (...) This combination of AI-powered tools and AWS expertise represents a fundamental shift from reactive to proactive operations, helping you prevent issues before they impact your business.
On Reddit, many developers are skeptical about the planned change from Developer to the new Business+ plan; others are concerned about the billing impact or the quality of the LLM models. A user comments:
On the AI side non-AWS models and agents are better at AWS than AWS is. Rather ironically, AWS support is the last place I want to be pawned off to an inferior chatbot.
Development
AWS expanded its "migration and modernization" solutions with AI-powered capabilities across AWS Transform, including a customizable service that automates code upgrades across entire repositories, full-stack Windows modernization tools that update applications and their dependencies, and mainframe transformation features that convert legacy systems into cloud-native architectures, automating significant portions of testing.
During the conference, the cloud provider introduced Kiro powers and Kiro autonomous agent to extend its specification-driven agentic IDE. Allen Helton, ecosystem engineer at Momento and AWS Hero, writes in his newsletter:
You activate a power and it's given MCP servers and steering docs to get it to perform a specific way. Love this idea and am fully on board with using it for targeted use cases (...) Super cool stuff coming from this team!
Machine Learning and Artificial Intelligence
The principal announcement was Amazon Nova 2, a new family of AI models for speech, reasoning, and multimodal tasks, including Nova 2 Sonic for speech-to-speech, Nova 2 Lite for fast reasoning, Nova 2 Omni for multimodal inputs, and Nova Forge for building custom frontier models.
The keynote also highlighted supporting services like Nova Act for UI automation, Bedrock enhancements for trusted AI deployment, and SageMaker updates for faster, elastic model training and serverless customization.
Furthermore, Amazon Bedrock added 18 managed open weight models, including the new Mistral Large 3 and Ministral 3 models. Stephen O'Grady, principal analyst and cofounder of RedMonk, notes:
AWS' use of open weights rather than open source to describe the models it had added support for was both interesting and notable.
InfoQ will provide a deeper review of these generative AI announcements in the AI, ML & Data Engineering queue in the coming weeks.
Resources
The AWS editorial team summarized their top announcements in an article. An unofficial, curated re:Invent feed is available at aws-news.com.
All the keynotes and many sessions are already freely available on the dedicated YouTube channels.