AWS Podcast Episode #748: re:Invent 2025 – Matt Garman Keynote Recap

Release Date: December 3, 2025
Hosts: Simon Leish & Gillian Ford

Episode Overview

In this special re:Invent 2025 keynote recap, hosts Simon Leish and Gillian Ford break down the key announcements from Matt Garman’s keynote address. The episode is jam-packed with major cloud innovations, especially in AI/ML, developer experience, infrastructure, modernization, support, security, and storage. The hosts contextualize these launches for practitioners, with memorable moments, personal insights, and plenty of real-world use case relevance.

Key Topics & Discussion Highlights

1. AI & Machine Learning Innovations

• AWS Clean Rooms – Synthetic Data for Privacy-Safe ML Training

  • New capability to generate synthetic datasets retaining statistical patterns without exposing original records.
  • Simon Leish (01:08): “This feature generates synthetic training datasets that preserve the statistical patterns of the original data without the model having access to original records.”

• Amazon Nova 2 Lite

  • Fast, cost-effective reasoning model supporting 1M token context window, multi-modal input (text, images, video, docs), and built-in tools (Web Grounding, Code Interpreter).
  • Simon Leish (01:54): "You can also customize this as well... it has two built-in tools: Web Grounding and Code Interpreter."

• Amazon Nova 2 Sonic

  • Advanced speech-to-text model with real-time, expressive voice support (multi-gender, multi-language including code switching).
  • Gillian Ford (04:02): “I see customers, especially in healthcare, who want to add voice interactions...super helpful when patients are multilingual.”

• Amazon Nova Forge & Nova Act

  • Nova Forge: Build custom frontier models with supervised fine-tuning and RL, leveraging anonymized and curated datasets.
  • Nova Act (now GA): AI agents automate complex web workflows via secure, isolated deployment and no-code web UI.
  • Gillian Ford (05:34): “Having a really nice UI ... makes it a lot easier for domain experts to be able to work with those AI agentic developers.”

• Amazon Bedrock: Model Choice & Agent Core Enhancements

  • Agent Core: New quality evaluations, policy controls, and expanded memory and bi-directional streaming for trusted AI agents.
  • Bedrock Model Expansion: Addition of 18 managed open-weight models, including new Mistral models and Google Gemma.
  • Simon Leish (09:10): “There are now over 100 serverless models available in Bedrock...the same API, same way of interacting.”

• Amazon S3 Vectors (GA)

  • Store/search up to 20 trillion vectors per bucket, major performance/scalability improvements.
  • Gillian Ford (08:57): “This is just one of those... lower cost, lower latency, higher throughput—it just checks all the boxes.”

2. Developer Productivity & Serverless

• Amazon SageMaker AI with Serverless MLflow

  • Serverless experiment tracking simplifies infrastructure and scales automatically.

• New Compute Innovations

  • Memory-optimized EC2 X8AEDZ instances with 5 GHz AMD EPYC CPUs and 3 TiB RAM.
  • Gillian Ford (11:26): “Ideal for electronic design automation workloads and memory-intensive databases...”

• Lambda Managed Instances

  • Blending serverless and EC2—access specialized compute, zero infra management.

• Lambda Durable Functions

  • Native capability for durable multi-step workflows; supports up to one year suspension; no payment for idle compute.
  • Gillian Ford (14:22): “The paradigm...you can build now with serverless really starts to bend a bit...all without paying for idle compute during wait times.”

• Container Orchestration with Amazon EKS Capabilities

  • Managed Kubernetes-native solutions with integrated open-source tools: ArgoCD, AWS Controllers for Kubernetes, Kube Resource Orchestrator.

3. Database and Analytics

• Database Savings Plans

  • Commit to hourly usage, reduce costs on AWS-managed databases by 3.35%.

• Amazon RDS Upgrades

  • SQL Server Developer Edition support (free, full enterprise functionality for non-production).
  • M7I and R7I instances on RDS for SQL Server (cost/perf. optimizations).
  • Up to 256 TB storage per RDS instance.

• Amazon OpenSearch

  • GPU acceleration for vector DBs: 10x faster, 25% of cost; auto-optimization for latency/quality tradeoff.

4. Cloud Support & Operations

• Revamped Support Offerings

  • Business Support Plus:
    • AI-powered, 30 min critical response, starts at $29/month (71% cheaper than old plan).
  • Enterprise Support:
    • Designated technical account manager, incident planning, AI-assist, 15 min response, starts at $5,000/month (67% lower).
  • Unified Operations Support:
    • Top-tier, 5-min response, multi-expert team for context-aware support.

• CloudWatch Unified Data Management

  • Consistent analytics with open schema and OpenTelemetry integration; S3 Table format allows querying with Athena/SageMaker.

5. Modernization, Migration & Mainframe

• AWS Transform Custom & Mainframe
  • Define and automate code transformations across large codebases for Java, Node.js, Python.
  • Simon Leish (24:37): “Customers using AWS Transform Custom have achieved up to 80% reduction in execution time in many cases, which frees up your developers.”
  • Mainframe modernization: From monolith to microservices; adds data lineage, auto-testing, SQL Server migration, modernization of old UI frameworks.

6. Networking & Partner Ecosystem

• Amazon Route 53 Global Resolver

  • Unified, secure, Anycast DNS resolution globally; supports encrypted etc. Centralized logging, preview status.

• AWS Partner Central in Console

  • Discover and onboard AWS partners inside AWS Management Console.

7. Security, Identity & DevOps

• AWS DevOps Agent (Preview)

  • Autonomous AI agent for incident response, root cause analysis, stakeholder comms, and integrates with Slack.
  • Gillian Ford (30:55): “...a teammate that can be working 24 hours a day, doesn't need constant intervention, is massively scalable, and yes, needs no coffee.”

• Amazon GuardDuty Expanded Detection

  • Unified detections across IAM, S3, EKS, now EC2 and ECS.

• AWS Security Hub Analytics

  • Real-time and historic risk analytics, unifies prioritization and incident management; up to one year of findings.

• AWS Security Agent (Preview)

  • Proactive app security from design to deployment, automated pen-testing and reviews.

• IAM Policy Autopilot

  • AI-driven policy creation based on analyzed code SDK calls.
  • Simon Leish (35:52): “I think this is a big deal because I don’t know about you, but I’m not the best at creating IAM policies the first time around.”

8. Storage

• Amazon FSx for NetApp ONTAP --> S3 Integration

  • Unified access for AI workloads; seamless data integration.

• Amazon S3 Table: Replication and Intelligent Tiering

  • Three storage tiers (frequent, infrequent, archive), automatic transitions, no impact on compaction or snapshotting.

• S3 Storage Lens Enhancements

  • Perf metrics for billions of prefixes; exports to S3 Table for analysis.

Notable Quotes & Moments

• Gillian Ford (30:55):
  “The first one I think this is something that every single person who has been doing DevOps and AWS has wanted, which is a teammate that can be working 24 hours a day, doesn't need constant intervention, is massively scalable, and yes, needs no coffee.”
• Simon Leish (08:57):
  "The 'I don't have to do anything to get the benefit of it' is always a big tick."
• Gillian Ford (35:52):
  “Like when your application needs IAM policies, your coding assistants can call the IAM Policy autopilot to analyze the AWS SDK calls within your application and it will generate the required IAM policies. Like that is wow.”

Timestamps for Important Segments

• AWS Clean Rooms / Privacy-Enhancing ML – [01:08]
• Amazon Nova Models, Bedrock Model Choices – [01:54] to [10:44]
• Vector Database Breakthroughs – [08:57]
• Lambda Durable Functions – [14:22]
• Database & RDS Cost/Performance Features – [16:28]
• Support Plan Overhaul – [19:26] to [23:43]
• AWS Transform & Mainframe Modernization – [24:37]
• Route 53 Global Resolver – [29:17]
• AWS DevOps Agent & Security Hub – [30:52] to [35:52]
• Storage Tiering & S3 Table – [36:31] to [38:57]

Concluding Thoughts

Matt Garman's keynote ushered in a new era for AWS customers: AI and agents everywhere, massive infrastructure and developer productivity gains, aggressive support pricing redesign, and generational security improvements. As the hosts emphasize, re:Invent 2025 positions AWS as both a visionary and pragmatic provider—enabling customers to do more, safer, faster, and with less friction.

For more details or specific segment recommendations, consult the timestamps above.