AWS Podcast Episode #750: re:Invent 2025 – Day 3 Wrapup
Hosts: Simon Elisha and Gillian Ford
Date: December 4, 2025
Episode Overview
On the final day of AWS re:Invent 2025, Simon and Gillian look back at the year's biggest announcements, reflect on trends shaping the AWS ecosystem, and highlight key takeaways for developers and IT professionals. The episode features a deep dive into major AWS product launches—headlined by Graviton5 CPUs, advances in AI services, enhancements in storage, and evolving best practices. The hosts blend technical insight with personal perspective, sharing both excitement for breakthrough innovations and actionable advice for listeners.
Key Announcements and Discussion Points
1. Graviton5 – The New AWS-Designed CPU
[00:34–04:42]
- Launch: Introduction of Graviton5, AWS's most powerful and efficient CPU to date.
- Graviton-based EC2 M9 instances now offer the highest CPU core density in AWS EC2 (192 cores/package).
- 33% reduction in inter-core communication latency, supporting workloads like real-time gaming, high-performance DBs, and analytics.
- 5x larger L3 cache, yielding 2.6x more cache per core over Graviton4 for faster memory access.
“We often think about latency between regions… this is between cores on the die. It’s pretty, pretty cool.” — Simon [00:49]
-
Network & Storage Enhancements:
- Up to 15% higher network bandwidth, 20% higher EBS bandwidth, and up to double network bandwidth for largest instances.
-
Nitro System Integration:
- New Nitro Isolation Engine: offloads virtualization, storage, networking; delivers formally verified isolation for workloads.
- Integration of 6th-gen Nitro cards into Graviton5 instances.
- Formal verification provides mathematical certainty — tying back to recent discussions with Byron Cook on automated reasoning.
“It’s not just about saying things are happening, it’s about being able to prove that they're happening, which is very, very cool.” — Simon [03:23]
-
3nm Process: The chip is built with cutting-edge 3nm technology.
-
Adoption:
- Over half of new AWS CPU capacity comes from Graviton for the third consecutive year.
- 98% of the top 1000 EC2 customers use Graviton (companies include Adobe, Airbnb, Atlassian, Epic Games, SAP, and more).
“If you can use it, you should use it because it’s going to make life better for you.” — Simon [04:19]
- Instance Preview Info:
- M9G (general purpose): available in preview now.
- C9G (compute-optimized) & R9G (memory-optimized): launching next year.
2. AI & ML Platform Advancements
[04:43–08:31]
- Amazon Bedrock Updates:
- Bedrock now supports the new Responsys API (compatible with OpenAI API), enabling:
- Asynchronous inference for long-running tasks.
- Simplified tool integration for agent workflows.
- Stateful conversation management—no need to pass full conversation history per call.
- Streaming and non-streaming modes for flexible deployment.
- Seamless migration (only a base URL change for OpenAI SDK users).
- Bedrock now supports the new Responsys API (compatible with OpenAI API), enabling:
“The Responsys API enables them to automatically rebuild context without manual history management.” — Gillian [05:42]
-
Model Support:
- Immediate support for OpenAI GPT-OSS-20B/120B models, with others (like Llama v3, Mistral, etc.) coming soon.
-
Strands Agents SDK – TypeScript Preview & Extensions:
- TypeScript support (in preview) joins existing Python SDK.
- Idiomatic TypeScript experience: full type safety,
async/awaitpatterns. - Strands Agents can now run in browsers, server-side (Lambda/Bedrock), complete TypeScript stack (with CDK).
- New Features:
- Edge device support (GA): Run agents on small/local devices, incl. support for providers like Llama.cpp.
- Strand Steering (experimental): Modular prompting to guide agents toward outcomes.
- Strand Evaluations (preview): Validate and measure agent behavior during development.
“It just lets you build a lot of cool stuff with a fair bit of control, but without having to do all the boilerplate stuff.” — Simon [08:12]
3. Year-in-Review Reflections
[08:42–16:40]
A. Key Architectural and Platform Trends
- Year of IPv6:
- Industry has finally reached broad IPv6 support across all AWS services.
“We've been talking as an industry about moving from IPv4 to IPv6 for a long time… you can pretty much use it across the board now.” — Simon [09:38]
- Explosive Growth in Amazon Connect:
- Fast pace of releases and adoption for voice/AI-driven applications, especially in healthcare and customer experience.
“Voice, especially with AI, has become a trend… With Connect being really this ecosystem... I’m very excited for customers.” — Gillian [10:27]
- Capacity Increases Across Core Services:
- Major upgrades to RDS (up to 256 TB), S3 (50 TB objects), Kinesis, SNS, SQS, including throughput increases.
- No customer action required—cloud delivers improvements automatically.
“The stuff I use constantly gets better and I don’t have to do anything to make it better.” — Simon [11:28]
- Rise of MCP (Model Composition Patterns):
- New tools emerging from AWS Labs and GitHub; pace of innovation exceeding documentation.
“I think it’s a testament to just the pace of innovation...” — Gillian [12:26]
- Rapid Change in AI/GenAI Best Practices:
- AI prompting, context management, and abstractions have “completely changed” within a single year.
“I kid you not, I've rewritten this thing three times now because... something comes out that renders most of the work I did obsolete.” — Simon [13:21]
B. Practical Customer Advice
- Education & Community:
- AWS continues investing in customer education—via workshops, summits, and the podcast.
- Highlighting the “AI League” and prize competitions at re:Invent for knowledge sharing.
“AWS has historically just really heavily invested in educating customers early on… as we adapt and evolve to what these new best practices are.” — Gillian [14:07]
- Storage as Foundation:
- S3 performance, S3 Tables, and advanced tiering create a strong foundation for scale.
- Vector support in S3 highlighted as transformative for data & AI workloads.
“The vectors one was a huge one. You’re right. That’s a really good callout.” — Simon [15:56]
4. Actionable Guidance & Best Practices
[17:02–19:44]
- Security First:
- Patch management should be automatic thanks to AWS features.
- Respond quickly to vulnerabilities—cloud allows faster patching than traditional IT.
“There’s no excuse to be running things that aren’t patched properly anymore… the speed of patching is also really, really fast in the cloud.” — Simon [17:30]
- Optimize Regularly:
- Review EC2 instance types and service choices every 3–6 months for cost savings.
“If you’re not doing that at least every six months... you’re just missing out.” — Simon [18:00]
-
Update Best Practices:
- Rapid changes (e.g., Lambda’s long-running functions) mean yesterday’s architectures may not be optimal today.
- Be adaptable—continue learning and re-evaluating.
-
Visibility Tools:
- CloudWatch and Security Hub updates help customers understand environments.
- AWS MCP servers provide projections/cost optimization.
“It’s never been easier to… optimize your architecture.” — Gillian [19:38]
5. Final Reflections & Motivation
[19:44–21:57]
- Embrace Experimentation:
- AI for DevOps and Security were major themes in 2025 re:Invent launches.
- Encouragement to explore AI/ML tools, regardless of role.
“It's never been a better time to really experiment… especially with using AI for your own productivity.” — Gillian [20:06]
- Community and Podcast:
- Podcast is a passion project for both hosts, who are full-time solutions architects.
- Recognition of the importance of connection with customers and the hands-on learning it provides.
- Looking Forward:
- Teasers for upcoming episodes and encouragement to catch up on 'Interview Pro' segments.
- Big message: share the podcast and keep building.
Memorable Quotes
- “I always tell people you need to go to re:Invent at least once.” — Simon [00:24]
- “The actual formal verification to provide mathematical certainty that your workloads are isolated from each other… I want to know how they did that.” — Gillian [02:34]
- “It's just amazing stuff… customers don't have to do anything to get the benefit of it.” — Simon [11:20]
- “What I love is that… the stuff I use constantly gets better and I don’t have to do anything.” — Simon [11:28]
- “None of us can really know everything. So just being able to hear someone else's experience. I always learn something.” — Gillian [16:40]
- “Don't get overwhelmed… the fundamentals will hold you in good stead.” — Simon [17:02]
- “It’s never just been faster to be able to run all of these tests, to be able to get the data that you need.” — Gillian [19:01]
- “It's never been a better time to just learn and experiment, especially with using AI for your own productivity.” — Gillian [20:22]
Key Timestamps
| Time | Segment | |-----------|------------------------------------------------------| | 00:34 | Graviton5 Announcement, Architecture Details | | 02:21 | Nitro System & Security Formal Verification | | 03:16 | Automated Reasoning, 3nm Tech, Customer Adoption | | 04:43 | Graviton M9G/C9G/R9G Instance News | | 05:41 | Amazon Bedrock Responsys API Launch | | 06:26 | Strands Agents SDK TypeScript & New Features | | 08:42 | Podcast & AWS Year-In-Review | | 09:20 | IPv6 Year, Capacity Increases, Amazon Connect | | 12:22 | MCP Trends and Rapid Change in GenAI Practices | | 14:37 | S3 & Storage–AI Vector Support, Lambda Durable Fn's | | 16:40 | Reflections on Podcast Guests & Learning | | 17:02 | Best Practices: Security, Optimization, Flexibility | | 19:01 | Faster Decision-Making With AWS Analytics Tools | | 20:04 | Reflection & Advice on AI Experimentation | | 20:38 | Community Thanks, Podcast Insights |
Summary
This episode delivers a tech-rich wrap-up of AWS re:Invent 2025, contextualizing major launches like Graviton5 and AWS AI advancements within a broader landscape of accelerating change and innovation. Simon and Gillian blend actionable technical detail with practical advice—urging listeners to optimize, continuously re-evaluate, and not get overwhelmed by the pace of transformation. The passionate, down-to-earth tone, combined with their hands-on experience, makes this a must-listen for anyone shaping, building, or running on AWS.
Key Takeaway:
Cloud is evolving faster than ever—embrace the change, invest in learning, take advantage of new capabilities, and “keep on building.”
