Key Insights from AWS Summit 2025: What Stuck with Us (and Why)

By Samuel Fanoukoe, Alexander Pevzner

We joined thousands of cloud and AI professionals at AWS Summit Amsterdam for a full day on the future of cloud, data, and GenAI. From packed sessions to live demos, the event brought together experts, builders, and decision-makers to explore the latest AWS tools and real-world applications. While GenAI took the spotlight, we also came away with insights on architecture, hybrid deployments, and cost optimization. Here’s what stuck with us, and the questions we think every data-driven team should be asking now.

The Big Themes That Defined the Summit

GenAI Everywhere: As expected, a large part of the summit focused on GenAI. Out of 98 sessions, more than 30 explored tools like Amazon Bedrock, SageMaker, Amazon Q, RAGs, and knowledge bases. An interactive quiz booth in the summit hall confirmed just how central GenAI has become, with over 80% of the questions touching on its application across the AWS ecosystem.

📸 Paul putting his AWS knowledge to the test

Modern Data Architectures: Sessions covered architectural patterns like serverless, decoupling, choreography vs. orchestration, and data vs. control flow. These presentations sparked quite some discussion as you can hardly find a single right decision for these topics.

Cost Optimization in Focus: AWS shared a lot of practical advice on reducing costs across services. We've seen this firsthand too, AWS support has helped us solve tricky cost-related challenges outside the summit.

Kubernetes at Scale: No cloud event is complete without Kubernetes. Sessions tackled managing cluster fleets, building high-performance EKS environments, and reducing overhead.

Something for Every Role: The summit delivered value for every type of attendee:

IT executives got an overview of current trends, challenges, and opportunities

Solution architects found inspiration in modern cloud patterns, best practices, and cost techniques

Cloud developers and data engineers gained practical insights into the most efficient tools, and their limitations


What Stuck With Us: Sessions, Demos & Lessons

The summit gave us a closer look at AWS’s current development directions, especially when it comes to GenAI, system design patterns, and hybrid infrastructure. Here are a few sessions and moments that really stood out to us:

Seeing RAGs in Action

One memorable session was with Sachin Kulkarni, Senior Solution Architect at AWS, who walked us through how to build RAG workflows using Amazon Bedrock. He emphasized the importance of preventing hallucinations in LLMs. For example, he showed how an LLM might return completely different results to the same request, and how connecting a knowledge base to Bedrock can help resolve this.

Getting Our Hands Dirty: Unicorn GameDay🦄

We also participated in the AWS Generative AI Unicorn Party GameDay, a hands-on team challenge using Bedrock to build AI applications. We completed two quests:

  • Voice of the Unicorn: We used Amazon Transcribe to convert support calls to text, then applied Bedrock’s models to analyze complaints and resolution times.
  • Mystic Code: We developed a Streamlit app that generates unicorn images, containerized it with Docker, and deployed it to ECS.

This exercise showed how AWS services can be integrated to solve real business problems, end to end.

Smarter Cloud Patterns

Two sessions stood out on system design and best practices. In “Best Practices for Serverless Developers,” Gunnar Grosch and Yan Cui shared practical tips for AWS Lambda and Step Functions:

  • When to use sync vs. async execution
  • Warming up Lambda functions to reduce cold starts
  • Optimizing code for multi-core environments
  • Using Step Functions to manage complex workflows

Another great session, “Integration Patterns for Distributed Systems” by Michael Gasch and Maximilian Schellhorn, explored how to decouple systems using queues, pub/sub, and orchestration. They gave a super nice reasoning for the pros and cons of choreography vs. orchestration, and how both apply in services like SQS, SNS, and EventBridge.

🧩 Why Hybrid Still Matters

AWS continues to invest heavily in hybrid solutions. Being a customer no longer means you're limited to the cloud. Tools like Amazon EKS Anywhere and AWS for VMware support on-prem workloads, especially valuable in compliance-heavy environments.

One standout session was “A Platform Engineering Journey of Large-Scale Kubernetes Migration” by Marco Cambon, Sebastian Wermann (Webfleet), and Amir Asfandyarov (AWS). They shared how they migrated a multi-tenant Kubernetes fleet from Azure to AWS, solving cost spikes by switching NAT mode without downtime. A deployment model we really liked: clusters pulling their own config instead of relying on push.

Also unforgettable was “IoT at Scale, Security, and Governance” by Sam Kool (AWS) and Marchin Dzwonkowicz (Royal Philips), who showed how Philips built a secure, compliant IoT platform to manage over a million connected healthcare devices. A strong example of how AWS goes far beyond the old “cloud-only” model, enabling developers to deliver on ambitious goals at scale.


The challenges discussed at the Summit are ones many teams are already facing. Understanding current trends is essential to anticipate and address what’s next.

The GenAI trend opens a lot of opportunities. Almost every operation that was previously manual may now be reconsidered using GenAI. But as it becomes more integrated, it also raises new questions for data strategists:

  • Are we ready to provide reliable data sources with proper integrations for building a knowledge base?
  • We're currently comfortable sharing small code snippets with tools like ChatGPT, but will we feel the same about sharing an entire codebase with a GenAI system?
  • How much can we trust the output of a GenAI tool? What kind of results are we prepared to accept, can we build a banking solution on GenAI?

The shift toward on-premises and hybrid solutions also brings new options for making systems more secure and distributed. But it introduces fresh challenges for cloud engineers, who now need to manage increasingly complex Kubernetes-based and bare-metal architectures.


Where This All Points Next

AWS Summit 2025 confirmed what we’re already seeing across our projects: GenAI is no longer just a feature, it's becoming a foundational layer. And cloud-native thinking is expanding to include hybrid environments, stricter governance, and smarter cost control.

We're already applying some of these learnings in our projects, and we’re always happy to share ideas, learnings, and approaches with others navigating the same challenges.

Want to explore how these trends could apply to your business? Let’s connect.