3
3Day 1–2: AI Takes Center Stage
The opening keynotes set a clear direction: AI agents, not just assistants , are the next major shift in cloud computing. AWS CEO Matt Garman framed this shift as moving beyond chat interfaces to autonomous systems that execute workflows independently — a theme that resonated throughout the week.
Top Announcements
- Amazon Nova 2 family: New models like Nova 2 Sonic for speech-to-speech experiences, Nova 2 Lite for cost-effective reasoning, and Nova 2 Omni for multimodal AI.
- Nova Forge: A major leap in enterprise AI — build custom foundation models using AWS checkpoints and your own data.
- Nova Act: AI agents that automate browser workflows with high reliability.
- Bedrock AgentCore: Production-ready controls, memory, and quality evaluations for enterprise agentization.
- S3 Vectors GA: Scalable vector search hitting billions of vectors with fast latencies and up to 90% lower cost vs specialized DBs.
- Multicloud preview with AWS Interconnect: Secure private links to other clouds such as GCP.
These innovations weren’t surface-level demos — they were practical tools executives and engineers can apply immediately to build, automate, and scale.
Infrastructure & Compute Evolutions
AWS also doubled down on performance and cost-efficiency:
- Graviton5: AWS’s most powerful CPU yet, offering significantly better compute for databases, analytics, and high-performance workloads.
- Trainium3 UltraServers: Powerful new AI silicon enabling faster training and inference at lower unit cost.
- Lambda Durable Functions & Managed Instances: Let cloud teams build long-running workflows without paying for idle time — a big win for serverless efficiency.
Behind the scenes, talks emphasized that performance and cost control still matter as much as AI innovation. Throughout re:Invent there were deep dives into cloud governance, observability, and optimization best practices.
Cloud Cost Conversations — Latte on Cloud Costs Live at re:Invent
One of Cloudkeeper’s highlights was taking “Latte on Cloud Costs” live from the re:Invent floor. These on-camera discussions with Rick Orcs, and Lee Wagner brought practical strategies from Cloudkeeper’s team to a broader audience — and resonated with builders struggling with runaway bills.
We covered:
- Cost visibility across AI workloads
- Rightsizing compute with Graviton & Trainium silicon
- Finance + DevOps alignment for predictable billing
The coffee-cup chats turned out to be a perfect setting for honest, tactical conversations that developers and finance partners both loved.
Sessions, Community & the Overall Vibe
Outside keynotes and product news, the energy was unmistakable:
- Crowded breakout rooms on AI engineering best practices
- Multi-cloud resilience talks amid broader enterprise concerns
- Security sessions on new IAM innovations like IAM Outbound Identity Federation and Policy Autopilot tools that help reduce risk while scaling development velocity.
And of course, re:Play and hallway conversations reminded us why re:Invent isn’t just about tech — it’s about people, community, and the shared commitment to innovate.
Final Thoughts
AWS re:Invent 2025 wasn’t about the next shiny thing. It was about making innovation practical and operationalizable. Whether your team is building AI applications, optimizing cloud spend, or tightening security at scale, there was something here to elevate your strategy.
Cloudkeeper returned with learnings, relationships, and content that will fuel insights for months to come — starting with the episodes we recorded live for you.
Stay tuned for the upcoming episodes of the special edition of Latte on Cloud Costs, recorded at re:Invent 2025!
