re:Invent 2025: The DevOps Highlights

re:Invent 2025 wrapped up a couple of weeks ago. Following the same format as our 2024 recap, this post covers the announcements we think actually matter for the day-to-day work of running production workloads on AWS.

Note: This recap is written based on the broad themes that emerged at re:Invent 2025. We’d encourage readers to validate specific service availability, pricing, and feature details against current AWS documentation before architectural decisions.

The themes that actually mattered

Three threads ran through nearly every keynote and matter for DevOps teams:

1. Agents are now an operational discipline

The agentic AI story moved decisively from “what can it do?” to “how do you operate it?” New tooling around evaluation, cost monitoring, and auditing for AI agents got significant airtime. For teams already running Bedrock Agents in production, this is overdue.

The substantive question for your team in 2026: are your agent operations mature enough to keep up with the pace of capability improvement? If you don’t have an evaluation pipeline, this is the year to build one.

2. IaC authoring keeps getting more agentic

Q Developer’s evolution toward true agentic IaC authoring continued. Iterative author-validate-fix loops for CloudFormation, multi-file refactors, and tighter integration with CodePipeline review.

For teams with significant CloudFormation or Terraform footprints, this materially changes the day-to-day authoring experience. Worth piloting on a non-critical project early in 2026.

3. FinOps tooling is consolidating

A clear consolidation push around cost-management tooling: Cost Explorer, Compute Optimizer, Trusted Advisor, and per-service dashboards are slowly being unified into a more coherent experience, with AI-driven recommendations layered in.

For teams that have been managing cost across multiple disparate tools, this is welcome. Don’t migrate workflows immediately (let the new tools mature for a quarter), but plan to consolidate in 2026.

Service-level announcements worth your attention

Without claiming completeness, the announcement areas to follow up on:

• Aurora DSQL continued maturation toward production-ready multi-region SQL
• S3 Tables ecosystem broadening: more engines, more management features
• Bedrock Agents governance and evaluation tooling
• Q Developer capability expansion into deeper IaC and pipeline workflows
• Cost-management tooling consolidation under a unified experience
• Multi-region resilience patterns: supporting services and reference architectures

For each of these, the right next step is the same: read AWS’s published documentation, evaluate against your current architecture, and pilot before committing.

What we’re skipping

As in past years, plenty of announcements got airtime without changing what most teams do day to day:

• New model variants without substantial capability gains
• Console UI improvements
• Niche service announcements relevant only at very large scale

Useful for someone, not load-bearing for most.

What we’re prioritizing in early 2026

Internally, our priorities going into the new year:

1. Building out evaluation pipelines for client Bedrock Agent deployments
2. Piloting expanded Q Developer agentic capabilities on internal IaC work
3. Beginning a planned migration of one client workload to Aurora DSQL now that the production-readiness story is clearer
4. Doing FinOps reviews using the consolidated tooling for several clients

Closing

re:Invent 2025 was less about new categories than about maturing the categories announced in the previous two years. That’s a healthy signal: the platform is consolidating around the right primitives. The engineering work shifts from “what’s new?” to “how do we operate this well?”

If you’d like to talk through what any of this means for your AWS environment heading into 2026, we’re here.