AWS re:Invent 2026 runs November 30 to December 4 in Las Vegas, according to AWS’s event page. The keynote schedule and speaker list had not been published at the time of writing; AWS says they are “coming soon.” Full-price registration is listed at $2,499, with early-bird pricing having ended on August 25 and a 10% discount for groups buying 10 or more tickets in one transaction.

For platform and FinOps teams, re:Invent matters less as a conference than as a pricing and roadmap event. Launches announced in the first week of December routinely land after annual budgets are drafted and right before change freezes, which means the useful work is preparing now so you can evaluate announcements quickly rather than rewriting your 2027 plan in January.

Nothing in this preview predicts specific 2026 launches. AWS does not pre-announce them, and speculation is not a planning input. Instead, it looks at what last year’s event changed and the categories worth watching.

What re:Invent 2025 Changed for Platform Teams

AWS’s own recap of the 2025 event is a useful baseline, because several of those launches are now mature enough that a re:Invent 2026 follow-up is likely to extend them.

Amazon EKS Capabilities

AWS announced Amazon EKS Capabilities, a set of fully managed cluster features. According to the EKS documentation, they cover declarative continuous deployment with Argo CD, AWS resource management with AWS Controllers for Kubernetes (ACK), and Kubernetes resource composition with kro. AWS runs these controllers rather than you running them in your cluster.

For a platform team, this changes the build-versus-buy math on a GitOps workflow. A managed Argo CD removes upgrade and availability work from your backlog, at the cost of AWS deciding versions and integration points. If your platform runs on EKS, watch for how these capabilities expand, and re-run that comparison after the event.

Lambda Managed Instances

AWS Lambda Managed Instances lets Lambda functions run on EC2 compute so you can use EC2 pricing models and specialized hardware while AWS manages the infrastructure. It blurs the line between “serverless” and “instances,” which matters for FinOps: workloads that were too expensive on standard Lambda pricing may now fit, and commitment discounts behave differently than on pure per-invocation billing.

Graviton5 and Trainium3

AWS introduced Graviton5, its fifth-generation Arm CPU, and made EC2 Trn3 UltraServers available, built on its first 3nm AI chip. Per-generation performance claims come from AWS and should be validated with your own benchmarks before you plan a migration around them. Our EC2 instance types guide covers how to run that comparison.

What FinOps Teams Should Watch

Commitment models

The most consequential 2025 launch for cost teams was arguably Database Savings Plans, which AWS describes as a pricing model that stays flexible across database services and deployment options. Before it, discounting database spend usually meant per-engine Reserved Instances. Third-party analyses disagree on details such as covered services and term options, so read AWS’s Savings Plans documentation directly rather than relying on summaries.

That launch is a template for what to watch this year. Any new commitment model, or an extension of an existing one to new services, changes how much of your bill is discountable. The FinOps preparation is simple:

  • Know your current commitment coverage and expiry dates by service, before December. A new plan type is only useful if you have room to buy it.
  • Avoid locking in large new commitments in November. A purchase made the week before a pricing announcement is the most common avoidable regret in cloud cost management.

Instance generations and price-performance

New instance families shift the cost floor. Teams already on Graviton should track whether new generations offer better price-performance at the same list price; teams still on x86 should treat each new Arm generation as a prompt to re-test, not an automatic migration. Our cloud cost optimization strategies and spot instance guide cover the broader levers.

AI and GPU cost

Inference spend is the fastest-growing line on many cloud bills. Watch for anything that changes GPU or accelerator availability, pricing or commitment options, and for managed services that bundle inference with orchestration. Pair announcements with the unit-cost approach in our FinOps for AI inference pipelines piece: a new chip is only cheaper if cost per request falls.

Data transfer

Egress pricing rarely gets keynote time, which is exactly why it is worth checking the pricing pages after every major event. See our egress audit guide for a repeatable process.

How to Process re:Invent Without Losing December

re:Invent produces more announcements than any team can evaluate. A filter keeps it manageable:

  1. Before the event, list the AWS services that make up most of your bill and the platform components you plan to change in the next two quarters. Only announcements touching that list get a follow-up.
  2. During the week, have one person per area (compute, containers, data, cost tooling) skim the launch posts daily and log anything relevant with a link and a one-line “why it matters to us.”
  3. After the event, separate generally available launches from previews. Previews do not belong in a 2027 budget. Many re:Invent features stay in preview for months, and some change significantly before GA.
  4. Re-check pricing pages, not keynote slides. Prices and regional availability are defined on the service pricing page, which is the only source you should put in a business case.

Remember that AWS is the source for its own product claims, not an independent judge of them. Treat performance and savings figures in launch posts as hypotheses to test against your workloads.

If you do one thing before November 30, export your current commitment inventory with expiry dates and coverage by service. That single spreadsheet turns every pricing announcement that week into a quick yes-or-no question instead of a research project.

Sources: AWS re:Invent event page (checked October 2026); Amazon EKS Capabilities documentation; AWS’s “Top announcements of AWS re:Invent 2025” recap.