re:Invent 2025: The Age of the AI Agent Is Here

Versent was on the ground at re:Invent 2025, right in the middle of the energy, the announcements and the hallway debates that shaped the week. And this year truly felt like a turning point. Every keynote, every side conversation and every breakout session pointed to a cloud landscape shifting faster than ever before.

But the biggest surprise of all was not an AI reveal. In a year overflowing with autonomous agents, multimodal models and frontier-scale infrastructure, the loudest applause of the entire conference exploded for something far more humble but deeply practical: Database Savings Plans. A cost optimisation update delivered the most emotional reaction at re:Invent 2025. That irony set the tone for the week and reminded everyone that while AI is the future, the foundations still matter.

With that grounding moment, AWS unveiled its clearest vision yet of where the cloud is heading. We are no longer simply scaling environments. We are entering the Age of the AI Agent, where systems can work beside us, for us and sometimes entirely without us. 

Below is a breakdown of the major announcements that are reshaping the cloud landscape and what they mean for customers and partners.

The Rise of Agentic AI: Cloud that Works Ahead of You

Agentic AI stole the spotlight. These are not simple assistants. They are autonomous workers that can operate for hours or days, break down complex problems, sequence tasks and adapt to context. They represent a historic shift in how value will be created on the cloud.

For many attendees, this was the first time AI felt not just powerful, but independent.

Key AI Agent and Foundation Model Announcements

Amazon Nova 2 Model Family

AWS expanded the Nova model line with serious upgrades in performance and capability.

• Nova 2 Lite is the new cost-effective reasoning workhorse.
• Nova 2 Omni brings fully multimodal capabilities across text, image, video and speech.

 Amazon Nova Forge

A bold move towards openness, Nova Forge lets organisations participate in the training process by combining their own high-value data with curated Nova datasets. For those waiting for more transparent and adaptable foundation models, this was a standout moment.

Frontier Agents: A New Class of Autonomous Workers

AWS introduced three agents that feel like the beginning of a new cloud era.

• Kiro Autonomous Agent, the virtual developer that understands context and tackles complex engineering tasks.
• AWS Security Agent, a relentless reviewer that never tires and never misses.
• AWS DevOps Agent, a real-time problem-solver that identifies issues before they escalate.

AgentCore: The Governance Brain

AWS reinforced that autonomy cannot exist without safety and control.

• Policy in AgentCore gives teams real-time deterministic guardrails.
• AgentCore Evaluations ensures agents stay aligned, predictable and trustworthy.

 This is AWS signalling that the AI agent revolution will be enterprise-ready from day one.

New Skills, New Specialists, New Expectations

Agentic AI Competency

A new competency that recognises partners who can actually design, build and operate real agents in real environments.

Generative AI and Serverless Microcredentials

• GenAI Essentials focuses on safe, smart adoption.
• Serverless Modernisation aligns future cloud applications to event-driven, AI-ready patterns.

 The message was clear. Skills will define the winners in the Age of Autonomy.

Next-Generation Cloud Infrastructure: Built for an AI World

AWS Graviton5

Faster, more efficient and designed for the hybrid mix of AI and traditional workloads that define modern applications.

Trainium3 UltraServers

A monster training platform with 4.4 times more compute capacity and four times greater energy efficiency than Trainium2. AWS is running at the frontier scale now.

Amazon S3 Vectors

Native vector search built straight into S3, instantly simplifying generative AI architectures and removing the need for yet another specialised database.

The Announcement That Shook the Room: Database Savings Plans

In a week obsessed with AI, nothing landed quite like this.

When AWS announced Database Savings Plans, the entire room erupted. Cheers, applause and even laughter at the sheer relief of it.

This update hit home because customers have been waiting years for exactly this.

Why the excitement?

• Up to 35 per cent savings on Aurora, RDS and DynamoDB.
• Automatic application across engines and configurations, unlike Reserved Instances that forced you into rigid commitments.
• A simple spend-based model that reflects how modern architectures actually work.
• Perfect alignment with autoscaling, serverless and variable workloads.

The irony was not lost on anyone. At the most AI-heavy re:Invent in history, the loudest applause celebrated a cost optimisation feature.

Cloud Operations and Security for an Autonomous Era

• AI-Native CloudWatch now gives visibility into how agents and models behave.
• GuardDuty Extended Threat Detection unifies analysis across EC2 and ECS, matching how attackers actually move.

AWS is quietly turning observability and security into AI-first disciplines.

AI Factories: Turning AI into a Production Line

One of the strongest themes across customer talks was the rise of AI Factories. These are not physical spaces, but digital supply chains that:

• prepare data
• train models
• evaluate behaviour
• deploy agents
• govern outcomes

All using integrated pipelines, S3 Vectors, Trainium3 compute and AgentCore governance.

For many organisations, AI Factories felt like the missing piece. The structure behind the ambition. A way to stop experimenting and start industrialising.

Interoperability and Ecosystem Moves

AWS Interconnect with Google Cloud

A surprising but welcome announcement. Private, high-bandwidth, automated cloud-to-cloud connectivity. Multicloud is not a theory anymore. It is becoming a default.

NVIDIA Partnership

Support for NVLink Fusion and integration of Nemotron open models into Bedrock reinforces that the AI ecosystem is not just expanding, it is converging.

The Cloud Has Become Autonomous

re:Invent 2025 marked the moment the cloud stepped into its next era. AI agents, AI factories, multimodal models and intelligent infrastructure are now part of the core cloud experience.

But perhaps the real lesson was this. Even in a year where autonomy stole the spotlight, it was a simple cost optimisation announcement that reminded everyone to stay grounded.

The future is autonomous. The foundations still matter.

How Versent Helps You Become Future Ready

Stepping into the Age of Autonomy requires more than enthusiasm. It requires the right foundation.

Versent was on the ground in Vegas with a 12-strong team of engineers, architects, product leaders and executives, engaging in deep conversations about how organisations can lead in the Agentic AI cloud era.

We help organisations build AI-ready cloud platforms and AI solutions designed for performance, sovereignty, compliance and cost efficiency. We bring together the latest AWS innovations including Nova 2 models, Frontier Agents, AgentCore governance, AI Factories, Graviton5 compute, Trainium3 UltraServers and S3 Vectors.

Our architecture approach is secure by design. Our cost engineering ensures predictability. Our experience in digital sovereignty ensures your data stays protected. And our modernisation expertise means your organisation can adopt generative and agentic AI safely and confidently from day one.

We help you move from experimentation to execution and from cloud-enabled to truly autonomous. And if the thunderous applause was anything to go by, we will even help you master Database Savings Plans, because apparently the fastest way to win hearts at an AI conference is to discount Aurora.

Ready to get started? Reach out to us today!