I've been tracking AWS releases all year, and honestly, 2025 felt different. Not just "here's another managed service" different, but fundamentally "we're changing how you build software" different. Between re:Invent and the steady stream of updates throughout the year, there's a lot to unpack.
This article is my attempt to summarise the key announcements across every major category. I won't cover everything - AWS announced hundreds of updates - but I'll hit the ones that actually matter for most developers and architects.
1. The Big Picture: What Defined 2025
If I had to describe 2025 in three themes, it would be:
- Agentic AI everywhere - AWS went all-in on autonomous agents that do things on your behalf
- Custom silicon at scale - Graviton5, Trainium3, and the infrastructure to run them
- Multicloud is real now - That AWS-Google partnership wasn't on my bingo card
The common thread? AWS is positioning itself not just as infrastructure, but as the platform where AI agents live and operate. Whether you buy into that vision or not, it's clearly where they're headed.
2. Global Infrastructure Expansion
AWS continued expanding its global footprint in 2025:
New Regions Launched:
- Mexico (Central) - AWS's first region in Mexico, launched January 2025. They've committed $5 billion over 15 years, so they're serious about Latin America.
- Thailand (Asia Pacific) - Also launched January 2025 with three availability zones.
Also Launched in 2025:
- Taiwan (Asia Pacific - Taipei) - Launched June 2025 with three availability zones and $5 billion investment
- New Zealand (Asia Pacific) - Launched September 2, 2025 (ap-southeast-6) with three availability zones and NZ$7.5 billion investment
Coming Soon:
- Saudi Arabia (expected 2026)
- AWS European Sovereign Cloud - Launching December 2025 in Brandenburg, Germany with €7.8 billion investment
As of December 2025, AWS operates 120 availability zones across 38 geographic regions. Remember - you don't pay for an AWS account, just the resources running in it. But regional availability still matters for latency and data residency requirements.
3. Compute: Where Things Got Interesting
This was a big year for compute. Not just incremental improvements, but some genuinely useful new capabilities.
Graviton5 Processors
AWS introduced Graviton5, their most powerful custom chip yet. The new EC2 M9g instances deliver up to 25% higher performance than the previous generation while using less energy. If you're not already on Graviton for compatible workloads, the price-performance gap just got wider.
Trainium3 UltraServers
For AI training workloads, Trainium3 UltraServers can host up to 144 Trainium3 chips per server, providing up to 362 MXFP8 PFLOPs of compute. The chips themselves are claimed to be 40% more energy efficient than the previous generation. I haven't run benchmarks myself, but the specs are impressive.
Lambda Durable Functions
This one's probably the most practical announcement for everyday developers. Lambda Durable Functions lets you build applications that coordinate multiple steps over extended periods - from seconds to up to one year - without paying for idle compute time.
Think workflows that wait for human approval, or processes that need to poll external systems periodically. Before this, you'd need Step Functions or some custom orchestration. Now Lambda handles it natively.
Lambda Managed Instances
A new capability that lets you run Lambda-like functions on EC2 hardware. It's an interesting middle ground - serverless simplicity with EC2's power and flexibility. Useful when you need more control over the underlying compute but don't want to give up the Lambda programming model.
New EC2 Instance Types
Memory-optimised instances powered by 5th Gen AMD EPYC processors hit the market, offering up to 5 GHz speeds and 3 TiB of RAM. These are aimed at heavy workloads like databases and EDA tools.
4. AI and Machine Learning: The Agentic Era
This section could be its own article - actually, I'm writing one - but here are the highlights:
Amazon Nova 2 Models
AWS expanded the Nova family significantly:
- Nova 2 Lite - Fast, cost-effective reasoning for everyday workloads
- Nova 2 Pro - The most capable model for complex, multi-step tasks (currently in preview)
- Nova 2 Sonic - Speech-to-speech model supporting seven languages
- Nova 2 Omni - The industry's first reasoning model that processes text, images, video, and speech while generating both text and images
All Nova 2 models support extended thinking with adjustable intensity levels. You can dial up the reasoning depth when you need it and keep it light for simpler queries.
Amazon Nova Forge
This one's fascinating - it's essentially "build your own frontier model." You start from Nova model checkpoints, blend your proprietary data with Nova's training data, and get a custom model that combines Nova's capabilities with your domain knowledge. Reddit apparently already built their own model using it.
Nova Act
Browser automation agents. Powered by a custom Nova 2 Lite model, it delivers 90% reliability on browser-based tasks. If you've been building web scrapers or automation tools manually, this might be worth exploring.
Amazon Bedrock Updates
Bedrock now has over 100 foundation models, including 18 new open-weight models added in December. The big capability addition was reinforcement fine-tuning, which uses feedback-driven training to deliver 66% accuracy gains without needing massive labelled datasets.
Amazon Bedrock AgentCore
AgentCore went from preview (July) to GA (October) to feature-rich (December). It's the infrastructure layer for building, deploying, and operating AI agents at scale. The SDK has been downloaded over 2 million times in just 5 months.
Key components:
- Runtime - Session isolation, bidirectional streaming for voice agents
- Memory - Now includes episodic memory for agents that learn from experience
- Gateway - Convert existing APIs to MCP-compatible tools
- Identity - OAuth integration and secure token storage
- Observability - CloudWatch dashboards for agent monitoring
- Policy - Real-time tool call interception using Cedar policies
- Evaluations - 13 built-in evaluators for quality monitoring
Amazon Q Developer
The AI coding assistant got significant updates:
- C#, C++, and 11 additional languages for customisation
- GitLab Duo integration (GA)
- GitHub integration (preview - no AWS account required)
- MCP support in the CLI
- Conversation history that persists between sessions
- Pro Tier available in Frankfurt for EU data residency
Kiro IDE
AWS released Kiro, an agentic AI IDE that Amazon now uses internally. Built on VS Code, it features spec-driven development where you write requirements in markdown and the agent scaffolds everything. One internal project reportedly went from 30 developers over 18 months to 6 developers over 76 days. I'm still evaluating it myself, but the early signs are promising.
5. Database: Savings and Scale
Database Savings Plans
Finally - a single, flexible commitment that applies across RDS, Aurora, DynamoDB, ElastiCache, Neptune, and DocumentDB. No more juggling separate Reserved Instance portfolios per engine. You can reduce costs by up to 35% with a one-year commitment.
Aurora DSQL
Cluster creation now takes seconds instead of minutes. Useful for rapid prototyping and testing scenarios.
RDS Storage Expansion
RDS for SQL Server and Oracle now support up to 256 TiB of storage (up from 64 TiB), with a 4x improvement in IOPS and I/O bandwidth. If you're running large on-prem databases, the migration path just got easier.
OpenSearch Enhancements
GPU-accelerated vector indexing makes index creation 10x faster at one-quarter the cost. Auto-optimised vector indexes automatically evaluate different KNN algorithms to balance recall quality against query performance.
6. Networking: The Multicloud Moment
AWS Interconnect
This is the one that surprised everyone. AWS partnered with Google Cloud to offer managed, high-speed private connections between the two platforms. You can provision dedicated bandwidth on demand and establish connectivity in minutes.
The service includes quad-redundancy and MACsec encryption. Microsoft Azure support is coming in 2026.
I've been saying for years that multicloud is mostly marketing, but this is... actually useful. If you genuinely have workloads across AWS and GCP, this beats managing your own Direct Connect and Cloud Interconnect setup.
Route 53 Global Resolver
Now in preview - secure anycast DNS resolution that simplifies hybrid DNS management. One service instead of managing resolvers in each VPC.
7. Storage: Vectors and Tables
Amazon S3 Vectors (GA)
S3 now natively supports storing and querying vector embeddings. It scales up to 2 billion vectors per index (40x the preview capacity), supports up to 20 trillion vectors per bucket, and reduces costs by up to 90% compared to specialised vector databases.
For RAG, semantic search, and agentic workloads, this removes the need for a separate vector database. I'm looking at you, Pinecone bills.
S3 Tables
Built-in Intelligent-Tiering support and replication for Apache Iceberg-native tables. Makes it easier to run analytics on S3 without complex ETL pipelines.
FSx for NetApp ONTAP
Now integrates with S3, allowing file-system data to be accessed via S3 APIs. Useful for plugging existing storage into analytics and ML services without copying data around.
8. Containers and Kubernetes
Amazon EKS Capabilities
Fully managed platform capabilities for workload orchestration and cloud resource management. The goal is eliminating infrastructure maintenance while maintaining enterprise-grade reliability. If you're running vanilla Kubernetes and spending too much time on cluster management, this might help.
ECS Express Mode
Simplified container deployments for ECS. I haven't tried this one yet, but it's on my list.
9. The Deprecation Corner
AWS introduced the Product Lifecycle page in 2025, which consolidates all service availability information in one place. About time, honestly.
Key deprecations to be aware of:
- AWS Cloud9 - No longer accepting new customers. AWS recommends VS Code with remote extensions.
- AWS WAF Classic - No new WebACLs after March 31, 2025. Fully retired September 30, 2025.
- AWS Proton - Support ends October 7, 2026. New customers blocked after October 7, 2025.
- AWS SDK for JavaScript v2 - End of support September 8, 2025. Migrate to v3.
- Amazon Linux 2 - End of support extended to June 30, 2026. Migrate to Amazon Linux 2023 before then.
If you're using any of these, now's the time to plan your migration.
The CodeCommit Reversal
In a rare move, AWS reversed the CodeCommit deprecation in November 2025 after listening to customer feedback. CodeCommit is back to full General Availability with new features planned including Git Large File Storage in early 2026 and regional expansion to additional regions starting Q3 2026. The reversal acknowledged that CodeCommit's deep IAM integration, VPC endpoint support, and seamless connectivity with CodePipeline provided significant value, especially for regulated industries. AWS explicitly apologised for the inconvenience caused to customers who had begun migration planning.
10. What Does 2025 Tell Us?
Looking at the year as a whole, a few patterns emerge:
AI agents are the new compute primitive. Just like we went from servers to containers to functions, we're now going from functions to agents. AWS is betting big that the future of cloud computing involves autonomous systems that act on our behalf.
Custom silicon matters. AWS keeps investing in Graviton and Trainium because they genuinely believe they can out-price and out-perform commodity hardware for specific workloads. The numbers so far suggest they're right.
Multicloud is becoming practical. The Google partnership signals that even AWS recognises customers have legitimate multicloud needs. Expect more interoperability announcements in 2026.
Developer experience is a priority. Q Developer, Kiro, and the various IDE integrations show AWS is serious about making AI-assisted development accessible. They're not just building infrastructure - they're building the tools developers use daily.
Whether you're excited or exhausted by the pace of change, 2025 was undeniably a significant year for AWS. And if the roadmap announcements are any indication, 2026 will be even more intense.
What announcements mattered most to you? I'd love to hear what you're planning to try first.
Top comments (0)