Serverless in 2025: What’s Changed and What’s Coming in 2026
Serverless has evolved beyond Lambda. In 2025, serverless encompasses databases, containers, AI inference, and more. Here’s the state of serverless on AWS and what to expect next year.
The Serverless Landscape 2025
| Category |
Serverless Options |
| Compute |
Lambda, Fargate, App Runner |
| Databases |
Aurora Serverless v2, DynamoDB, ElastiCache Serverless |
| Analytics |
Athena, Redshift Serverless, EMR Serverless |
| AI/ML |
Bedrock, SageMaker Serverless Inference |
| Integration |
EventBridge, Step Functions, SQS, SNS |
Lambda Updates in 2025
- SnapStart for Python/.NET: Sub-100ms cold starts
- Year-long coordination: Wait for events without paying
- S3 failed-event destinations: Better dead-letter handling
- Improved streaming: Response streaming for all runtimes
Modern Serverless Architecture
# 2025 serverless stack
Resources:
# API Layer
Api:
Type: AWS::Serverless::HttpApi
Properties:
StageName: prod
# Compute
Function:
Type: AWS::Serverless::Function
Properties:
Runtime: python3.12
SnapStart:
ApplyOn: PublishedVersions
Events:
Api:
Type: HttpApi
# Database
Database:
Type: AWS::RDS::DBCluster
Properties:
Engine: aurora-postgresql
ServerlessV2ScalingConfiguration:
MinCapacity: 0.5
MaxCapacity: 16
# AI
BedrockAgent:
Type: AWS::Bedrock::Agent
Properties:
FoundationModel: amazon.nova-pro-v1
2026 Predictions
What’s Coming
- SnapStart everywhere: All runtimes, faster cold starts
- Serverless GPUs: Pay-per-inference for AI workloads
- Deeper Bedrock integration: Serverless AI agents
- Cost improvements: Sub-millisecond billing
- Edge serverless: Lambda@Edge improvements
When to Go Serverless
| Go Serverless |
Consider Containers |
| Variable/unpredictable traffic |
Steady-state high throughput |
| Event-driven workloads |
Long-running processes |
| Rapid development priority |
Complex runtime requirements |
