serverless compute

AWS Lambda

AWS Lambda is the foundational event-driven serverless computing platform executing code in response to events and HTTP triggers.

What the product actually does

AWS Lambda executes code functions inside ephemeral Firecracker microVMs in response to hundreds of AWS event sources, including API Gateway HTTP requests, S3 bucket mutations, DynamoDB stream records, and SQS queue messages.

The runtime scales from zero to tens of thousands of concurrent execution environments in seconds, automatically allocating compute, network interfaces, and runtime dependencies without manual infrastructure management.

How pricing works once you are inside

Charges are billed strictly on execution count (per 1M requests) and duration measured in milliseconds multiplied by allocated memory (GB-seconds).

Arm-based Graviton2 processors provide up to 34% better price-performance compared to x86. Provisioned Concurrency, while eliminating cold starts, incurs a continuous hourly reservation fee.

Where it breaks down

Hard execution timeouts of 15 minutes prevent Lambda from handling continuous batch processing or persistent WebSocket connections.

VPC networking, while improved, can introduce initialization latency, and relational database connection pooling requires Amazon RDS Proxy to prevent port exhaustion.

What switching costs

Moderate. The application code is standard language code, but event payloads, IAM policies, and cloud infrastructure definitions (CloudFormation/Terraform) tie deeply into AWS.

Sources

AWS Lambda Developer Guide and Compute Pricing Documentation, October 2026.