Does AWS Lambda have limitations when processing training data larger than 100 MB?

Yes. Lambda’s memory limit is 10 GB and the maximum execution time is 15 minutes. If a training dataset exceeds 100 MB, a single invocation cannot load it entirely into memory. Instead, you must stream the data from S3, split it into smaller chunks, or use multiple Lambda invocations orchestrated by Step Functions. Additionally, the deployment package size limits (50 MB zipped) mean you cannot bundle large libraries directly; you would need to use layers or external dependencies. For truly large training jobs, services like SageMaker or EMR are more appropriate because they provide longer runtimes, larger memory, and built‑in distributed training capabilities.