You are designing a training job in an Azure Machine Learning workspace by using Automated ML. During training, the compute resource must scale up to handle larger datasets. You need to select the…

Compute instance is a single-node managed virtual machine and does not support multi-node automatic scaling for training workloads.

Endpoints are used for model deployment and real-time or batch inference serving, not as a training compute target.

Kubernetes cluster can support multi-node workloads but requires explicit setup, scaling policy configuration, and ongoing cluster management - it does not scale automatically out of the box in the same self-managed way.

Serverless compute abstracts cluster management entirely and dynamically scales to multiple nodes based on workload demand. This makes it the correct choice for Automated ML jobs that need to handle larger datasets without pre-configuring a fixed cluster size or managing scale-in and scale-out rules.