AWS raises EMR Serverless shuffle limit to 1 TB per job

AWS announced on October 1, 2026, that Serverless Storage on Amazon EMR Serverless now supports shuffle operations of up to 1 TB, raising the previous 200 GB per-job limit. AWS says the enhancement targets Apache Spark workloads with substantial shuffle data, including joins, aggregations and sorting (AWS announcement).

The update also adds spill support. AWS describes this as allowing memory-intensive operations to offload data to disk when necessary, and says it improves reliability for demanding analytical workloads. The announcement does not quantify that benefit.

A shuffle allowance, not a dataset-size ceiling

For capacity planning, read the 1 TB figure as a shuffle allowance rather than an input-dataset limit. AWS identifies joins across multi-terabyte datasets as a target use case; it does not present 1 TB as the maximum size of the data a job can process.

For previously constrained workloads, compare observed shuffle volume with the new allowance and examine spill behavior. That is a more relevant starting point than source-table size alone.

The announcement provides no benchmark results or pricing comparison. The higher supported shuffle volume therefore should not be treated as evidence that a particular job will run faster or cost less.

Availability depends on release and Region

AWS lists availability with emr-7.14, emr-spark-8.1 and later, in 18 AWS Regions where EMR Serverless is available. It directs customers to the Amazon EMR documentation for the supported Region list and applicable limits.

Before relying on the larger allowance, confirm the workload’s release, target Region and applicable limit. For migration planning, this is an expanded shuffle allowance—not a replacement for checking the job’s execution characteristics.