Amazon EMR Serverless expanded its worker configurations in 2026, offering larger instances with 32 vCPUs and up to 244 GB of memory, doubling previous capacity limits. This allows customers to run more compute- and memory-intensive Spark and Hive workloads, improving runtime performance and efficiency, notably for shuffle-heavy, skewed, or cache-dependent jobs.
Larger workers reduce data transfer overhead, decrease out-of-memory failures, and enable caching of more data in memory, ultimately enhancing job performance and cost profiles. The new instance sizes are available across all AWS Regions where EMR Serverless operates.
Source & further reading
Amazon EMR Serverless now supports larger worker sizes to run more compute and memory intensive workloads ↗Source published 2026-07-07. This report was prepared with AI assistance from the linked primary-source material. Claims and limitations should be checked against the original announcement or research.