S3 Express One Zone Expands to Seven New Regions to Accelerate AI and Interactive Analytics

Amazon Web Services has expanded its Amazon S3 Express One Zone storage class into seven additional geographic regions: Asia Pacific (Singapore, Sydney, Seoul), South America (São Paulo), US West (Northern California), Canada (Central), and Europe (Paris), bringing its total regional presence to 15 AWS regions. The S3 Express One Zone tier uses dedicated directory buckets deployed within a single AWS Availability Zone to deliver single-digit millisecond data access speeds—up to 10x faster than S3 Standard—alongside an 80% reduction in request billing costs. This expansion allows teams running compute in these additional markets to co-locate latency-sensitive workloads with fast, persistent object storage. Traditionally, object storage was engineered primarily for high durability and massive scale across multiple availability zones, accepting higher latency and standard per-request transaction costs as trade-offs. However, the rise of large-scale AI training loops, vector retrieval pipelines, real-time analytics engines, and high-concurrency microservices has exposed S3 request costs and tail latencies as major operational bottlenecks. When training clusters or analytics clusters repeatedly query small files or checkpoints, per-request fees and network traversals quickly dominate infrastructure bills. The industry response across major cloud providers has been to specialize object storage layers into distinct performance tiers, blurring the boundary between traditional object stores, in-memory caches, and local high-performance file systems. In practice, this expansion gives platform engineers and data architects a straightforward mechanism to eliminate intermediate caching layers like Redis clusters or specialized POSIX mounts for read-heavy, low-latency intermediate state. Workloads using PyTorch data loaders, Trino query engines, or distributed AI inference engines can ingest and stream batches directly from directory buckets located within the exact AZ of their compute nodes. Teams should note the architectural trade-off: because Express One Zone stores data within a single Availability Zone, disaster recovery plans for mission-critical datasets must still incorporate automated replication or periodic backup policies into multi-AZ standard storage tiers.