Google's Cloud Storage Gets Faster and Smarter for AI

At its Next '26 conference, Google unveiled substantial enhancements to its Cloud Storage offerings, designed to optimize data delivery and management for artificial intelligence (AI) workloads. The upgrades span performance, intelligence, and overall management capabilities, ensuring that cloud storage can keep pace with the increasing demands of AI models and applications. A key highlight is the introduction of Cloud Storage Rapid, a new object storage solution engineered for extreme throughput, frequent I/O operations, and sub-millisecond latency. This service, built on Google's Colossus distributed storage architecture, boasts impressive figures such as over 15 TB/s bandwidth and 20 million requests per second within a single zone. It supports high-performance gRPC and S3-compatible APIs, significantly accelerating checkpoint restoration by five times and checkpoint writing by 3.2 times compared to traditional object storage. Furthermore, Cloud Storage Rapid is natively compatible with popular AI frameworks like PyTorch and JAX, reducing GPU idle time by 50% and speeding up data loading for multimodal training by 2.5 times. For high-performance file storage needs, Google introduced Managed Lustre with a Dynamic Tier. Powered by DDN Lustre and EXAScaler, this parallel file system offers up to 10 TB/s throughput, marking a tenfold year-over-year increase and outperforming rival hyperscaler Lustre deployments by 4 to 20 times. The new Dynamic Tier, priced at $0.06 per GB monthly, provides low-latency performance through persistent disks, simplifying billing without complex tiering rules. Beyond performance, Google is enhancing the intelligence and management of its cloud storage. This includes automated metadata tagging, enabling more efficient data discovery and organization for AI agents. A Cloud Storage MCP server also facilitates standardized protocol-based data reading, writing, and analysis for AI agents. Storage Intelligence has been upgraded with zero-configuration dashboards that can detect cost anomalies and integrate Data Security Posture Management (DSPM) governance to identify storage security risks automatically. New object event and bucket activity tables provide deeper cost auditing and streamlined operations, while upgraded batch tools allow for bulk adjustments across billions of objects and multiple buckets. These comprehensive updates aim to deliver a smarter, more efficient, and more secure cloud storage environment for AI development and deployment.