Google Cloud Launches Cloud Storage Rapid: High-Performance Object Storage for AI Workloads
Key Takeaways
- ▸Google Cloud addresses critical storage bottleneck for AI/ML by launching Cloud Storage Rapid, combining high performance with object storage simplicity
- ▸Rapid Bucket delivers massive throughput (15+ TB/s), ultra-low latency (sub-millisecond), and 20M queries/second—designed to keep GPUs and TPUs fully saturated
- ▸Real-world performance gains include 50% less blocked GPU time, 5x faster checkpoint restores, and 3.2x faster writes compared to traditional object storage
Summary
Google Cloud announced Cloud Storage Rapid at Google Cloud Next '26, a new family of object storage capabilities designed specifically for data-intensive AI and analytics workloads. The offering consists of Rapid Bucket (a high-performance zonal object storage service in general availability) and Rapid Cache (which accelerates reads and collocates compute with data). Cloud Storage Rapid directly addresses storage bottlenecks that have become critical as organizations scale to trillion-parameter models and deploy inference at global scale.
Rapid Bucket delivers exceptional performance metrics: up to 20 million queries per second, sub-millisecond latency, and 15+ TB/s of aggregate read throughput from a single zonal bucket. The service combines block-like storage latency with parallel filesystem throughput while maintaining object storage's scalability and ease of use. It includes new capabilities like native appends, unlimited concurrent readers during writes, and vectored reads.
In real-world deployment scenarios, Rapid Bucket demonstrated significant efficiency gains: 50% reduction in blocked GPU time during multi-modal training, checkpoint restores up to 5x faster than traditional object storage, and 3.2x faster checkpoint writes. These improvements directly translate to higher accelerator utilization and reduced costs for expensive compute infrastructure during training and inference workloads.
- Positions Google to compete in AI infrastructure as organizations scale to trillion-parameter models and enterprise autonomous agents
Editorial Opinion
Cloud Storage Rapid tackles a fundamental problem in modern AI: storage is often the limiting factor preventing accelerators from reaching peak utilization. By offering 15+ TB/s throughput with sub-millisecond latency while maintaining the simplicity of object storage, Google is making a compelling case for consolidation. Organizations won't need to choose between specialized niche storage and cloud-native object stores—Cloud Storage Rapid bridges that gap in a way that could reshape infrastructure decisions for large-scale AI teams.
