Pure Storage Launches FlashBlade//EXA™ for AI and HPC

With legacy storage systems struggling to keep up with AI’s increasing need for speed and scalability, FlashBlade//EXA eliminates the metadata bottleneck, setting a new standard for high-performance storage solutions. Initial tests indicate that FlashBlade//EXA will deliver over 10 terabytes per second in read performance within a single namespace, making it the highest-performing data storage platform in the industry.

FlashBlade//EXA is purpose-built to meet the demands of modern AI and HPC workloads. Unlike traditional storage systems, which face challenges with parallel processing, metadata performance, and high concurrency, FlashBlade//EXA features a disaggregated architecture that independently scales data and metadata. This breakthrough enables enterprises to efficiently handle massive datasets and high-concurrency operations, ensuring seamless performance for AI and HPC applications.

The rapid advancement of AI, driven by more powerful GPUs, has led to an explosion in the size and complexity of AI models. As AI models become more sophisticated and process multimodal data, such as text, images, and videos, traditional storage systems struggle to meet these requirements. FlashBlade//EXA addresses this challenge by offering a massively parallel, disaggregated architecture that provides the flexibility and scalability needed to support evolving AI and HPC workloads. By optimizing metadata performance and ensuring high throughput, FlashBlade//EXA accelerates the pace of AI model training and inference.

With FlashBlade//EXA, Pure Storage is enabling enterprises to reduce the complexity of storage management while delivering unparalleled performance. The platform’s integration of Pure Storage’s metadata engine and Purity operating system, along with off-the-shelf data nodes, ensures a cost-effective, high-performance solution for AI and HPC environments.

FlashBlade//EXA delivers over 10 terabytes per second in read performance, setting a new benchmark for the industry. It supports high-speed AI requirements, with scalable metadata IOPS to handle the demands of large AI datasets. The platform’s ability to handle massive datasets without manual tuning or additional configuration simplifies storage management, while its disaggregated architecture offers flexibility for evolving AI and HPC needs.

“FlashBlade//EXA represents a paradigm shift in data storage,” said Rob Lee, Chief Technology Officer at Pure Storage. “It combines unmatched performance and scalability, providing organizations with the tools they need to accelerate AI and HPC workloads.”

NVIDIA’s Rob Davis, Vice President of Storage Networking Technology, commented, “FlashBlade//EXA enables organizations to fully leverage the potential of AI technologies while maintaining data security, scalability, and performance for model training and inference.”

Matt Kimball, VP & Principal Analyst at Moor Insights & Strategy, added, “With the explosive growth of AI, legacy storage systems are no longer sufficient. FlashBlade//EXA’s ability to manage high-performance storage and simplify metadata management is a game-changer for organizations looking to keep up with AI and HPC demands.”

FlashBlade//EXA will be available in Summer 2025, offering organizations a transformative solution to meet the challenges of modern AI and HPC data storage.