Enterprise-class SSD storage is a solid-state hard disk designed to cope with 7×24 hours uninterrupted operation and high-intensity workload of data centers, servers and enterprise storage systems. It can release excellent performance continuously, stably and reliably under extreme pressure.
 

In the AI model training scene, large models often need to deal with massive data of PB level, and the data must be transmitted to GPU at high speed. If the data is still stored in the traditional mechanical hard disk, the low reading speed will cause the expensive GPU cluster to "idle" frequently because of waiting for the data, which greatly affects the training efficiency. The storage system built with enterprise-class SSD memory can continuously deliver data to GPU with extremely high throughput, significantly shortening the model training cycle and effectively reducing the overall training cost.
 

Driven by the AI wave, the demand for enterprise-class SSD memory continues to rise and the market performance is strong. After the NAND Flash market experienced a reduction in production in the first half of 2025, the inventory was gradually depleted, and the supply and demand structure has been significantly improved. At the same time, upstream manufacturers shifted their production capacity to high-margin products, which led to the contraction of market circulation supply. On the demand side, enterprises have benefited from increasing investment in AI field and the batch shipment of NVIDIA's new generation Blackwell chips to jointly support the growth of enterprise-level SSD memory market. It is expected that its contract price will increase by 5% to 10% in the third quarter, and it is expected to continue to achieve 5% to 10% growth in the fourth quarter.