- DapuStor has debuted a 245TB PCIe Gen5 QLC SSD for AI workloads
- The firm joins other vendors targeting hyperscalers with ultra-capacity flash
- The move reflects how AI is driving demand for dense NAND storage
AI workloads are driving demand for much larger SSDs inside hyperscale data centers, and in response to this, DapuStor has announced a PCIe Gen5 QLC SSD which scales to 245TB.
122TB versions of the drive are already in use by customers, as the high capacity SSD is intended for AI data lakes, vector databases, and large storage pools where data stays online and is accessed frequently.
These workloads generate large volumes of embeddings, logs, and video during AI training and inference. That data is accessed repeatedly rather than archived, so dense flash storage is a better choice than mechanical drives.
DapuStor joins a select group
DapuStor’s SSD uses QLC NAND. By storing four bits per cell, it increases capacity per wafer compared with TLC, improving density and cost efficiency.
Earlier concerns around QLC’s endurance and performance have been addressed via controller design, firmware management, and data placement techniques.
This announcement makes DapuStor the eighth vendor to reveal an SSD at or around the 245TB mark.
Kioxia has shown a 246TB LC9 SSD, while Sandisk has unveiled a 256TB model for AI workloads. Solidigm has confirmed plans for 245TB drives, and Micron has outlined 122TB PCIe Gen5 SSDs as part of efforts to reduce reliance on hard disks.
Huawei has approached the problem differently, pairing high-capacity SSDs with controller techniques to reduce dependence on expensive HBM in AI systems.
Samsung has outlined roadmaps extending beyond current capacities, and its archrival, SK Hynix, has teased the PS1101, its own 245TB PCIe Gen5 enterprise drive.
The move toward higher-capacity SSDs reflects changes in how AI data is stored. Training datasets, inference results, and logs are accessed frequently, favoring flash that handles random I/O while using less rack space and power.
These SSDs are obviously not for consumers or even typical enterprises. Platform requirements, pricing, and scale keep them limited to hyperscale environments, where flash capacity is becoming a critical resource.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!
And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.
