Samsung has unveiled a 256TB QLC data-centre SSD, previewed NAND exceeding 400 layers and introduced zNAND-O as it targets the rising storage demands of AI systems.
Samsung has used this week’s memory-industry showcases to underline how aggressively it is targeting the AI storage market. At events in Shenzhen and Santa Clara, the company presented a series of enterprise products and next-generation NAND technologies aimed at data centres, liquid-cooled servers and on-device AI, as demand rises for denser storage, faster interfaces and lower power use.
Among the headline products is the BM1773, a QLC solid-state drive for data centres that Samsung says delivers 245.76TB of usable capacity, rounded up for marketing purposes to 256TB. Built in the compact E3.S form factor, it is being positioned as the smallest SSD of its class. Samsung says the drive uses ninth-generation V-NAND and a 16-channel PCIe 5.0 design, with sequential write speeds of up to 4.5GB/s, random write performance of up to 50,000 IOPS and an endurance rating of 0.6 drive writes per day, which the company claims is roughly double that of its previous QLC generation.
Samsung also highlighted the PM1763, a PCIe 6.0 enterprise SSD intended for intensive AI and server workloads. According to Samsung, the drive reaches up to 28.4GB/s on sequential reads and 21GB/s on writes, with capacities ranging from 3.84TB to 61.44TB. The company says the product is designed for liquid-cooled systems and includes security features such as post-quantum cryptography, SPDM and TDISP, reflecting growing concern about both throughput and protection in next-generation data-centre hardware.
Beyond finished products, Samsung used the events to preview its V10 BV-NAND platform, which the company says exceeds 400 layers and is built using wafer-bonding techniques that separate memory cells and peripheral circuitry. Samsung claims this approach lifts bit density by about 58% compared with the previous generation, while improving I/O speed by more than 33% and cutting power use by more than a quarter. In a separate technical note, the company said the architecture can deliver more than 4.8Gbps I/O speeds, but it has not yet announced commercial availability.
The company also unveiled zNAND-O, a storage technology aimed at running larger AI models locally on devices. Samsung says it offers 10 times the bit density of DRAM and is faster than conventional NAND flash for certain access patterns. In a simulation using Moonshot AI’s Kimi K2-1T model, Samsung said a system based on zNAND-O cut memory cost to one-quarter while keeping performance steady and supported a model nearly twice as large. Taken together, the announcements suggest Samsung is trying to cover the full memory stack, from ultra-high-capacity enterprise SSDs to specialised components for future AI systems.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
