Samsung Electronics said Wednesday that it has started mass production of PM1763, an enterprise solid-state drive built for the storage demands of AI data centers. The drive, announced in Seoul, pairs Samsung’s ninth-generation V-NAND flash memory with a newly developed 4-nanometer controller, and the company says it moves data more than twice as fast as its predecessor, the PM1753.
The headline numbers are the kind that make server architects pay attention. In its 16-terabyte configuration, the PM1763 delivers sequential read speeds of up to 28,400 megabytes per second and write speeds of 21,900 MB/s. Samsung says the drive can transfer a 40-gigabyte large language model in about 1.4 seconds, cutting the latency that slows training and inference when processors and AI accelerators wait on storage.
“PM1763 has successfully completed validation for next-generation AI platforms and is well positioned to support evolving AI infrastructure requirements,” said Jangseok Choi, vice president and head of memory product planning at Samsung. “As AI models continue to grow in size and complexity, PM1763 will serve as a key solution that enables customers to efficiently scale memory capacity and optimize AI operations.”
The drive is designed for liquid-cooled servers, a shift that is remaking data center design. Direct-to-chip cooling routes coolant to the hottest components, and Samsung engineered the PM1763 to sit in that environment without losing performance. Power efficiency is up more than 1.8 times versus the previous generation, a selling point for operators whose electricity bills have become a board-level topic.
Storage has quietly become a bottleneck in the AI build-out. Training runs generate data faster than conventional drives can absorb it, and inference workloads multiply that pressure at lower latency tolerances. Hyperscale operators have been rebuilding their storage tiers around faster, denser devices, and enterprise SSDs have become the fastest-growing segment of the NAND market. Samsung is defending its position there while competing for high-bandwidth memory orders in the same data centers.
The PM1763 also carries a security package aimed at AI-era threats. It supports post-quantum cryptography algorithms designed to resist future quantum computers, along with the TEE Device Interface Security Protocol for securing data pathways in virtualized environments. Analysts said security features of that kind are becoming table stakes in procurement decisions for government and financial customers.
The product cycle itself is part of the story. PM1763 is Samsung’s first enterprise drive built around the PCIe 6.0 interface, which doubles the data rate of the PCIe 5.0 standard that powered PM1753. Servers have been slower to adopt each new PCIe generation, but AI platforms are pulling the transition forward, and Samsung is betting that hyperscalers will rebuild their storage tiers around the faster interface rather than wait out the cycle.
Samsung’s timing reflects the shape of AI infrastructure spending. Training clusters are being built out at record pace, and inference, the process of running models once they are trained, is expanding even faster as chatbots and agents move into daily business use. Each of those workloads pulls data through storage, and vendors that offer speed with power efficiency are winning the sockets, according to analysts who track the storage market.
The competitive field is crowded. SK Hynix, Samsung’s domestic rival, is pushing its own enterprise SSD line backed by the same AI demand wave, Micron is shipping its 2650-series drives, and Intel’s former NAND business, now Solidigm, has been an early adopter of liquid-coolable SSD designs. Samsung’s pitch is vertical integration: the flash, the controller, and the packaging all come from its own fabs, which the company says gives it a cost and supply advantage rivals cannot easily match.
The launch also gives Samsung’s memory division something to point to beyond HBM, where it has been catching up after yield problems in the premium segment. NAND has its own cycle, and analysts expect supply discipline among the major makers to keep prices firm through the year even as AI demand pulls volumes upward. The company’s semiconductor business has leaned on memory pricing recovery to steady its results, and a flagship enterprise drive helps it hold share in the most profitable part of the NAND market.
Enterprise SSD demand is being driven by the same forces reshaping the rest of the memory industry. AI servers carry far more storage than conventional machines, and analysts who follow the NAND market expect enterprise drives to grow faster than the consumer segment for years. Samsung, SK Hynix, and Micron are all expanding production of high-capacity drives, betting that AI will do for storage what smartphones once did.
Samsung said the PM1763 will be offered in 4TB, 8TB, and 16TB capacities, with validation completed for next-generation AI platforms. The company did not disclose production volumes or customers, but said the drive is aligned with emerging server architectures, including high-density, liquid-cooled deployments. For operators racing to stand up AI capacity, the arrival of a faster, cooler drive from the largest memory maker is one less constraint to plan around.
