PCIe 6.0 Storage Moves From Roadmap To Server Rack

Samsung’s PM1763 is not the kind of SSD that drops into a gaming PC for a quick upgrade. As of July 21, 2026, it is a newly mass-produced enterprise drive aimed at AI and HPC servers, where storage has become part of the performance conversation alongside GPUs, accelerators, networking and memory. Samsung announced mass production on July 8, 2026, positioning the PM1763 as a PCIe 6.0-based enterprise SSD for next-generation AI infrastructure. (news.samsungsemiconductor.com)

Article contains affiliate links, commission may be earned.

The Headline Specs: Up To 28,400MB/s Reads

The main number is hard to ignore: the 16TB PM1763 configuration is rated for sequential reads of up to 28,400MB/s and sequential writes of up to 21,900MB/s, according to Samsung’s launch materials. The initial mass-production capacities listed by Samsung are 4TB, 8TB and 16TB, with the 16TB model more than doubling the performance of the previous PM1753 generation in Samsung’s comparison. (news.samsungsemiconductor.com)

  • Interface: PCIe Gen6-based enterprise SSD
  • Protocol/platform support: NVMe 2.1 and OCP 2.6 listed on Samsung’s product page
  • NAND: Samsung 9th-generation V-NAND
  • Controller: newly developed 4nm controller
  • Launch capacities: 4TB, 8TB and 16TB
  • Peak listed 16TB performance: up to 28,400MB/s sequential read and 21,900MB/s sequential write


See SANDISK Optimus GX PRO 8100 PCIe 5 SSD price

Why AI Servers Care About Storage Speed

The interesting part is not just that the PM1763 is fast; it is where that speed is being targeted. AI clusters have to move huge datasets, model weights, checkpoints and retrieval data between storage, CPUs, GPUs and accelerators. When those systems are waiting on data, expensive compute can sit underused. Samsung says a 40GB large language model can be transferred in about 1.4 seconds using the PM1763, which frames the drive around model-loading latency rather than everyday file-copy convenience. (news.samsungsemiconductor.com)

View NVIDIA DGX Spark Personal AI Desktop Supercomputer on partner website

The Controller Is Becoming A Bigger Part Of The Story

Enterprise SSDs are no longer just NAND packages behind a basic controller. The PM1763 pairs Samsung’s 9th-generation V-NAND with a 4nm controller, and that matters because high-end SSD controllers now have to handle signal integrity, power behavior, error recovery, telemetry and security at very high throughput. Samsung also says the drive improves power efficiency by more than 1.8 times versus its predecessor, based on its internal comparison. That is important in AI server rooms where storage performance is useful only if it can be sustained within tight power and thermal limits. (news.samsungsemiconductor.com)

Buy Crucial T705 13 GB/s NVMe SSD here

Built For Dense, Liquid-Cooled Infrastructure

Samsung describes the PM1763 as optimized for liquid-cooled server environments through Direct-to-Chip cooling, a design direction that fits the broader move toward denser AI racks. Its product page also lists support for E1.S, E3.S and U.2 form factors, while noting that U.2 support is tied to PCIe Gen5 rather than PCIe Gen6. Security is also part of the platform pitch, with support listed for PQC, SPDM 1.4, CNSA 2.0 and TDISP-based link encryption. (news.samsungsemiconductor.com)

This makes the PM1763 less of a consumer storage story and more of a signpost for AI infrastructure. The drive is meant for systems where model movement, checkpointing and accelerator feed rates can become real constraints. It will not make sense for ordinary desktops, but for server builders, cloud platforms and HPC operators, PCIe 6.0 SSDs like this show how storage is being pulled closer to the compute bottleneck fight.

See Samsung 990 PRO NVMe SSD price