Samsung PM1763 PCIe Gen6 SSD Enters Enterprise Production
Samsung Electronics has entered mass production of the PM1763, its first enterprise SSD based on the PCIe 6.0 interface. Designed for AI and high-performance computing (HPC) infrastructure, the PM1763 targets the rapidly emerging generation of servers built around PCIe Gen6-capable CPUs, accelerators, and storage platforms.
PCIe Gen6 doubles the signaling rate of PCIe Gen5, creating substantially more bandwidth for storage devices. As hyperscale and AI platforms begin adopting processors and accelerators with PCIe Gen6 connectivity, enterprise SSDs must evolve accordingly to prevent storage from becoming a bottleneck.
The PM1763 combines Samsung’s 9th-generation V-NAND with a new 4nm storage controller and is designed specifically for high-throughput, latency-sensitive workloads.
⚡ PCIe Gen6 Performance for AI and HPC #
The 16TB version of the Samsung PM1763 delivers sequential read performance of up to 28,400 MB/s and sequential write performance of up to 21,900 MB/s.
That represents more than twice the performance of Samsung’s previous-generation PM1753 enterprise SSD, highlighting the bandwidth increase possible when NVMe storage moves from PCIe Gen5 to PCIe Gen6.
Storage bandwidth for accelerator-heavy workloads #
Samsung estimates that the PM1763 can transfer a 40GB large language model in approximately 1.4 seconds under its stated performance conditions.
This type of workload illustrates why PCIe Gen6 SSDs are becoming increasingly relevant to AI infrastructure. Modern servers continuously move large datasets and model checkpoints between storage, host memory, CPUs, GPUs, and other accelerators. Increasing storage bandwidth can reduce the time spent waiting for data and improve overall pipeline utilization.
The transition also reflects a broader shift in how enterprise storage performance is evaluated. Rather than focusing primarily on traditional file-transfer workloads, AI and HPC deployments increasingly emphasize sustained throughput, low latency, and rapid movement of large model and dataset artifacts.
💧 Liquid-Cooling Support for Next-Generation Servers #
The PM1763 is also designed with liquid-cooled server environments in mind.
This is becoming increasingly important as AI servers move toward higher rack-level power densities. Liquid cooling allows system designers to remove more heat from CPUs, GPUs, accelerators, and other high-power components without relying exclusively on large volumes of airflow.
Storage devices increasingly need to fit into these thermal architectures rather than assuming conventional air-cooled server layouts.
Samsung’s liquid-cooling-oriented design therefore reflects a broader change in data-center architecture: the storage subsystem must be thermally compatible with the same high-density infrastructure used by modern AI accelerators.
🔐 Post-Quantum and Virtualization Security #
Beyond performance, the PM1763 incorporates security technologies intended for long-lived enterprise deployments.
The SSD supports post-quantum cryptography (PQC) algorithms designed to provide protection against future cryptographic threats posed by sufficiently capable quantum computers. For enterprise storage expected to remain in production for years, preparing for emerging cryptographic requirements can be an important part of platform security planning.
The drive also supports the TEE Device Interface Security Protocol (TDISP), which is designed to protect device communication paths in virtualized and confidential-computing environments.
These capabilities extend the PM1763 beyond raw PCIe Gen6 bandwidth by addressing security requirements that increasingly accompany modern cloud, AI, and virtualization infrastructure.
📦 4TB, 8TB, and 16TB Capacity Options #
Samsung plans the PM1763 in 4TB, 8TB, and 16TB configurations.
Rather than positioning the drive primarily as a maximum-capacity storage product, these capacities emphasize high-performance local storage inside compute servers. This strategy is consistent with previous PCIe generations, where early-generation enterprise NVMe drives have typically focused on extracting maximum performance from newly introduced host interfaces.
For AI and HPC systems, local NVMe storage can provide a high-bandwidth staging layer for datasets, model weights, checkpoints, temporary processing data, and other frequently accessed workloads.
🚀 PCIe Gen6 Storage Signals the Next Server Transition #
The most significant aspect of the PM1763 announcement may be less about the SSD itself and more about what its production status says about the server market.
PCIe Gen6 systems are beginning to move from roadmap discussions toward deployment. As next-generation CPUs and accelerators expose PCIe Gen6 connectivity, storage vendors can now build products capable of exploiting that additional I/O bandwidth.
The Samsung PM1763 therefore represents an early step in the PCIe Gen6 NVMe transition. With 28.4 GB/s-class sequential reads, liquid-cooling compatibility, and security features aimed at long-lived enterprise infrastructure, it is designed for the increasingly demanding storage requirements of AI and HPC servers.