Samsung PM1763 Enterprise SSD: Specs, Performance, and Nvidia Vera Rubin AI Platform Details

• Sequential Read: Up to 28,400 MB/s
• Sequential Write: Up to 21,900 MB/s (as reported by multiple press sources; Samsung's product page lists up to 21,000 MB/s)
• Performance vs. Predecessor: Approximately 2× faster than the PCIe 5.0-based PM1753
• LLM Transfer Time: A 40 GB large language model (comparable to a 4-bit quantized Llama 3 70B model) can be transferred in approximately 1.4 seconds
• Interface: PCIe 6.0 with PAM4 signaling, doubling bandwidth over the previous generation (PCIe 5.0)

Capacities and Form-Factors

The PM1763 is initially offered in three capacities :

• 4 TB
• 8 TB
• 16 TB

The official Samsung Semiconductor product page also lists support for capacities up to 64 TB in E1.S, E3.S, and U.2 form factors, enabled by the drive's 9th-generation V-NAND and a new 4-nanometer controller .

Cooling Technology

Next-generation AI servers are increasingly moving toward liquid-cooled architectures to manage thermal output. The PM1763 is designed for this environment with Direct-to-Chip (D2C) liquid cooling .

• How it works: A cold plate attaches directly to the SSD die, enabling stable operation under high load
• Benefit: Maintains peak performance over extended periods during demanding AI training and inference workloads

Power Efficiency

Improving power efficiency is a critical goal for data centers, where energy costs and cooling are major operational expenses. The PM1763 delivers a 1.8× improvement in power efficiency over the PM1753, helping to reduce overall operating costs .

Security Features

Recognizing the evolving threat landscape for AI-driven data centers, Samsung equipped the PM1763 with advanced security capabilities designed to protect data at rest, in transit, and in use .

• Post-Quantum Cryptography (PQC): Includes PQC encryption algorithms to protect data against future quantum-computing-based attacks
• TEE Device Interface Security Protocol (TDISP): Supports Trusted Execution Environment (TEE) security for virtualized and AI-driven data center infrastructures, providing link-level encryption
• Additional Protocols: Also supports SPDM (Security Protocol and Data Model) 1.4 and CNSA (Commercial National Security Algorithm Suite) 2.0

Validation and Supply Status

The PM1763 has passed Nvidia's qualification for the Vera Rubin platform and entered mass production as of July 7-8, 2026 . According to industry reports, Samsung and Micron are currently the two companies mass-producing eSSDs for Vera Rubin, while SK Hynix is still in the preparation stage and has not yet begun supply for this platform .

Broader Context: HBM4 Certification and Industry Trends

The PM1763 launch is part of a larger strategy by Samsung to provide a "total memory solution" for Nvidia's AI platforms, extending beyond its role as an HBM4 supplier .

HBM4 Certification: On June 5, 2026, Nvidia CEO Jensen Huang confirmed that all three major memory manufacturers — Samsung, SK Hynix, and Micron — have been certified to supply HBM4 memory for the Vera Rubin platform . This marks a significant milestone for the AI supply chain, ensuring a diversified and competitive supply base.

Industry Trends: The shift toward AI inference workloads is expected to significantly increase NAND flash demand. High-performance eSSDs like the PM1763 are designed to reduce data-transfer bottlenecks between storage, CPUs, and GPUs in both training and inference pipelines, enabling faster and more efficient AI processing .