Samsung Not Only Supplies HBM, AI Server SSD for Nvidia Begins Mass Production

SSD PM1763. (Photo: Samsung.)

JAKARTA - Samsung Electronics has begun mass-producing enterprise SSDs for Nvidia's next-generation AI platform, Vera Rubin. This move makes the South Korean chip giant not only rely on high bandwidth memory or HBM, but also go deeper into AI server storage.

Korea JoongAng Daily, quoted Thursday, July 9, reported that the new SSD is called PM1763. This product is based on PCIe 6.0 and is designed for AI servers that require very large data supplies.

Samsung previously supplied HBM for the Vera Rubin platform. HBM is a high-speed memory that is widely used in AI systems. With PM1763, Samsung's product line for AI servers is becoming wider, from fast memory to large-scale data storage.

PM1763 was first introduced at Nvidia's GTC 2026 developer conference in March. At the time, Samsung showcased the SSD as a memory solution for Vera Rubin along with the sixth-generation HBM4 and Small Outline Compression Attached Memory Module 2 or Socamn2.

Enterprise SSDs are SSDs for servers or data centers, not for regular laptops. These devices store large amounts of data on AI servers, then send it quickly to GPUs and CPUs.

In AI systems, data transfer speed is now increasingly important. Generative AI models continue to grow, so system performance is not only determined by computing power, but also by how fast data can be transferred.

PM1763 uses Samsung's ninth-generation V-NAND and a newly developed 4-nanometer controller. For a 16TB configuration, this SSD has sequential read speeds of up to 28,400MB per second and write speeds of up to 21,900MB per second.

That's more than twice as fast as its predecessor, the PM1753. According to Samsung, the PM1763 can transfer a 40GB large-language model file in about 1.4 seconds.

Its power efficiency is also more than 1.8 times better than the previous generation. Samsung designed this SSD for liquid-cooled AI server environments, a system commonly used to handle the heat from AI workloads.

PM1763 also supports security features, including post-quantum cryptography. This technology is prepared to face future security threats as quantum computing develops further.

Vice President and Head of Memory Product Planning at Samsung Electronics Choi Jang-seok said the PM1763 has completed validation for the next generation AI platform.

"Built on industry-leading performance, PM1763 has successfully completed validation for the next-generation AI platform and is well positioned to support the growing AI infrastructure needs," Choi said, quoted by Korea JoongAng Daily.

According to Choi, the ever-increasing size and complexity of AI models make the need for memory capacity and operational efficiency increase.

"As AI models continue to grow in size and complexity, PM1763 will be a key solution that allows customers to efficiently increase memory capacity and optimize AI operations," he said.