'Delivers 1.8x Better Power Efficiency': Samsung starts mass production of AI SSD for Nvidia Vera Rubin

Samsung has begun production of the PM1763 solid-state drive (SSD) for Nvidia Corp.'s upcoming Vera Rubin AI platform, which is expected to launch later this year. The PM1763 is built on the PCIe 6.0 interface, doubling the data transfer bandwidth compared with PCIe 5.0.

Samsung said the 16TB model will deliver sequential read speeds of up to 28,400 MB/s and sequential write speeds of up to 21,900 MB/s, offering roughly twice the performance of its predecessor.

The company also disclosed that it will supply HBM4 memory, SOCAMM 2 modules, and the PM1763 SSD for the Vera Rubin platform, allowing Samsung to provide a broad portfolio of memory technologies for Nvidia's next-generation AI platform.

Demand for faster storage and better efficiency

As AI servers continue to grow in scale, faster storage has become increasingly critical. Modern GPUs rely on a constant flow of data from storage, and higher SSD performance helps minimize bottlenecks, allowing AI systems to process workloads more efficiently.

More from Tech

Samsung added that the PM1763 improves power efficiency by more than 1.8 times compared with the previous generation, helping lower operating costs for AI data centres. The SSD also supports liquid-cooled server environments and incorporates enhanced security features, including encryption and safeguards for virtualized data pathways.

Samsung's Q2 earnings

Samsung recently reported preliminary revenue of 89.4 trillion won ($58.4 billion) for the April-June quarter, up from 57.2 trillion won ($37.58 billion) in the previous quarter and 46.8 trillion won ($30.7 billion) a year earlier.

The company also projected preliminary operating profit of 17.1 trillion won ($11.23 billion), compared with 8.79 trillion won ($5.78 billion) in the first quarter and 7.46 trillion won ($4.91 billion) a year ago.

Reports suggest Samsung's soaring earnings have been driven by runaway global demand for memory chips used in artificial intelligence infrastructure. The AI boom has created severe supply shortages, giving the company significant pricing power and boosting both selling prices and profit margins.

What is the Vera Rubin AI platform?

Expected to launch later this year, Nvidia's Vera Rubin AI platform is made up of more than 1.3 million components and is designed to deliver up to 10 times greater performance per watt than its predecessor. Nvidia previously told CNBC that the system is a complex web of components sourced from around the world.

At its core, the platform features 73 Rubin graphics processing units (GPUs) and 36 Vera central processing units (CPUs). The remaining components—including liquid-cooling systems, power infrastructure, and compute trays—are supplied by more than 80 vendors across at least 20 countries, including China, Vietnam, Thailand, and Mexico.

Meta has already announced plans to deploy the Vera Rubin platform in its data centres by 2027. Other expected customers include OpenAI, Anthropic, Amazon, Google, and Microsoft.