What AI memory technologies did SK hynix showcase at the OCP Global Summit 2026?

SK hynix showcased MARS, or Memory-Centric AI Rack-Scale System, High Bandwidth Flash, and a second-generation 256-gigabyte memory product based on CXL 3.2. The technologies are intended to support AI infrastructure and next-generation server ecosystems beyond conventional HBM.

How does SK hynix’s second-generation CXL 3.2 memory improve on its predecessor?

The second-generation 256-gigabyte CXL 3.2 memory product provides more than twice the bandwidth of the previous generation and improves energy efficiency by approximately 30 percent.

What is SK hynix’s MARS architecture designed to do?

MARS is a disaggregated rack-scale architecture designed to coordinate computing and memory resources seamlessly. Kim Ho-sik, a senior vice president at SK hynix, introduced the system during the company’s exhibition at the summit.

SK hynix showcases next-generation AI memory technologies beyond HBM

Listen

Summary

SK hynix showcased next-generation AI memory technologies at the OCP Global Summit 2026 in California from Tuesday to Thursday. The company presented MARS, a disaggregated rack-scale architecture, and progress in High Bandwidth Flash to expand AI infrastructure beyond conventional HBM. It also unveiled a second-generation 256-gigabyte CXL 3.2 memory product with more than twice the predecessor’s bandwidth and approximately 30 percent better energy efficiency.

Key Facts

The OCP Global Summit is an annual conference organized by the Open Compute Project Foundation for engineers and industry leaders.
SK hynix divided its exhibition into a front memory section for flagship products and a larger rear system demo section for next-generation server memory solutions.
MARS stands for Memory-Centric AI Rack-Scale System and is designed to coordinate computing and memory resources across a rack.
The company is developing High Bandwidth Flash to address growing demand for memory capacity.
The second-generation 256-gigabyte CXL 3.2 product delivers more than twice the bandwidth of its predecessor and about 30 percent higher energy efficiency.

By Yonhap

Published Oct 16, 2026 10:52 am KST

SK hynix's booth at the OCP Global Summit 2026 held in California / Courtesy of SK hynix

SK hynix said Friday that it has showcased its next-generation artificial intelligence (AI) memory technologies aimed at expanding the boundaries of AI infrastructure beyond conventional high-bandwidth memory (HBM).

The company presented the technologies at the OCP Global Summit 2026, held in California from Tuesday to Thursday (local time). The summit is an annual conference organized by the Open Compute Project Foundation that brings together engineers and industry leaders.

The Korean chip giant said it divided its exhibition space into two sections.

The front section, dubbed the memory section, showcased the company's flagship products, while the larger rear section, called the system demo section, featured memory solutions designed to support next-generation server ecosystems.

Kim Ho-sik, a senior vice president at SK hynix, introduced MARS (Memory-Centric AI Rack-Scale System), a disaggregated rack-scale architecture designed to coordinate computing and memory resources seamlessly.

Kim highlighted progress in High Bandwidth Flash (HBF), a technology aimed at addressing the industry's growing demand for memory capacity.

The company also unveiled its second-generation 256-gigabyte memory product based on Compute Express Link (CXL) 3.2, offering more than twice the bandwidth of its predecessor and an approximately 30 percent improvement in energy efficiency.

Explore More