Samsung Electronics Unveils AI Memory and Storage Innovation Strategy, Going Beyond HBM to CXL and SSDs

Jointly Demonstrates CXL-Based Memory Expansion Technology with AMD
Executive Shim Hojun: "Speed, Capacity Expansion, and Data Management Are Key"
Expands Global Partnerships Through Memory, Storage, and Software Optimization

Samsung Electronics is stepping up its efforts to capture the artificial intelligence (AI) infrastructure market, leveraging next-generation high-bandwidth memory (HBM), Compute Express Link (CXL)-based memory expansion technology, and enterprise solid-state drives (SSDs). The company aims to expand its partnerships with global AI companies through a system optimization strategy encompassing memory, storage, and software, moving beyond performance improvements to individual memory products.

Samsung Electronics said it is participating as a Diamond Sponsor, the event's highest sponsorship tier, in the OCP Global Summit 2026, which opened in the United States on October 12 (local time), and has unveiled next-generation memory and storage technologies as well as results from collaborations with global partners. Under the theme "AI System Design-Innovation, Together," the company is showcasing HBM5, customized HBM, CXL-based memory expansion technology, and next-generation SSDs.

At the exhibition, Samsung is jointly demonstrating a CXL-based memory expansion solution with AMD and showcasing Groq language processing unit (LPU) wafers and NVIDIA's Groq 3 LPX system.

Shim Hojun, executive director of the AE Team at the Memory Business of Samsung Electronics' Device Solutions (DS) Division, speaks at the OCP Global Summit 2026 in Silicon Valley, U.S., on Oct. 12 (local time). Samsung Electronics.

Shim Hojun, managing director of the AE Team at Samsung Electronics' Memory Business, DS Division, delivered a presentation that day titled "Scaling Innovation in the AI Era: Maximizing Computing Performance with Scalable Memory" and outlined a memory architecture strategy to address bottlenecks in AI infrastructure. As the spread of AI agents rapidly increases memory traffic and KV cache requirements during inference, he explained, it is necessary to increase memory bandwidth, flexibly expand capacity, and optimize data storage tiers.

In response, Samsung presented three pillars of its next-generation AI memory strategy: ▲improving data processing speeds ▲expanding memory capacity ▲improving data storage and management efficiency.

In the high-performance memory field, Samsung introduced next-generation technologies such as HBM5, customized HBM, and processing-in-memory (PIM). For memory expansion, it outlined ways to flexibly increase system memory capacity and improve resource utilization using CXL and CMM-D. In storage, it unveiled a strategy to improve the storage performance and management efficiency of large-scale AI data through technologies including next-generation SSDs, QLC, and FDP.

In particular, the joint demonstration with AMD showcases how CXL-based memory expansion technology can meet growing memory demands during AI inference. CMM-D-based memory pooling technology enables multiple systems to use memory resources efficiently and flexibly secure the capacity they need.

The exhibition space is organized around technologies required for AI computing and inference. The "Tokenization Solutions" zone features HBM4E, HBM5, customized HBM concept products, and LPDDR5X-PIM. A demonstration of an LPDDR5X-PIM-based edge AI accelerator showcases technology that performs some computations within memory, reducing data movement between the processor and memory and improving power efficiency.

The "KV Cache Solutions" zone features CXL memory, enterprise SSDs, KV cache, and storage optimization technologies. Samsung is also presenting ways to improve data processing efficiency in AI systems through a range of demonstrations, including AI memory nodes, vector databases (VectorDB), direct storage, and liquid cooling.

Samsung Electronics is also expanding its joint design efforts with global customers. Through Samsung MAX Lab, which validates memory and storage technologies in real-world AI environments, the company can evaluate PIM, CMM-D, and next-generation SSDs before commercialization and incorporate the system requirements of individual customers. Through its Global Open Ecosystem Team (GOST), Samsung also participates in open-source software development and standardization, while pursuing system-level technology collaborations in areas such as KV cache optimization and direct storage.

Samsung Electronics has also unveiled next-generation memory and storage technologies at major global AI and semiconductor events this year. At NVIDIA GTC in March, it showcased the NVIDIA Vera Rubin platform equipped with HBM4, SOCAMM2, and PM1763. At Computex in Taiwan in June and FMS in the United States in August, it unveiled a technology roadmap for AI infrastructure, including HBM4E, HBM5, and next-generation three-dimensional memory architectures.

Samsung Electronics is expanding its product portfolio beyond HBM to include system memory, enterprise SSDs, and NAND flash, while broadening its technological reach through three-dimensional (3D) architectures combining memory and logic, PIM, CXL, and advanced packaging. Its strategy is to improve AI system performance and power efficiency through the integrated optimization of memory, storage, and software, going beyond improvements to individual semiconductor performance, and to expand collaborations with global customers.

This content was produced with the assistance of AI translation services.