Samsung showcases next-generation AI memory at OCP, highlights AMD partnership

Samsung Electronics showcased next-generation memory and storage technologies in the United States on Monday, highlighting AI infrastructure and cooperation with AMD. The company jointly demonstrated a CXL-based memory expansion solution with AMD and displayed products associated with Groq and Nvidia. Samsung also presented customized HBM, HBM5, processing-in-memory technology, 2-nanometer wafers and AI-focused storage solutions. The showcase underscored Samsung’s expansion beyond HBM into system memory, SSDs and NAND flash.

The CXL-based solution uses Samsung’s CMM-D memory pooling technology to let multiple systems share memory more flexibly during AI inference.

Samsung’s processing-in-memory technology performs certain calculations within memory chips to reduce data transfers and improve energy efficiency.

The company displayed Groq language processing unit wafers, Nvidia’s Groq 3 LPX system and 2-nanometer gate-all-around wafers.

Samsung’s MAX Lab enables customers to test memory and storage technologies under real-world AI workloads before commercialization.

AMD CEO Lisa Su discussed potential cooperation with Samsung in memory and foundry businesses during her visit to Korea on Oct. 7, while the companies signed a related memorandum of understanding in March.

Published Oct 13, 2026 11:06 am KST

Shim Ho-jun, vice president of the AE Team in the Memory Business at Samsung Electronics' Device Solutions division, delivers a presentation at the 2026 OCP Global Summit held in Silicon Valley, Calif., Monday (local time). Courtesy of Samsung Electronics

Samsung Electronics showcased next-generation memory and storage technologies at the 2026 OCP Global Summit in the United States on Monday (local time), highlighting its efforts to improve artificial intelligence (AI) infrastructure and expand cooperation with global technology companies.

The Korean tech giant jointly demonstrated a memory expansion solution with AMD and displayed products related to AI chipmakers Groq and Nvidia at the event, where Samsung participated as a Diamond Sponsor.

As AI models grow larger and AI agents handle increasingly complex tasks, demand is rising for memory capacity, faster data processing and more efficient storage. Samsung presented technologies designed to address these challenges across three areas: computing speed, memory expansion and data storage.

A key highlight was the joint demonstration with AMD of a Compute Express Link (CXL)-based memory expansion solution. CXL allows processors and other computing components to share memory resources, helping AI systems access additional capacity without relying solely on memory installed directly in individual servers.

Using Samsung's CMM-D-based memory pooling technology, the solution enables multiple systems to share memory more flexibly, helping them handle growing workloads during AI inference, when trained models generate responses.

Samsung also showcased high-performance memory products, including customized high-bandwidth memory (HBM) and HBM5, as well as processing-in-memory technology, which performs certain calculations within memory chips to reduce data transfers and improve energy efficiency.

The company displayed 2-nanometer gate-all-around wafers, Groq language processing unit wafers and Nvidia's Groq 3 LPX system, highlighting its capabilities spanning memory and advanced chip manufacturing.

For data management, Samsung presented enterprise solid-state drives and technologies designed to improve the storage and retrieval of information used during AI inference. These included key-value cache optimization, which helps AI models reuse previously processed information to generate responses more efficiently.

Samsung also emphasized collaboration with customers to optimize entire AI systems, rather than focusing solely on individual semiconductor products. Its Samsung MAX Lab allows customers to test memory and storage technologies under real-world AI workloads before commercialization.

The showcase follows Samsung's technology presentations at major industry events, including Nvidia's GTC in March, COMPUTEX in June and the Future of Memory and Storage conference in August.

Samsung is expanding its portfolio beyond HBM, a critical component in AI accelerators, into system memory, SSDs and NAND flash.

The company is also deepening ties with AMD. During AMD CEO Lisa Su's visit to Korea on Oct. 7, she discussed potential cooperation in memory and foundry businesses. The two companies signed a memorandum of understanding in March to expand collaboration in next-generation AI memory and computing technologies.

三星电子周一在美国展示了其新一代内存和存储技术，重点介绍了人工智能基础设施以及与AMD的合作。该公司与AMD联合演示了基于CXL的内存扩展解决方案，并展示了与Groq和Nvidia相关的产品。三星还展示了定制的HBM、HBM5、内存内处理技术、2纳米晶圆以及面向人工智能的存储解决方案。此次展示凸显了三星的产品线已从HBM扩展到系统内存、固态硬盘和NAND闪存领域。

基于 CXL 的解决方案采用三星的 CMM-D 内存池技术，使多个系统在 AI 推理期间能够更灵活地共享内存。

三星的内存内处理技术在内存芯片内执行某些计算，以减少数据传输并提高能源效率。

该公司展示了 Groq 语言处理单元晶圆、英伟达的 Groq 3 LPX 系统和 2 纳米全环栅晶圆。

三星的 MAX 实验室使客户能够在商业化之前，在真实的 AI 工作负载下测试内存和存储技术。

AMD 首席执行官苏姿丰 (Lisa Su) 于 10 月 7 日访问韩国期间，与三星探讨了在存储器和代工业务方面的潜在合作，而两家公司已于 3 月份签署了相关的谅解备忘录。

发布于2026年10月13日上午11:06（韩国标准时间）

三星电子器件解决方案事业部存储器业务AE团队副总裁沈浩俊（Shim Ho-jun）于周一（当地时间）在加州硅谷举行的2026 OCP全球峰会上发表演讲。图片由三星电子提供。

周一（当地时间），三星电子在美国举行的 2026 年 OCP 全球峰会上展示了下一代内存和存储技术，重点介绍了其在改善人工智能 (AI) 基础设施和扩大与全球科技公司合作方面所做的努力。

这家韩国科技巨头与 AMD 联合展示了内存扩展解决方案，并在活动中展示了与人工智能芯片制造商 Groq 和 Nvidia 相关的产品，三星作为钻石赞助商参与了此次活动。

随着人工智能模型规模的扩大和人工智能代理处理的任务日益复杂，对内存容量、更快的数据处理速度和更高效的存储的需求也随之增长。三星展示了旨在应对这些挑战的三大技术：计算速度、内存扩展和数据存储。

此次活动的一大亮点是与AMD联合演示了基于Compute Express Link (CXL)的内存扩展解决方案。CXL允许处理器和其他计算组件共享内存资源，帮助人工智能系统获取额外的容量，而无需完全依赖直接安装在各个服务器上的内存。

该解决方案利用三星基于 CMM-D 的内存池技术，使多个系统能够更灵活地共享内存，从而帮助它们在 AI 推理期间处理不断增长的工作负载，当训练好的模型生成响应时，这些系统就能更好地应对。

三星还展示了高性能内存产品，包括定制的高带宽内存 (HBM) 和 HBM5，以及内存内处理技术，该技术可在内存芯片内执行某些计算，以减少数据传输并提高能源效率。

该公司展示了 2 纳米全环栅晶圆、Groq 语言处理单元晶圆和英伟达的 Groq 3 LPX 系统，重点展示了其在存储器和先进芯片制造方面的能力。

在数据管理方面，三星展示了企业级固态硬盘以及旨在提升人工智能推理过程中信息存储和检索效率的技术。这些技术包括键值缓存优化，它可以帮助人工智能模型重用先前处理过的信息，从而更高效地生成响应。

三星还强调与客户合作，优化整个人工智能系统，而不仅仅是专注于单个半导体产品。其三星MAX实验室允许客户在商业化之前，在真实的人工智能工作负载下测试内存和存储技术。

此次展示是继三星在各大行业盛会上进行技术展示之后进行的，包括 3 月份的英伟达 GTC 大会、6 月份的台北国际电脑展 (COMPUTEX) 以及 8 月份的未来内存与存储大会。

三星正在将其产品组合从人工智能加速器中的关键组件 HBM 扩展到系统内存、固态硬盘和 NAND 闪存。

该公司也在深化与AMD的合作关系。10月7日，AMD首席执行官苏姿丰访问韩国期间，双方探讨了在存储器和代工业务方面的潜在合作。今年3月，两家公司签署了一份谅解备忘录，旨在扩大在下一代人工智能存储器和计算技术领域的合作。