Samsung Showcases Next-Generation AI Semiconductor Innovations at COMPUTEX 2026

Held in Taipei, Taiwan, from June 2–5, COMPUTEX 2026 once again reinforced its position as Asia’s premier ICT exhibition, bringing together more than 1,400 exhibitors and approximately 110,000 visitors from around the world. Under the theme “AI Together,” this year’s event highlighted the growing interconnectedness of AI technologies and industry ecosystems. With NVIDIA GTC Taipei taking place alongside COMPUTEX, global attention was focused on the rapid evolution of AI infrastructure and semiconductor technologies.

Integrated memory and storage solutions for AI systems

Under the theme “Integrated AI Semiconductor Solutions,” Samsung Electronics showcased a comprehensive portfolio spanning HBM, system memory, and storage technologies designed to support the entire AI system. At the center of the booth, Samsung featured HBM4E chips and wafers alongside mock-up and samples demonstrating HPB (Heat Path Block), a next-generation thermal architecture technology for future HBM products. A wall exhibit based on a next-generation AI system and memory hierarchy architecture illustrated how Samsung’s memory and storage solutions support the entire system stack. HBM4 was positioned in the GPU memory layer, while SOCAMM2, an LPDDR5X-based server module, represented the system memory layer. The storage layer featured Samsung’s AI-optimized SSD portfolio, including PM1763, PM1753, and PM9D3a. In particular, PM1763, Samsung’s next-generation PCIe Gen6 enterprise SSD, is designed to deliver the performance and power efficiency required for AI server environments. Together, these technologies demonstrated Samsung’s capability to deliver integrated memory and storage solutions across the entire AI memory hierarchy, from HBM and system memory to storage.

Next-generation semiconductor innovation for AI system competitiveness

On June 2, the opening day of COMPUTEX 2026, Samsung Electronics CTO Jaihyuk Song met with members of the media to discuss how the semiconductor industry is evolving in the AI era. During the briefing, he emphasized that semiconductor innovation is evolving beyond memory performance alone to encompass system-level optimization across packaging, storage, thermal management, and AI system architecture. He also highlighted Samsung’s unique position as an integrated device manufacturer (IDM), with capabilities spanning memory, foundry, logic, and advanced packaging, allowing the company to drive innovation across the entire AI system stack.

Samsung also unveiled HBM4E, its next-generation HBM solution designed for future AI systems. Built with Samsung’s most advanced 1c DRAM-based Core Die and a Base Die fabricated on Samsung Foundry’s 4nm process technology, HBM4E supports speeds of up to 14Gbps per pin, with scalability to 16Gbps and bandwidth exceeding 4TB/s in the future. The solution also delivers more than 30% higher capacity than the previous generation, addressing the growing demands of large-scale AI models and data-intensive workloads.

The D2D PHY (Die-to-Die Physical Layer), which enables high-speed communication between HBM and GPUs, is one of the primary heat-generating areas within the Base Die. As data throughput and operating speeds continue to increase, heat generation in this region becomes a growing challenge. To address this, Samsung introduced HPB (Heat Path Block), a technology that incorporates a dedicated thermal path to more efficiently transfer and dissipate heat generated in the D2D PHY. By reducing thermal resistance and improving operational stability, HPB enables reliable performance even in high-bandwidth, high-density environments. Samsung is currently validating HPB technology on HBM4E and plans to adopt it in future HBM5 products. Through this unveiling, Samsung highlighted its vision for next-generation HBM technologies and reinforced its leadership in AI memory innovation.