Upcoming Event: Storage and Memory Summit 2025
Future of Memory and Storage Conference Highlights Innovations for AI-Driven Computing
The Future of Memory and Storage (FMS) conference, held from August 5-7, showcased the latest advancements in memory and storage technology. This year's event featured a focus on Compute Express Link (CXL), chiplets, and advanced chip packaging as critical enablers for scaling AI workloads and next-generation computing.
Compute Express Link (CXL) emerged as a foundational technology, enabling scalable, coherent memory architectures essential for AI and HPC systems. CXL provides an open, high-bandwidth, low-latency interconnect between CPUs, accelerators, memory expansion modules, and storage, allowing efficient offloading of memory management tasks and improving data delivery to GPUs and accelerators.
One of the innovative approaches using CXL for memory expansion can alleviate DRAM bottlenecks in managing petabyte-scale SSD metadata by offloading index management to external modules, thus enabling much larger capacity SSDs by 2030 with improved efficiency. The CXL Consortium continues to drive such innovations and ecosystem partnerships, including integration with optical interconnects in advanced chip packaging.
Chiplets and advanced chip packaging technologies are pivotal for integrating multiple dies to boost system performance, flexibility, and yield. The packaging innovations range from 2.5D and 3D integration to interposers and optical interconnect bridges. These packaging methods support dense stacking and heterogeneous integration of compute, memory, and I/O components. Celestial AI, showcased at FMS, demonstrated their Photonic Fabric™ technology, which incorporates optical chiplets, interposers, and multi-chip interconnect bridges that provide up to 10x performance and energy efficiency improvements over electrical interconnects.
The broader packaging landscape is evolving as semiconductor foundries and OSATs (Outsourced Semiconductor Assembly and Test) race to deliver more complex, multi-chip solutions that integrate vertical memory architectures, chiplet-based designs, and new transistor technologies. Packaging is now recognized as a critical design component enabling higher performance and better thermal management for AI and high-speed memory/storage solutions.
Additional highlight winners and innovations at FMS 2025 included breakthroughs in high-capacity storage solutions, such as petabyte-capacity SSDs using hardware-accelerated compression combined with CXL-enabled memory expansion, and AI-driven storage platforms optimizing data throughput and efficiency.
Other notable demonstrations at the conference included a Cold-Plate SSD designed for liquid-cooled servers, a Chiplet System Architecture being armed, a Storage Accelerator speeding up NVMe arrays, and high-capacity solid-state disks (SSDs) being showcased. The conference also had some preshow announcements, such as the unveiling of UCIe, an open chiplet ecosystem at the package level.
In summary, the key advances at FMS 2025 in memory and storage technology converge on enabling scalable, efficient AI and HPC systems through CXL, chiplet and advanced packaging technologies, and innovative storage architectures. These integrated innovations position the memory and storage ecosystem as a vital foundation for future AI-driven computing.
Embargoed details will be posted when the show starts, and videos shot at the show will be posted later. Stay tuned for more updates from FMS 2025!
[1] CXL Consortium website [2] Semiconductor Engineering article [3] AnandTech article [4] Celestial AI press release [5] Data Center Knowledge article
- The Future of Memory and Storage (FMS) conference, with a focus on Compute Express Link (CXL), chiplets, and advanced chip packaging, underscores the significant role of data-and-cloud-computing technology in promoting AI-driven computing by enabling scalable, efficient systems.
- These advancements in memory and storage technology, such as CXL, chiplets, and packaging, highlight the crucial impact of technology in overcoming bottlenecks in managing massive datasets and improving AI performance, as showcased at FMS 2025.
