Longsys Shifts Storage Strategy to Meet Edge AI Demands

The rise of edge AI has created what Dr. Chen describes as a "thousand-handed challenge," where diverse AI models and hardware architectures render traditional, one-size-fits-all storage solutions obsolete. To solve this, Longsys is integrating hardware design, firmware development, and OS-level optimization to create custom storage pathways for PCs, robotics, and industrial systems. By applying full-stack engineering, the company aims to move beyond simple product manufacturing into a model that aligns storage performance with specific CPUs and SoCs.

Central to this strategy are three new technical innovations. HLCache™ improves multitasking on mobile devices by managing data between UFS and DRAM, while the iSA™ and AISSD™ combination allows AI PCs to offload model parameters to SSDs, effectively reducing DRAM reliance. For high-performance edge systems, the new AIDIMM™ architecture provides a 256-bit interface capable of reaching 307.2GB/s bandwidth. These tools are designed to lower the barrier for developers struggling with memory bottlenecks and high costs in edge environments.

Under the banner of "Collabovate," Longsys is now opening its technology stack to partners across the semiconductor and AI industries. By treating storage as a foundry service rather than a commodity, the company seeks to accelerate the deployment of complex AI applications. This shift marks a strategic pivot for the firm, which has expanded its global manufacturing footprint through acquisitions like Lexar and Zilia to support the specialized demands of the current AI-driven hardware landscape.