Panmnesia’s CXL-Based GPU Memory Expansion Kit eliminates the GPU memory capacity limitations with innovative Compute eXpress Link (CXL) technology for large AI models such as LLMs and GenAI. The CXL empowers the GPU to enjoy terabytes of memory space by connecting it to multiple memory expanders through the CXL network. Such ample memory resources allow AI service providers to run large AI models with 34 times less system CAPEX. In addition, our kit employs Panmnesia’s optimized CXL controller that exhibits the world’s first two-digit nanosecond latency, offering high speed for memory expansion.