MemSilicon is tackling the AI "memory wall," where moving data between memory and compute can lag processing, with RRAM and processing-in-memory architectures designed to combine storage and computation while reducing power and increasing density. It reports embedded-IP supply contracts worth several million yuan and strategic and long-term standalone-RRAM supply agreements, while its latest more-than-RMB-400-million Series A is earmarked for RRAM industrialization and a first-generation LPU for large-model inference.