On November 4, Jiemian News, China Securities Journal and other media reported from informed sources that HOMOmo AI, a compute-in-memory AI chip company, has recently closed a Series B round worth tens of billions of RMB. The round was jointly backed by upstream and downstream industrial capital, state-owned capital and international investors, with multiple existing shareholders adding further commitments—a signal of market recognition for the compute-in-memory technical route and its commercial prospects.
This is the latest large round in the on-device AI funding wave. According to incomplete statistics, nearly ten on-device AI chip companies in China announced new financing rounds in September alone—investor attention is shifting from cloud compute clusters to the chips inside phones, computers and robots.
Technology and product details
HOMOmo focuses on an SRAM-CIM (compute-in-memory on SRAM) architecture. Its edge-and-device LLM chip M50 has been in production for less than a year, yet it has already locked in more than 150 design points and served over 100 customers. By merging storage and compute into the same unit, the compute-in-memory architecture sharply reduces the memory-access and power bottlenecks of on-device LLM inference, bringing large models to real scenarios such as phones, computers and intelligent terminals.
According to Jiemian, the round will mainly be used to scale M50 volume delivery, advance the next-generation M6X and 3D CIM products, and continue to strengthen the software ecosystem—moving from a single chip to a combined capability of chip, toolchain and ecosystem.
From an industry perspective…
On-device AI is transitioning from storytelling to large-scale realization. As large models shift from the cloud to terminals, unit compute cost, energy efficiency and software ecosystems are becoming the focus of competition. New architectures such as compute-in-memory are running in parallel with general-purpose NPUs and co-processors, meaning the on-device compute landscape is becoming far more diverse; whoever turns compute into deployable products fastest will lead the race.
Junyi Microelectronics business relevance
Junyi Microelectronics builds on mainstream platforms such as RK3588, RK3576 and RK3568, offering J-3568/J-3576/J-3588 core boards, K-series edge intelligent motherboards, and finished products including the 5G-EDGE-3399Pro, AI-BOX-3588 edge computing host, IPC-i5-1235U industrial mini PC and touch panel PCs—key carriers for on-device AI compute in industrial sites. As on-device LLMs and new compute chips mature, edge hosts can carry higher local inference loads and run AI directly on production lines, inspection and terminal devices.
Source: Jiefang Daily (2026-11-04)