PIM Chip

About PIM Chip

Founded in February 2021, PIM Chip is a semiconductor company headquartered in Beijing that concentrates on developing AI accelerator chips using processing-in-memory (PIM) technology. The company was established to address the performance bottlenecks of the traditional von Neumann architecture in AI applications, where the separation of processing and memory units leads to significant energy consumption and data transfer delays. By integrating computation directly within memory, PIM Chip aims to create high-performance, low-power AI computing engines.

The founding team combines deep academic knowledge with extensive industry experience. The initiative was led by Dr. Chen Yiran, a professor at Duke University and a renowned expert in new memory and AI chips, who serves as the chairman. The co-founding team includes CEO Dr. Yang Yue, a Tsinghua University alumnus with a Ph.D. from the University of Toronto, who previously worked as a principal systems architect at Micron Technology. Co-founder Dr. Zhang Yaojun, the first doctoral graduate of Prof. Chen, specializes in MRAM commercial chip development. The team is rounded out by co-founder Xu Zhenlong, who brings over two decades of chip design experience from his time at TSMC.

PIM Chip’s business model revolves around designing and providing hardware units and related IC design services that boost the performance of deep learning computations. The company targets the edge computing market, with products designed for applications in smart wearables, drones, security, robotics, and smart homes, all of which require low power consumption and long standby times. Its revenue is generated through the sale of these specialized chips and providing tailored AI hardware acceleration solutions to clients, which include partnerships with major manufacturers in the smart wearables and image recognition sectors.

The company's product portfolio is built on its PIM technology. One of its key products is the PIMCHIP-S300, a multi-modal intelligent perception and decision-making AI chip that leverages an SRAM-based in-memory computing acceleration unit. This design significantly enhances energy efficiency by minimizing data movement. Another offering is the PIMCHIP-N300, a neural network processing unit designed for machine learning and AI, supporting mixed-precision computing to balance power consumption and performance. These chips are developed using 28nm and 22nm process nodes, a strategic choice that balances performance with manufacturing costs, ensuring a stable supply chain. The company provides a complete development toolkit, enabling clients to shorten their product development cycles.