Yizhu Tech / 亿铸科技

About Yizhu Tech / 亿铸科技

Yizhu Tech (亿铸科技, Suzhou Yizhu Intelligent Technology) is a Suzhou-based semiconductor company that designs and mass-produces large-compute AI chips built on a fully digital compute-in-memory (processing-in-memory) architecture using emerging non-volatile memory such as memristors. The architecture is intended to solve the memory wall and energy wall that limit conventional AI chips, and the company's stated vision is to lead the global compute-in-memory chip field and build a new ecosystem around the technology.

Its chips are aimed at data-center, cloud computing and central server AI workloads, with the goal of delivering several times better price-performance, higher energy efficiency and more headroom for compute growth than conventional AI accelerators. The company has demonstrated a memristor-based, high-precision, low-power compute-in-memory proof-of-concept chip made on a conventional process node, and reports third-party-verified energy efficiency more than ten times above the average of traditional-architecture AI chips. It has also proposed a compute-in-memory super-heterogeneous architecture as a development path for domestic AI compute chips and has presented its approach at industry events such as WAIC and provincial AI application conferences.

The team covers the full chain of process devices, architecture design, circuit design and software ecosystem, and its core engineering members have on average more than 25 years of high-end integrated circuit design and volume production experience. Its position is as one of the few Chinese companies able to independently design and mass-produce this class of compute-in-memory chip, backed by well-known domestic semiconductor investors.

Yizhu sells AI accelerator chips and related solutions to cloud, server and data-center customers in China, competing in the domestic AI compute chip market against GPU and ASIC suppliers on cost per unit of compute and power consumption.