Tsing Micro, a semiconductor company developing reconfigurable chips, has completed the IPO counseling process and entered the counseling acceptance stage. Filings reviewed by IPO Zaozhidao show the company plans to list on the ChiNext board of the Shenzhen Stock Exchange.
The move brings Tsing Micro closer to formally submitting its prospectus and could make it China’s first listed company focused on a non-GPU chip architecture, according to IPO Zaozhidao.
Founded in 2018, Tsing Micro focuses on R&D and commercial applications of domestically developed reconfigurable chips, known as reconfigurable processing units, or RPUs. These chips are designed to provide computing power for artificial intelligence computing centers, large models, autonomous driving, and smart manufacturing. The company aims to build a domestic ecosystem for general-purpose computing based on RPUs and expand the use of reconfigurable chips across device, edge, and cloud computing nodes.
IPO Zaozhidao describes Tsing Micro as the developer of the world’s first commercial reconfigurable chip and as the commercial reconfigurable computing chip company with the largest shipment volume globally. Commercial reconfigurable chips follow a different technology path from traditional GPUs and can be used for AI computing and deep learning model acceleration. Reconfigurable AI chips are designed to retain the flexibility of GPUs while using dynamic operator reconfiguration to approach the energy efficiency advantages of dedicated AI chips, such as tensor processing units (TPUs). For this reason, some in the industry describe them as “general-purpose TPUs.”
Tsing Micro has begun mass-producing its TX8 series, a cloud computing chip for computing centers and other cloud environments, and its TX5 series, an energy-efficient chip for computer vision applications in edge computing. Together, the products are intended to support device, edge, and cloud computing scenarios.
The TX8 series is built on Tsing Micro’s self-developed reconfigurable dataflow architecture and targets the cloud market. Its TX81 AI compute chip uses chip-to-chip mesh technology, which allows data to move directly among chips, servers, and racks without first passing through network switches. The company said this design addresses three common bottlenecks in AI computing: computational efficiency, interconnect bandwidth, and memory access.
Tsing Micro’s cumulative reconfigurable chip shipments have reportedly exceeded 30 million units, and its chips have been deployed at scale in more than a dozen computing centers in China with at least 1,000 cards each.
At the software level, Tsing Micro is one of the few companies that has achieved full-stack compatibility with all core components of FlagOS, according to the report. Its adaptation scale ranks among the top two non-GPU architectures, alongside Huawei Ascend. That compatibility means AI applications developed on FlagOS can run on Tsing Micro chips. Last April, ten Chinese AI chip companies, including Tsing Micro, achieved same-day adaptation of the DeepSeek V4 model based on FlagOS and released model inference images through FlagRelease.
Since its founding, Tsing Micro has attracted investors including national funds, state-owned capital, industrial investors, and market-oriented investment firms. Together, these backers give the company a shareholder base spanning policy support, industrial coordination, capital resources, and ecosystem links. The investor base could support Tsing Micro as it expands R&D, deployment, and ecosystem development for AI chips built on emerging architectures.
This article was adapted based on a feature originally written by Stone Jin and published on IPO Zaozhidao. KrASIA is authorized to translate, adapt, and publish its contents.