We develop general-purpose computing on graphics processing units (“GPGPU”) chips and GPGPU-based intelligent computing solutions to provide the foundational computing power required by artificial intelligence (“AI”). By integrating self-developed GPGPU-based hardware and proprietary BIRENSUPA software platform, our solutions support the training and inferencing of AI models in a broad range of applications from cloud to edge. In particular, strong performance and high efficiency for large language models (“LLMs”) pre-training, post-training and inference of our GPGPU-based solutions, which possess high technology barriers, provide us with key competitive advantages among domestic players. Our technology forms a critical infrastructure to enable AI and advance artificial general intelligence (“AGI”), addressing the surging computational demands across various industries to drive productivity, innovation and transformation.
Innovation and technology excellence are our core competencies. With the rapid development of AI, especially through LLMs and generative AI, many businesses have an increasing need for computing solutions to meet their surging demand for computing power and harness the power of AI. To meet such demand, we have self-developed our Specialist Technology Product which is an integrated intelligent computing solution comprising of two components, namely (i) hardware systems based on our GPGPU architecture and chips, and (ii) BIRENSUPA, a computing software platform. The development of our solution requires careful planning and coordination across various stages, we are primarily engaged in marketing requirements analysis, defining chip specifications, architecture design, chip design, hardware system integration and testing, in-house development of our software stack, sales and marketing, and technical support. To facilitate the chip design process, we utilize various items, tools and support services provided from third-party suppliers and may choose to outsource certain backend and physical design to design services vendors. We operate under a fabless model, and we engage third-party contract manufactures for manufacturing, assembly, testing and packaging of our semiconductor wafers and final products. To better address our customers’ urgent demands for high-performance computing and intelligent applications, our Specialist Technology Product can be offered as large-scale intelligent computing clusters, which consist of a large number of interconnected GPGPU units and that work together to perform parallel processing tasks and controlled by our BIRENSUPA software platform.
We have built our solutions upon five foundational pillars: a self-developed GPGPU architecture, system-on-chip (“SoC”) design, hardware system, software platform, and cluster deployment optimization at large scale. Specifically:
‧ Our technology capabilities and solution excellence are underpinned by our in-house developed GPGPU architecture, which is purpose-built for handling large-scale AI workloads, especially LLM workloads, to accommodate expanding model sizes, parameters and complexities, while offering high-performance and superior general- purpose flexibility, energy-efficiency and scalability. The unified and continuously evolving GPGPU architecture is the core of our platform strategy and lays a solid foundation for fast iteration and development of next generation computing platform.
‧ Based on the self-developed GPGPU architecture, we design and launch a series of chips. According to CIC, we are the first company in China to package dual AI computing dies using 2.5D chiplet technology, supported by our superior SoC design and execution capabilities. We are among the first in the industry to support advanced interconnection specifications, according to CIC. Our SoC design methodology and workflow ensure successful VLSI (Very Large Scale Integrated Circuit) execution and first-time-right tape-out, which help us achieve mass production and commercialization with our first generation products.
‧ We have developed a comprehensive portfolio of high-performance hardware systems in various form factors containing our self-developed GPGPUs chips, such as PCIe Card, open accelerator module (“OAM”), and servers. Our hardware systems support both air-cooled and liquid-cooled solutions, helping to reduce the power usage effectiveness (“PUE”) and optimize energy efficiency of data centers and comply with applicable energy-saving requirements. We provide enterprises the mission-critical large-scale computing infrastructure that offers high performance, reliability, and scalability.
‧ We have developed the BIRENSUPA software platform, bridging all of our hardware systems with diverse AI applications and scenarios. BIRENSUPA enables our hardware features, optimizes their performance and manages large-scale GPGPU clusters. It offers user a friendly programming interface, high performance libraries, training and inference frameworks and comprehensive set of tool chains to streamline the development and deployment of AI solutions. Furthermore, BIRENSUPA is compatible with other third-party GPGPU computing software platforms, significantly reducing the migration cost to our GPGPU products.
‧ We have developed comprehensive solutions of large-scale intelligent computing cluster by integrating our hardware systems and software platform with other hardware infrastructure such as servers, storage, and networking equipment provided by partners. Our cluster management platform, BIRENCUBE, is designed to manage extensive AI hardware infrastructure, allowing us to help customers construct GPU clusters comprising of over one thousand, or even ten thousand, GPU chips.
As AI adoption continues to expand, a growing number of companies across diverse industries are creating innovative AI-enabled products and services, significantly increasing the demand for computing power. Key sectors, including AI data centers, AI solutions and Internet, are at the forefront of the race, significantly increasing their investment in computing power and related infrastructure. Moreover, leading companies within these industries account for the majority of capital expenditures on computing power. Hence, we implement the strategy that targets key industries with high demand for computing power, and form strategic partnerships with large customers in each industry. These selected key industries include AI data centers, telecommunications, AI solutions, energy and utilities, financial technology, and Internet. With localized expertise and on-the-ground customer support, our solutions are designed to address unique needs of these customers.
‧ We started to generate revenue from our intelligent computing solutions in 2023.
During the year ended December 31, 2024 and for the six months ended June 30, 2025, we had 14 and 12 customers for our Specialist Technology Product, respectively, contributing a revenue amounted to RMB336.8 million and RMB58.9 million, respectively.
‧ As of the Latest Practicable Date, we had 24 unfulfilled binding orders for our Specialist Technology Product with a total value of approximately RMB821.8 million.
‧ In addition, as of the Latest Practicable Date, we have entered into five framework sales agreements and 24 sales contracts for our Specialist Technology Product with a total value of approximately RMB1,240.7 million, which will contribute to our future revenue when realized.
Source: Biren Technology (06082) Prospectus (IPO Date : 2025/12/22) |