K

Kunlunxin

Semiconductors · Yanqing District, China · Founded 2011

Kunlun Core Beijing Technology, operating under the brand name Kunlunxin, is a developer of artificial intelligence processors spun out of Baidu. The company originated as the tech giant's Intelligent Chip and Architecture Department, which began its foray into AI acceleration back in 2011. In 2021, the department completed its independent financing to officially establish Kunlunxin, with Baidu retaining a majority stake and Ouyang Jian, Baidu's former chief chip architect, taking the helm as CEO. Kunlunxin focuses on the design and manufacturing of high-performance, general-purpose AI chips, addressing the growing demand for computing power in the AI industry. The firm's core business revolves around providing these semiconductor products to a range of enterprise clients in sectors such as internet services, intelligent industry, smart transportation, and finance. Its revenue is generated from the sale of its AI chips and associated hardware, like accelerator cards. The company serves as a critical domestic alternative in China's technology supply chain, underscored by a significant order from telecom giant China Mobile for its AI projects. The company's product portfolio includes multiple generations of its 'Kunlun Core' AI processors. The first-generation chip, Kunlun 1, was built on a 14nm process and saw mass production in 2020. Its successor, Kunlun Core 2, utilizes a 7nm process and a self-developed second-generation XPU architecture, reportedly boosting performance by two to three times over its predecessor. These chips are engineered to handle both AI training and inference tasks and are deployed in various form factors, including accelerator cards (K-series and R-series) and server configurations. A key selling point is the software's compatibility with NVIDIA's CUDA platform, which simplifies the transition for developers. Kunlunxin's technology has been deployed in Baidu's own services, such as its search engine and the ERNIE Bot, as well as in industrial smart factories and the Baidu Apollo autonomous driving platform. Keywords: AI chips, semiconductor design, artificial intelligence processors, XPU architecture, cloud computing, edge computing, AI inference, AI training, Baidu, Ouyang Jian, general-purpose GPU, machine learning hardware, data center acceleration, autonomous driving chips, CUDA compatible, domestic AI technology, Kunlun Core, intelligent computing, smart transportation, intelligent finance