Kunlunxin (Beijing) Technology Co., Ltd.
Category: AI Chips / Semiconductors
A leading AI semiconductor company that designs and produces high-performance general-purpose AI processors based on its proprietary XPU architecture for cloud-to-edge computing. Kunlunxin (Beijing) Technology Co., Ltd. was founded in 2021. The company is led by Ouyang Jian. Based in Beijing, China. Team size: 500+. Total funding raised: $650.0M. Latest round: Series D ($283.0M, Jul 2025). Key investors include ["Baidu","CPE Yuanfeng","IDG Capital","Legend Capital","GGV Capital","BYD","China Mobile (Hecheng Fund)"].
- Founded
- 2021
- Headquarters
- Beijing, China
- Team size
- 500+
- Total funding
- $650.0M
Value proposition
Delivers a high-efficiency, cost-effective alternative to global GPUs with a software-hardware co-optimized stack that enables seamless workload migration with minimal code changes.
Products and solutions
["Kunlunxin Generation 1 AI Chip (XPU-K, 14nm)","Kunlunxin Generation 2 AI Chip (XPU-R, 7nm)","Kunlunxin Generation 3 / P800 Series (XPU-P, for large-scale model training)","K-Series & R-Series AI Acceleration Cards (K100, K200, R200-8F)","R480-X8 AI Accelerator Cluster","Kunlunxin SDK (Software Development Kit for framework compatibility)"]
Unique value
Developed the 100% proprietary 'XPU' architecture, which is specifically engineered to balance the flexibility of general-purpose GPUs with the high efficiency of specialized AI accelerators (NPUs).
Target customer
Hyperscale internet companies, telecommunications operators, financial institutions, autonomous vehicle manufacturers, and government-led smart city projects.
Industries served
["Internet & Cloud Computing","Financial Services (Banking/Fintech)","Smart Manufacturing & Industrial AI","Autonomous Driving & Intelligent Transportation","Telecommunications","Smart City & Public Infrastructure"]
Technology advantage
Maintains a mature software-hardware co-design ecosystem; features 'one-line code' migration from NVIDIA ecosystems and is the only domestic chip proven by massive-scale deployment within Baidu’s search engine and 'Wenxin' LLM workloads.
How they differentiate
Kunlunxin differentiates through its 100% proprietary 'XPU' architecture, which balances the high efficiency of specialized AI accelerators (NPUs) with the flexibility of general-purpose GPUs. It offers a mature software-hardware co-design ecosystem that enables 'one-line code' migration from NVIDIA's CUDA environment and deep integration with Baidu's PaddlePaddle AI framework.
Main competitors
["Huawei HiSilicon (Ascend)","Cambricon (寒武纪)","Biren Technology (壁仞科技)","Enflame (燧原科技)"]
Key partnerships
["Baidu (Parent and lead ecosystem partner)","China Mobile (Strategic partner in 'Xinhe' Open Lab for intelligent computing)","BYD (Investor and partner for automotive/industrial applications)","PaddlePaddle (Deep framework-level integration)","Key enterprise clients: China Merchants Bank (CMBC), Southern Power Grid, and Geely Auto."]
Notable customers
["Baidu","China Mobile","China Merchants Bank","Southern Power Grid","Geely Auto","BYD"]
Major milestones
["2011: Project inception within Baidu's Intelligent Chip and Architecture Department.","2018: Launch of the first-generation Kunlun 1 AI chip.","2021-04: Independent spin-off from Baidu with a Series A valuation of 13 billion RMB.","2024: Secured a landmark 1 billion RMB contract with China Mobile.","2025-07: Completed Series D funding, raising 2.1 billion RMB at a 21 billion RMB valuation.","2026-01: Formally filed for a Hong Kong IPO."]
Growth metrics
Valued at approximately 21 billion RMB (~$3 billion) as of early 2026; 2025 annual revenue surpassed 2 billion RMB with forecasts predicting a 6x surge in revenue following its 2026 IPO filing.
Market positioning
A leading domestic challenger in China's high-performance AI chip market, positioned as a key alternative to global GPUs for cloud-to-edge computing, especially in large-scale model training (LLMs) and inference.
Geographic focus
Primarily focused on the Greater China market, driven by the national push for semiconductor self-reliance and the replacement of restricted foreign high-end AI silicon.
Patents and IP
Holds a leading position in Chinese AI chip patents (leveraging Baidu's top-ranked IP portfolio); core XPU architecture innovations have been published in top-tier global forums including Hot Chips and ISSCC.
About Ouyang Jian
Ouyang Jian is a prominent semiconductor architect who served as the Chief Chip Architect at Baidu for over a decade. He joined Baidu in 2009 and initiated the Kunlun AI chip project in 2011. Under his leadership, the team developed several generations of AI processors, including the Kunlun 1 and Kunlun 2, transitioning the project from an internal department to a multi-billion dollar independent entity.
Official website: https://www.kunlunxin.com/