Core Viewpoint - Shanghai Fangqing Technology has become a new focus in the domestic AI computing track after securing significant funding, emphasizing its innovative "decoupled AI computing architecture" [1][2]. Funding and Investment - The company completed a Pre-A round financing exceeding 500 million yuan in December 2025, followed by a 1 billion yuan Pre-A3 round in March 2026, with investments from eight institutions including NIO Capital and Guokai Science and Technology Innovation [1][2]. - The rapid capital influx reflects strong confidence in the company's core technology direction [2]. Technology Innovation - Fangqing Technology addresses the efficiency challenges of large model inference by decoupling the "Attention mechanism" and "Feedforward Neural Network (FNN)" into separate modules, allowing for optimal hardware allocation [2][4]. - This innovative approach is expected to improve computing resource utilization by over 50% and reduce power consumption by 25%-30%, effectively tackling the "efficiency anxiety" in large model inference [4]. Leadership and Expertise - The company is led by Liang Jun, a veteran with over 20 years of chip development experience, known for his role in developing the Kirin SOC chip at Huawei and the first 7nm AI training chip at Cambricon [6]. - Liang's leadership is seen as a significant asset, providing a clear path for technology implementation from architectural innovation to commercial deployment [6]. Future Outlook - Fangqing Technology aims to continue focusing on technology as its core, seeking to transform existing AI hardware design logic and contribute to the diversified development of the domestic AI computing ecosystem [6].