NOVASILICON: Raises Nearly RMB 100M in Pre-A+ Led by Shunwei

On 17 November 2026, NovaSilicon raised nearly RMB 100 million in a Pre-A+ round led by Shunwei Capital, joined by GGV Capital and Matrix Partners China.

NewTimeSpace News: On 17 November 2026, NovaSilicon (Shanghai Xinxingyuan Intelligent Technology Co., Ltd., unlisted), an AI-native chip design platform, completed a pre-A+ round of nearly RMB 100 million led by Shunwei Capital, with GGV Capital and Matrix Partners China following and LightSource Capital acting as sole financial adviser. The exact amount and post-money valuation were not disclosed. It is the third round since the company was founded in June 2026; earlier backers include 5Y Capital, SenseTime Guoxiang Capital, Shanghai Future Industry Fund and Shanghai Science and Technology Innovation Group. Proceeds will fund development of its Circuit Design Kit (CDK), tape-out and delivery of the first IP, and team expansion.

CDK packages chip design experience and agents into one system that handles the iterative work between a chip specification and the GDS layout. The company compares it with the Process Design Kit that a foundry supplies to clients, and calls the resulting division of labour Designless, where chip companies define specifications while implementation is handled by CDK and the engineers using it. CDK has an analogue line called Vega and a digital line called Sirius.

Two recent papers from the company report that, on 40 analogue layout tasks, Novara within Vega scored 0.8154 on average with a 72.5% acceptance rate, the highest among systems tested, against 0.7483 and 70.8% for GPT-6 Astra. On an industrial design with 99 hard macros, Nova-SI within Sirius averaged 57.97 on floorplanning, about 1.9 times the 31.17 scored by GPT-6 Astra. The company notes these are benchmark results from the papers, not verification through tape-out.

Commercially, the company uses CDK to deliver IP and chip design services, starting with UCIe PHY IP and 3D compute-in-memory design services; Sirius is already used in the latter and has generated revenue. Systems are deployed locally so customer design data does not pass through third-party model services. The company says Vega has cut PLL design cycles from months to weeks. Founder and CEO Li Linyang studied under Professor Qiu Xipeng of Fudan University and Professor Lin Dahua of the Chinese University of Hong Kong; co-founder Xue Jianxi leads the digital side and co-founder Chen Jianqiu the analogue side. Shunwei Capital said the company reworks the underlying logic of chip design with reinforcement learning rather than following the conventional route of adding AI to EDA tools.

×
Share to WeChat

Open WeChat, use the "Scan", and share to my Moments.