China's quant funds are turning their trading compute into AI labs
Shanghai Securities News names four quant managers with their own institutes, supercomputing centers or open source models, and asks which will produce the next DeepSeek.
Shanghai Securities News (上海证券报) reported on 23 March 2026, under reporter Ma Jiayue (马嘉悦), on quant private funds (量化私募) moving from trading into AI research. The headline translates as “The quant ‘technology corps’: who is preparing the next DeepSeek?” The article is in Chinese, and the translation is ours.
| Firm | What the report says |
|---|---|
| Jiukun Investment (九坤投资) | Set up the Zhizhi Innovation Research Institute (至知创新研究院), separate from investment research. Open-sourced a code model that writes, debugs and explains code. A paper was a NeurIPS 2025 Spotlight |
| Mingshi Fund (鸣石基金) | Founded its G-Lab AI lab in 2021 and its own supercomputing center in 2022, now through three phases |
| Huanfang Quant (幻方量化) | Built a supercomputing room in 2019 about the size of a basketball court |
| Lingjun Investment (灵均投资) | An affiliate launched an AI health product in January 2026 that reportedly drew nearly 10,000 users in three months |
Yuan Yu (袁宇) of Mingshi is quoted: the creator of the next DeepSeek “need not be a quant fund boss. It can be any top talent.” An unnamed veteran founder says that “replicating DeepSeek is hard” (复刻DeepSeek很难).
The article does not describe how these firms use agents in factor research, and it gives no GPU counts, staff totals or benchmark results. Chinese brokers deployed DeepSeek-V4 within hours of its release a month later.
Sources
- Shanghai Securities News paper.cnstock.com