Changlei Capital
常垒资本
Changlei Capital (常垒资本) is an early-stage Chinese venture capital firm that appears to focus on technology and AI infrastructure investments. It co-led the Pre-A round of Feijiekesi (飞捷科思), a physics AI engine startup founded by a former NVIDIA PhysX co-founder, alongside Yunqi Capital in May 2026 . The firm also invested in Tanxu Technology (探序科技), a Fudan University scientific research commercialization project , and participated in the August 2025 angel round of MosuMosu, an AI-powered personalized information platform .
Changlei's founding and managing partner Shi Mao (石矛) expressed a cautious stance on large language model investments in mid-2023, noting that the window for investing in foundational LLMs had closed and that the monetization path remained unclear . The firm also maintains a Middle East presence; its Middle East partner described Saudi Arabia as the primary market of interest, dismissing the UAE as merely a "springboard" comparable to Shanghai versus Guangdong Province in scale . Investment partner Peng Yiren (彭羿人) was quoted in late 2024 describing a personal investment in a project that Changlei's investment committee had rejected .
AI-generated — may contain errors, please verify.
Coverage
Yunqi Capital Co-Leads Round in Feijiekesi: NVIDIA PhysX Co-Founder Rebuilds Foundational Physics AI Engine | Yunqi Portfolio
From the Ground-Up Engine, Looking at Physical AI Opportunities
云启资本·Here are 14 startups that haven't been covered yet.
Always celebrate the risk-takers.
暗涌Waves·Where Are the Bold State Investors?
Breaking Down the Most Active Local State-Owned Capital.
暗涌Waves·2024, When Everyone Stands at a Crossroads: 96 Investors' Decisions
Deciding is easy. Waiting is hard.
暗涌Waves·Trekking to Riyadh
Opportunity, wealth, and a surging tide of people.
暗涌Waves·After 200+ Days of AIGC Fervor, Investors Reach Three Broad Consensuses
By Elune, edited by Jiumu, curated by Eason On July 19, Meta, Facebook's parent company, released the Llama 2 large language model, making it freely available for both research and commercial use — widely hailed as the most powerful open-source alternative to GPT-4. This marks a significant shift from the previous landscape, where numerous global large language models were built on Llama but constrained by its inability to be used commercially at no cost.
线性资本·
