Mingshi Capital's Founder on Finding AI's 'Goosebump' Founders

#vc-strategy #ai-agents #founder-mindsetmingshi-capital-ai-investment-strategyminimax ideal-auto genspark lovart openclaw

💡Learn how top VCs identify AI winners and why they believe 'AI-native' founders must be willing to bet on non-consensus.

⚡ 30-Second TL;DR

What Changed

Investment success in AI requires identifying founders who are 'obsessively rational' and willing to bet everything on non-consensus bets.

Why It Matters

The article provides a rare look into the decision-making process of a top-tier VC, suggesting that the next wave of AI winners will be those who move beyond linear extrapolation and embrace the rapid, often chaotic, iteration cycles of AI agents.

What To Do Next

Track your daily token consumption for complex AI tasks; if you aren't spending significant resources on agent execution, you lack the necessary context to build for the next wave.

Who should care:Founders & Product Leaders

Key Points

•Investment success in AI requires identifying founders who are 'obsessively rational' and willing to bet everything on non-consensus bets.
•AI investment strategy must involve staying close to top-tier model developers to anticipate capability unlocks.
•Founders should prioritize deep, hands-on experience with AI agents, specifically tracking high-token-consumption user behaviors.
•Age is irrelevant; 'AI-native' thinking and the ability to iterate rapidly are the primary indicators of a successful founder.

🧠 Deep Insight

Web-grounded analysis with 14 cited sources.

🔑 Enhanced Key Takeaways

•Mingshi Capital, established in 2010, operates as a quantitative hedge fund primarily utilizing AI for its investment strategies within China's A-share market, focusing on generating alpha and employing intraday high-frequency trading.
•Huang Mingming is characterized as an 'optimist' regarding the future of AI agents, predicting that China's extensive mobile internet product development capabilities will lead to two-thirds of the world's leading AI agents originating from the country.
•Mingshi Capital is actively expanding its internal AI capabilities by recruiting top AI engineers for its 'Genesis AI Lab,' indicating a strategic commitment to in-house AI research and development to bolster its investment processes.
•The firm employs a 'quantitative factory model' for its investment research, segmenting the process into factors, AI, optimization, risk control, and trading, ensuring a systematic and coordinated team-based approach to strategy development.
•The broader AI investment landscape experienced record-breaking venture funding in Q1 2026, with significant capital concentration in foundational horizontal AI platforms like OpenAI, Anthropic, and xAI, alongside a growing trend towards specialized vertical AI solutions.

🛠️ Technical Deep Dive

AI agent tasks are significantly more expensive than traditional code reasoning or chat, consuming up to 1000 times more tokens, with input tokens being the primary cost driver.
Token usage in AI agent tasks is highly variable and stochastic, with costs for identical tasks potentially differing by up to 30 times, and increased token consumption does not consistently correlate with higher accuracy.
Frontier AI models frequently fail to accurately predict their own token usage, often systematically underestimating the actual costs incurred during operations.
Huang Mingming noted that the computational cost for AI inference has decreased by approximately 280 times, with the Mixture of Experts (MoE) architecture further reducing this cost by an additional 80%.

🔮 Future ImplicationsAI analysis grounded in cited sources

AI will become an essential national infrastructure layer globally.

NVIDIA CEO Jensen Huang posits that AI should be viewed as critical infrastructure, akin to electricity and roads, enabling every nation to develop its own AI capabilities using open models and local expertise.

The financial management of AI will shift towards token-based consumption models, necessitating new FinOps strategies.

As AI adoption scales, enterprises will face volatile and non-linear token consumption costs, requiring robust FinOps discipline and hybrid infrastructure models for sustainable AI deployment and cost control.

The global competition for top AI engineering and research talent will intensify, with Chinese firms actively challenging Western dominance.

Chinese quantitative hedge funds, including Mingshi Capital, are aggressively recruiting AI talent and offering competitive salaries to secure the expertise needed for AI breakthroughs and market expansion.

⏳ Timeline

2010

Mingshi Investment Management founded.

2020-07

Mingshi Investment Management, an AI-powered hedge fund, approaches $1 billion in assets under management.

2025-03

Mingshi Investment Management is noted among Chinese asset managers intensifying AI research, with its Genesis AI Lab actively hiring computer scientists.

2025-09

Huang Mingming participates in a roundtable, identified as an 'optimist' investor in AI agents, having invested in projects like GenSpark, Lovart, and Sheet0.