DeepSeek Founder's Fund Surges 57%: The Quant AI Revolution

The intersection of high-frequency trading and frontier AI model development has just produced a staggering metric: 57%. That’s the average return Liang Wenfeng’s quantitative hedge fund, Zhejiang High-Flyer Asset Management, posted for 2025.

While global markets grappled with volatility, High-Flyer didn’t just survive; it dominated. This massive influx of capital isn’t just a win for investors—it’s a direct injection of “firepower” into DeepSeek, the AI lab that has been redefining the cost-to-performance ratio of Large Language Models.

Key Takeaways

• Massive Returns: High-Flyer’s funds averaged a 56.6% return in 2025, nearly double the 30.5% average of its Chinese quant peers.
• The “War Chest” Effect: Liang Wenfeng owns both High-Flyer and DeepSeek, creating a unique self-funding ecosystem where trading profits subsidize massive AI compute costs.
• Quant Dominance: AI-driven trading strategies are increasingly outperforming traditional human-led investment vehicles.

The “High-Flyer” Flywheel

In the world of AI development, compute is currency. OpenAI has Microsoft; Anthropic has Amazon. DeepSeek has High-Flyer.

Liang Wenfeng’s dual role as a hedge fund mogul and an AI lab founder creates a powerful feedback loop. High-Flyer uses advanced algorithms—likely benefiting from the same underlying research that powers DeepSeek—to extract alpha from the markets. These profits, in turn, fund the enormous GPU clusters required to train the next generation of DeepSeek models.

As reported by Bloomberg, this performance makes High-Flyer the second-best performer among Chinese “10 billion yuan club” private quant funds. This financial independence allows DeepSeek to pursue aggressive research directions without being beholden to traditional venture capital timelines.

From Quant Funds to Autonomous Agents

The success of High-Flyer is a real-world validation of algorithmic supremacy. It demonstrates that automated, data-driven systems can consistently outperform human intuition in complex, high-stakes environments.

This is the exact thesis behind Google’s AlphaEvolve and the broader move towards agentic finance. If a centralized quant fund can achieve 57% returns using proprietary algorithms, imagine the potential of decentralized, multi-agent systems that can govern themselves.

This brings us to Project GOOSE.

While High-Flyer represents the pinnacle of centralized quant trading, the GOOSE Intelligent Trading System represents the future of agentic market intelligence. Built on a 5-agent architecture (Market Analyst, News Sentinel, Risk Manager, Execution Agent, and Coordinator), GOOSE brings the same rigorous, automated decision-making to prediction markets.

The Future is algorithmic

The era of the “star trader” relying on gut instinct is fading. Whether it’s Liang Wenfeng’s High-Flyer fund crushing the Shanghai turnover or autonomous AI agents managing prediction market portfolios, the signal is clear: The future of finance is code.

DeepSeek’s ability to leverage this financial engine makes them a dangerous competitor in the AI arms race. They aren’t just building models; they are proving that their underlying intelligence can generate the capital needed to sustain itself.

For enterprises, the lesson is simple: Adopt agentic workflows and data-driven decision engines now, or be outpaced by those who do.

Sources: Bloomberg, Business Times