Liang Wenfeng's dual identity: How does the 56.55% return of Fantom Quant support DeepSeek's AI dream

2026-01-14 07:40:07

DeepSeek founder Liang Wenfeng has a somewhat unique background. He is not only the helm of this AI company but also the founder of the well-known domestic quantitative private fund, Fantasia Quantitative. Recent data shows that Fantasia Quantitative achieved a 56.55% return in 2025, with assets under management exceeding 70 billion yuan. This substantial capital and technological accumulation are becoming a solid backing for DeepSeek.

How Strong Are Fantasia Quantitative’s Performance

According to the latest news, Fantasia Quantitative ranks second among Chinese quantitative private funds managing over 10 billion yuan, only behind Lingjun Investment, which posted a 73.51% return. What does this achievement signify? The market performance during the same period makes it clear.

In 2025, the Shanghai Composite Index, Shenzhen Component Index, and CSI 300 Index rose by only 18%, 30%, and 18%, respectively, while Fantasia Quantitative achieved a 56.55% return. This is not just luck but a validation of long-term capability. Data shows that Fantasia Quantitative’s average return over the past three years is 85.15%, and over the past five years, it’s 114.35%.

Indicator	Fantasia Quantitative	Lingjun Investment	CSI 300
2025 Return	56.55%	73.51%	18%
3-Year Average Return	85.15%	-	-
5-Year Average Return	114.35%	-	-
Assets Under Management	70 billion+	-	-

Liang Wenfeng’s Quantitative DNA

To understand why Fantasia Quantitative is so profitable, it’s essential to know Liang Wenfeng’s background. He founded Fantasia Quantitative in 2008 while studying Information and Communication Engineering at Zhejiang University. This wasn’t an arbitrary idea but stemmed from a deep understanding of mathematics, computing, and systems engineering.

From its inception, Fantasia Quantitative has been characterized by a strong technical gene. In October 2016, it officially launched real-time trading and began using advanced Graphics Processing Units (GPUs) for computation, which was a bold choice at the time. Previously, the industry mainly relied on traditional machine learning algorithms and CPUs, but Liang Wenfeng had already recognized the advantages of GPUs in compute-intensive tasks.

How forward-looking was this decision? Just look at the importance of GPUs in AI training today.

From Quantitative to AI: A Leap

In 2019, Fantasia Quantitative’s assets under management surpassed 10 billion yuan, and in 2021, it briefly exceeded 100 billion yuan. Over the years, it has accumulated not only capital but also a deep understanding of computing, algorithms, and data. In July 2023, Liang Wenfeng spun off the AI laboratory originally belonging to Fantasia Quantitative and established DeepSeek.

In other words, DeepSeek didn’t appear out of nowhere; it grew from the technical soil of quantitative trading. Some analysts believe that Fantasia Quantitative’s impressive performance provided sufficient R&D funding for Liang Wenfeng’s DeepSeek. Does this logic hold? With a management scale of 70 billion yuan and a 56.55% annual return, Liang Wenfeng indeed has ample capital to support a money-burning AI project.

How Quantitative Trading’s Technical DNA Transforms into AI Advantages

What benefits does Liang Wenfeng’s experience in quantitative trading bring to AI? Here are a few key points:

Computing Efficiency: Quantitative trading requires processing massive amounts of data and making decisions in extremely short timeframes, demanding optimal use of computing resources. This mindset directly transfers to AI model training and inference.
Algorithm Design: The core of quantitative trading is designing better algorithms to discover market patterns, which follows a similar logic in optimizing AI models.
Data Handling: The sensitivity and processing capabilities of quantitative funds remain core competitive advantages in the AI era.
Hardware Application: Liang Wenfeng recognized early on the value of GPUs, which directly benefits DeepSeek’s choice to use GPU clusters for model training.

Future Highlights

According to the latest news, DeepSeek plans to release a new flagship AI model, DeepSeek V4, in February. This model boasts powerful programming capabilities and is expected to significantly impact the current AI competitive landscape.

This timing is quite interesting. Fantasia Quantitative just delivered its 2025 performance report, and now DeepSeek’s new model is about to be launched. The parallel development of these two lines demonstrates Liang Wenfeng’s unique entrepreneurial path: making money through quantitative trading while using that wealth and accumulated technology to develop AI.

Summary

Liang Wenfeng’s story exemplifies a unique entrepreneurial model: he didn’t start from zero but built upon an existing successful foundation. Fantasia Quantitative’s 56.55% return in 2025 and its management scale of 70 billion yuan are proof of his personal capability and serve as the financial and technological foundation for DeepSeek.

This model is relatively rare in the AI field. Usually, AI entrepreneurs come from large tech companies or academic backgrounds, but few have established a hundred-billion-level fund in quantitative trading first and then leveraged that platform to pursue AI. When DeepSeek V4 is released, we will see how far this model can go.

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.