In the hyper-competitive world of artificial intelligence, where billion-dollar corporations and academic powerhouses jostle for dominance, a new name has emerged from the East, sending ripples of both excitement and unease through the established order: Liang Wenfeng, the founder of DeepSeek.
For those in the know within China’s burgeoning tech scene, Liang Wenfeng is hardly a newcomer. He’s been a significant figure in the country’s quantitative finance sector for years, a realm often shrouded in secrecy and technical complexity to outsiders. However, for many in the West, and particularly in the United States, Liang Wenfeng and his company DeepSeek arrived seemingly out of nowhere, like a “mysterious Eastern power” suddenly making its presence felt on the global stage.
DeepSeek’s recent unveiling of its cutting-edge AI models, particularly the DeepSeek-V3 and the open-source DeepSeek-R1, has sparked a global conversation. These models aren’t just technically impressive; they’ve been touted for their remarkable efficiency, achieving performance comparable to, and in some areas even surpassing, industry giants like OpenAI’s ChatGPT, but at a fraction of the cost. This feat has earned DeepSeek the moniker of the “Pinduoduo of AI,” a reference to the Chinese e-commerce platform known for its disruptive, value-driven approach that shook up the established giants of online retail in China. According to a report by 华尔街见闻 (Wall Street Insights), this aggressive pricing strategy has forced larger companies to respond.
But who is Liang Wenfeng, the man behind this AI phenomenon? For an American audience just beginning to grapple with China’s growing technological prowess, understanding the story of Liang Wenfeng and DeepSeek is crucial to comprehending the shifting dynamics of the global AI race. He’s not your typical Silicon Valley founder, nor does he fit neatly into the mold of a Chinese tech titan. Instead, Liang Wenfeng emerges as a unique blend of quantitative finance acumen, deep-seated technical passion, and a somewhat enigmatic, almost minimalist personal style that has captivated observers both within China and abroad.
This is the story of Liang Wenfeng, a Guangdong native who, with a team of mostly young, homegrown Chinese programmers, has built DeepSeek into a force to be reckoned with, challenging the very foundations of the AI landscape and prompting a global reassessment of what it takes to lead in this transformative technology.
From Guangdong to the Quant World: The Early Years
Born in 1985 in Zhanjiang, a city in Guangdong province, Liang Wenfeng’s early life offered little hint of the technological disruption he would later unleash. He himself describes his hometown as a “fifth-tier city,” although by today’s standards, Zhanjiang would likely be classified as a third-tier city in China, indicating a city undergoing significant development but still outside the top tier metropolises like Beijing or Shanghai. His father was a primary school teacher, a modest profession in a society rapidly transforming economically. Details about his early schooling are scarce, adding to the air of mystery surrounding his persona. However, it is known that even as a young student, Liang Wenfeng demonstrated a keen interest in mathematics and particularly in mathematical modeling.
In 2002, at the age of 17, Liang Wenfeng embarked on a journey that would fundamentally shape his future, enrolling at Zhejiang University, one of China’s most prestigious institutions, to study Electronic and Information Engineering. Zhejiang University, located in the vibrant city of Hangzhou, is often likened to MIT in China, renowned for its engineering and technology programs. He pursued both his undergraduate and master’s degrees there, specializing in Information and Communication Engineering for his postgraduate studies. His master’s thesis focused on “Target Tracking Algorithm Research Based on Low-Cost PTZ Cameras,” hinting at his early engagement with machine vision and artificial intelligence concepts.
Even during his university years, Liang Wenfeng’s entrepreneurial spirit and technical curiosity were already evident. In 2008, while still at Zhejiang University, he teamed up with fellow students to begin gathering market data, financial market information, and macroeconomic indicators. This wasn’t just academic curiosity; Liang Wenfeng and his team were already experimenting with machine learning techniques to explore the then-nascent field of fully automated quantitative trading. This was a particularly prescient move, as 2008 marked the onset of the global financial crisis, a period of immense market volatility that paradoxically underscored the potential of data-driven, algorithmic approaches to finance.
Legend has it that during this early period, Liang Wenfeng was even approached by Wang Tao, the founder of DJI, the world-leading drone manufacturer, with an invitation to join his burgeoning venture. DJI, also based in Shenzhen, was then in its early startup phase, and joining at that point could have set Liang Wenfeng on a very different path. However, Liang Wenfeng’s conviction in the transformative power of artificial intelligence was already firmly rooted. He chose to forge his own path, driven by a belief that AI would fundamentally reshape the world.
Venturing into the Quant Frontier
Upon graduating with his master’s degree in 2010, Liang Wenfeng didn’t follow the well-trodden path of many of his peers into established tech giants. Instead, he embarked on a more unconventional and challenging route, venturing into the then-uncharted territory of quantitative investment. He moved to Chengdu, a major city in Western China known for its laid-back culture and growing tech scene, and began his entrepreneurial journey in solitude, renting a small apartment as his initial base of operations.
Quantitative investment, at that time, was still a relatively nascent field in China, perceived by many as an unproven and even risky approach. Many doubted whether relying solely on computer algorithms and data models could truly generate profits in the complex and often unpredictable financial markets. However, Liang Wenfeng was deeply drawn to the scientific logic underpinning quantitative investing. He held a firm belief that beneath the seeming chaos of market fluctuations, there existed patterns and rules that could be captured and exploited through sophisticated mathematical models. The backdrop of the global financial crisis only reinforced his conviction in the need for more robust and data-driven investment strategies.
These were undoubtedly challenging early days. Faced with skepticism and the inherent difficulties of building sophisticated trading systems from scratch, Liang Wenfeng drew inspiration from the words of quant finance pioneer James Simons, the founder of Renaissance Technologies: “There is always a way to model price.” This mantra became his guiding light, fueling his relentless efforts as he immersed himself in oceans of data, tirelessly experimenting with algorithms and models, seeking the key to unlock financial success through quantitative methods.
After countless nights of dedicated research and iterative testing, Liang Wenfeng’s perseverance began to pay off. The year 2010 proved to be a pivotal moment for quantitative investing in China with the launch of the CSI 300 Index futures. This landmark event provided quant investors with powerful new tools and opportunities, acting as a catalyst for the sector’s growth. Seizing this opportune moment, Liang Wenfeng, leveraging his deep technical expertise and sharp market insight, led his team to capitalize on this emerging landscape. The quantitative trading strategies they developed proved remarkably successful, leading to rapid growth in their proprietary capital, exceeding 500 million RMB (approximately 70 million USD at the time).
This early success not only validated Liang Wenfeng’s vision and hard work but also firmly established him in the burgeoning quantitative investment scene in China. It laid a solid foundation for his future endeavors, marking the beginning of his ascent to becoming a leading figure in both finance and, eventually, artificial intelligence.
Quant Finance Titan and the AI Awakening
Liang Wenfeng’s journey into AI was not a sudden pivot but rather a gradual evolution, deeply intertwined with his success in quantitative finance. In 2013, he co-founded Jacobi Investment Management in Hangzhou with Xu Jin, a fellow Zhejiang University alumnus. Two years later, in 2015, recognizing the transformative potential of China’s evolving financial market landscape, particularly the launch of CSI 500 index futures which ushered in “Quant Private Equity 2.0 Era”, Liang Wenfeng, along with Xu Jin, ventured into the private fund sector, establishing High-Flyer Technology in Hangzhou. This company would later become Zhejiang Jiuzhang Asset Management Co., Ltd., commonly known as HF Quant or HFF Investment Management, a name that would become synonymous with quantitative investment prowess in China.
From the outset, Liang Wenfeng envisioned HFF Investment Management as a technology-driven fund, aiming to become a world-class quantitative hedge fund powered by mathematics and artificial intelligence. Even in the face of the 2015 Chinese stock market “stock market crash,” HFF’s high-frequency quantitative trading strategies delivered impressive results, even with limited computational resources – initially just 10 GPUs. The firm quickly expanded, launching multiple products in rapid succession, demonstrating its growing fundraising capabilities.
According to 新浪科技, Liang Wenfeng was already recognized as a talent back in 2009.
By 2016, HFF Investment Management officially registered with the Asset Management Association of China (AMAC), signaling its commitment to industry standards and professional practices. That same year marked a critical milestone in Liang Wenfeng’s AI journey. HFF launched its first AI model, deploying GPU computing to execute trading strategies generated through deep learning. Prior to this, their algorithms primarily relied on linear models and traditional machine learning techniques, with computations largely handled by CPUs. By the end of 2016, just a year after its founding, HFF’s assets under management had already reached approximately 1 billion RMB (around 140 million USD), a testament to the rapid growth and success fueled by their early adoption of AI.
The following years saw continued expansion and deepening of HFF’s commitment to AI. In 2017, Liang Wenfeng spearheaded the expansion of both the AI algorithm research team and the AI hardware and software development teams. By the end of that year, nearly all of HFF’s quantitative strategies were powered by AI models, and assets under management had tripled to 3 billion RMB. In 2018, Liang Wenfeng solidified AI as the company’s primary strategic direction. HFF Investment Management received its first Golden Bull Award for private equity, the highest honor in China’s private securities sector.
However, this rapid AI-driven growth also brought new challenges, particularly in computational power. The increasing demands of AI model training began to strain HFF’s existing computing resources. Recognizing this bottleneck, Liang Wenfeng embarked on a bold and ambitious plan to build a large-scale AI supercomputing infrastructure.
The “Firefly” Era: Building AI Muscle
In 2019, Liang Wenfeng founded HF AI and invested a staggering 200 million RMB (approximately 28 million USD) to independently develop a deep learning training platform codenamed “Firefly No. 1.” This initial supercomputing cluster was equipped with 1,100 GPUs, a remarkable feat at the time. By this point, Liang Wenfeng had already led HFF Investment Management to become a prominent quant fund managing over 10 billion RMB in assets. HF Capital was also established in Hong Kong, obtaining a Type 9 license for asset management.
In August 2019, Liang Wenfeng delivered a keynote speech at the Golden Bull Awards ceremony titled “The Future of Quantitative Investment in China from a Programmer’s Perspective,” which resonated deeply within the financial industry. In his address, Liang Wenfeng emphasized the data-driven, algorithmic nature of quantitative investing, stating that in true quant firms, “there are no fund managers, fund managers are just a bunch of servers.” He openly acknowledged the high expectations and pressures from investors in the private equity sector, while also highlighting the significant fees charged for their services. He predicted that quantitative investing would increasingly encroach on the territory of fundamental analysis, ultimately aiming to improve the efficiency of China’s secondary markets – a mission statement for HFF.
“Firefly No. 1” became operational in 2020, with a lifespan of about 18 months. A HFF employee noted on the company blog that “the boss himself writes code and runs code every day,” highlighting a key differentiator for the firm – a deep technical immersion from the top down.
2021 proved to be a pivotal year for Liang Wenfeng and HFF. That year, they invested another 1 billion RMB to build “Firefly No. 2,” a significantly more powerful supercomputing cluster. In February 2021, Liang Wenfeng wrote a preface for the Chinese edition of “The Man Who Solved the Market: How Jim Simons Launched the Quant Revolution,” a biography of James Simons. In it, he wrote, “Whenever I encounter difficulties at work, I think of Simons’ words, ‘There must be a way to model price.’”
In August 2021, HFF Investment Management’s assets under management surpassed 100 billion RMB, firmly establishing it as one of China’s top quantitative private equity firms, alongside firms like NineQuants, Minghong Investment, and Lingjun Investment, collectively known as the “Four Heavenly Kings” of quant private equity. It had taken HFF just four years to reach 100 billion RMB in AUM, a remarkable growth trajectory under Liang Wenfeng’s leadership.
However, late 2021 also brought challenges. Performance fluctuations led HFF to temporarily close all fundraising channels in November. In December, Liang Wenfeng issued a public apology to investors, acknowledging that “HFF’s performance drawdown has reached a historical maximum, for which we are deeply ashamed,” attributing the setback to shortcomings in AI-driven investment decisions, particularly in timing trades. Reports indicated that HFF had proactively reduced its AUM by over 10 billion RMB within a month and a half, from nearly 100 billion RMB to around 80 billion RMB, gradually scaling back its overall size.
Despite these setbacks in the quant finance realm, 2021 was also the year Liang Wenfeng secured a significant advantage in the AI arena. HFF became one of the first private companies in China to obtain NVIDIA A100 GPUs, with its AI cluster boasting a “ten-thousand-card level” scale. According to 腾讯新闻, at the time only about 5 companies in China exceeded 10,000 GPU enterprise, and except the hff, the other four are Internet giants. This massive investment in AI infrastructure positioned Liang Wenfeng for his next, even more ambitious venture: DeepSeek.
The Birth of DeepSeek: A New Quest for AGI
In May 2023, at the age of 38, Liang Wenfeng announced his ambitious foray into the realm of general-purpose artificial intelligence (AGI). In July of the same year, he founded Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co., Ltd., the company behind DeepSeek. Liang Wenfeng became known as the “first person” among quantitative investors to venture into AI entrepreneurship, although his quant firm HFF had been leveraging AI technology for years.
In October 2023, Xu Jin, Liang Wenfeng’s long-time partner, faced personal issues that led to his temporary suspension from the company. By this time, HFF Investment Management’s AUM had decreased to around 40 billion RMB, a deliberate scaling back as Liang Wenfeng shifted his focus towards DeepSeek.
May 2024 saw the launch of DeepSeek’s mixed-expert language model DeepSeek-V2. Its API pricing was revolutionary – just 1 RMB per million tokens for input and 2 RMB per million tokens for output, a mere fraction of the cost of GPT-4 Turbo. DeepSeek-V2’s ultra-competitive pricing triggered a wave of price reductions from major Chinese tech companies like ByteDance, Alibaba, Baidu, and Tencent, earning DeepSeek the reputation of the “Pinduoduo of AI.”
In October 2024, HFF Investment Management announced to investors its plan to gradually reduce its hedge product investment positions to zero, further signaling Liang Wenfeng’s strategic shift towards AI.
Then, in December 2024, DeepSeek unveiled DeepSeek-V3, accompanied by a 53-page technical report co-authored by Liang Wenfeng and Luo Fuli, among others. DeepSeek-V3’s performance benchmarks, as showcased on the company’s website, “outclassed” numerous leading AI models, both domestically and internationally, particularly in math, coding, and Chinese knowledge Q&A, all while maintaining its cost-effectiveness. This solidified DeepSeek’s image as a “mysterious Eastern force” disrupting the AI landscape.
According to 雷峰网, Liang Wenfeng does not think of himself as a financial person, he thinks of himself as doing AI, but the use cases happen to be in finance.
OpenAI co-founder Andrej Karpathy publicly praised DeepSeek’s achievements, acknowledging their impressive research and engineering capabilities, particularly given their resource constraints. What made DeepSeek even more remarkable was its entirely homegrown R&D team – composed solely of Chinese programmers, many of whom were recent graduates or young professionals with only a few years of experience, bucking the trend of heavily relying on overseas returnees. Some reports even suggested DeepSeek intentionally avoided hiring senior technical professionals, fearing that those with extensive experience might be less innovative and burdened by established paradigms. As one industry insider aptly put it, “Innovation requires breaking free from inertia.”
DeepSeek’s talent pool was drawn from top Chinese universities like Peking University and Tsinghua University, with many making significant contributions during their internships. This echoed a similar approach adopted by OpenAI, which also prioritized hiring high-potential newcomers over established academics. Liang Wenfeng explained his talent philosophy: “If you are pursuing short-term goals, hiring experienced people is the right way. But if you are looking at the long term, experience is not that important; basic ability, creativity, and passion are more important.”
Liang Wenfeng recognized early on that DeepSeek’s research-first approach, prioritizing fundamental breakthroughs over immediate commercialization, would likely not attract traditional venture capital. VCs, with their focus on quick returns and exit strategies, would be impatient with DeepSeek’s long-term AGI vision. Therefore, DeepSeek’s AI endeavors have been primarily funded by the profits generated by his quant finance firm, HFF Investment Management.
By 2025, HFF Investment Management’s AUM had intentionally shrunk to under 30 billion RMB, falling outside the top six quant private equity firms in China, a clear indication of Liang Wenfeng’s unwavering commitment to his AI pursuit. In January 2025, following an earthquake in Xigaze, Tibet, HFF Investment Management donated 1 million RMB for disaster relief, demonstrating a continued commitment to social responsibility even amidst its strategic shift.
Intriguingly, despite rumors of a high-profile poaching attempt by Xiaomi’s Lei Jun, DeepSeek researcher Luo Fuli’s departure from the company was confirmed by a DeepSeek publication, although her potential move to Xiaomi remains unconfirmed.
Liang Wenfeng’s emphasis on technological originality is unwavering. He stated: “China must have someone standing at the forefront of technology… In the past thirty years of the IT wave, we have basically not participated in real technological innovation. We are used to Moore’s Law falling from the sky, lying at home and better hardware and software will come out in 18 months… We believe that with economic development, China should gradually become a contributor, not just a free rider… The real gap is not a year or two, but the difference between originality and imitation.”
Under Liang Wenfeng’s leadership, DeepSeek embodies this spirit of originality and innovation, offering a compelling example of China’s potential to compete with, and even challenge, Western dominance in the AI field, particularly against the backdrop of US chip export restrictions. According to 钛媒体, even Zhu Xiaohu, a partner at GSR Ventures whose AI investment attitude had a reversal, previously mentioned not to invest in general model company, but DeepSeek changed his attitude.
The “AI Catfish” and the Future of DeepSeek
Liang Wenfeng’s journey, from a math enthusiast in Guangdong to a quant finance titan and now an AI disruptor shaking Silicon Valley, is a testament to his unwavering belief in technology-driven innovation. His decision to apply AI in finance initially provided the financial foundation for his more ambitious AI research pursuits, creating a virtuous cycle of innovation and resource accumulation.
Remarkably, Liang Wenfeng has launched major ventures during periods often perceived as “capital winter” – 2008 for quant trading exploration, 2015 for HFF Investment Management, and 2023 for DeepSeek. Each time, he has not only survived but thrived, demonstrating a resilience and “strength in adversity” that underscores his deep-seated capabilities.
Having achieved significant financial success, Liang Wenfeng has now set his sights on AGI, viewing DeepSeek as a “curiosity-driven exploration, academic research” rather than a purely profit-seeking endeavor, a philosophy reflected in his decision to open-source DeepSeek’s technology. His unconventional management style, characterized by attracting high-potential talent, a flat organizational structure, and a lack of rigid KPIs, fostering employee autonomy and creativity, has also contributed to DeepSeek’s unique identity and rapid rise.
Still in his early forties, Liang Wenfeng’s calm determination and unwavering faith in artificial intelligence evoke a sense of awe and respect, traits often associated with truly transformative CEOs. DeepSeek, under his guidance, has emerged as a significant force, challenging the established narratives of AI and prompting a global re-evaluation of the landscape. Whether DeepSeek can sustain this momentum and truly become a long-term challenger to OpenAI remains to be seen, but one thing is clear: Liang Wenfeng has already made an indelible mark on the world of AI, cementing his place as a key figure in this technological revolution and earning the moniker of the “AI Catfish,” stirring up the waters and challenging the established order with his innovative and disruptive approach. And this article also sees him as the “Huang Zheng in the AI field”.
评论