Wang Wenfeng, the founder of DeepSeek, has emerged as a unique figure in China’s artificial intelligence sector. Unlike many AI entrepreneurs focused on rapid commercialization, Wang prioritizes fundamental research.
His company has inadvertently shaken up the industry, sparking a price war and challenging the business models of major tech firms like ByteDance, Alibaba, Tencent, and Baidu.
Despite the disruption, Wang insists that DeepSeek’s goal is research, not competition.
Wang Wenfeng’s Path to AI Innovation
Wang was born in 1985 in Zhanjiang, Guangdong. He attended Zhejiang University, where he pursued degrees in Electronic Information Engineering and Information & Communication Engineering.
His early interest in financial markets led him to found Hangzhou Yakebi Investment Management Co. and later High-Flyer Quant, one of China’s most successful hedge funds, managing $8 billion in assets.
Unlike many who enter AI for its business potential, Wang’s motivation has always been research-driven. DeepSeek operates independently from High-Flyer, focusing solely on developing general AI models.
The Unintended Price War
DeepSeek’s release of its V2 model disrupted the AI market. By pricing its API access at cost-recovery levels with minimal profit, it forced competitors to lower their prices.
Wang maintains that this was not intentional. “We simply priced according to our costs,” he explained. However, the move exposed the inefficiencies of larger tech firms, whose operational costs far exceed those of a lean, research-focused company like DeepSeek.
Pioneering AI Research in China
DeepSeek takes a different approach from most Chinese AI firms, which often build upon open-source models like Meta’s Llama. Instead, it focuses on foundational research, aiming to close the efficiency gap between Chinese and Western AI models.
Wang estimates that Chinese AI models currently require four times the computational resources of their American counterparts to achieve similar results.
The company’s development of Multi-Head Latent Attention (MLA) architecture is a breakthrough. This design reduces memory consumption to just 5-13% of conventional methods, significantly lowering training and inference costs.
Unlike many Chinese AI firms that rely on talent returning from Silicon Valley, DeepSeek’s team consists of young engineers from top domestic universities. Wang believes that China can build its own AI talent base rather than depending on overseas expertise.
A Different Culture of Innovation
DeepSeek’s management style contrasts with the rigid hierarchies of Chinese tech giants. Engineers have the freedom to allocate GPU resources without seeking approval, and projects emerge from individual curiosity rather than top-down directives.
Wang argues that true innovation comes from an open, research-driven environment rather than immediate monetization. “Chinese companies have spent the last 30 years obsessed with profitability at the expense of curiosity,” he said.
Despite open-sourcing much of its research, DeepSeek is not concerned about competitors copying its work. Wang believes that in transformative technologies, innovation matters more than secrecy.
The Pursuit of Artificial General Intelligence
Unlike many AI startups that balance research with enterprise solutions, DeepSeek remains focused on achieving artificial general intelligence (AGI). Wang sees large language models (LLMs) as a stepping stone, similar to how early neural networks led to deep learning.
DeepSeek is exploring various paths, including mathematical reasoning and multi-modal learning. However, Wang acknowledges that there is no single agreed-upon approach to achieving AGI.
Overcoming Challenges in a Competitive Market
Despite DeepSeek’s success, scaling remains a challenge. While Wang’s background in high-frequency trading has given him access to significant computing resources—DeepSeek stockpiled 10,000 GPUs as early as 2021—the company still operates at a disadvantage compared to AI giants like OpenAI and Google DeepMind.
The company also faces external pressures, including U.S. export controls on high-end AI chips. However, Wang remains optimistic, believing that real innovation does not necessarily come from the biggest budgets.
The Future of DeepSeek
DeepSeek has no plans to pivot toward enterprise applications or seek external funding. Wang is confident that the company’s focus on solving the hardest problems will attract top talent and sustain its long-term vision.
In a rapidly evolving AI landscape, where short-term profits often take priority, Wang Wenfeng is betting that deep, original research will prove to be the most valuable strategy. Whether this approach will pay off remains to be seen, but DeepSeek’s unconventional path has already set it apart from its competitors.
