DeepSeek vs. OpenAI: How China’s AI Underdog Is Winning Against the Odds

Jan 29

In the high-stakes world of artificial intelligence, where Silicon Valley giants like OpenAI and Google DeepMind dominate, a new contender has emerged from an unexpected place—China. DeepSeek, an AI-powered chatbot developed by Chinese hedge fund mogul Liang Wenfeng, is rapidly ascending the global AI ranks, leaving tech insiders questioning how a relatively obscure company has managed to defy U.S. export restrictions and develop a powerful AI model.

The AI Powerhouse That No One Saw Coming

DeepSeek-AI Assistant has skyrocketed to the top of Apple’s App Store charts in both China and the U.S., making it the most popular free app, according to San Francisco-based analytics firm Sensor Tower. With capabilities akin to OpenAI’s ChatGPT—answering questions, drafting content, and aiding in information retrieval—DeepSeek is proving that innovation isn’t exclusive to Silicon Valley.

The Hangzhou-based startup has captivated AI researchers with its DeepSeek R1 model, which has quickly become the fourth most popular AI model globally by user vote, according to Chatbot Arena, a platform developed in collaboration with researchers from the University of California, Berkeley. Notably, DeepSeek R1 ranks ahead of OpenAI’s o1 model, released in December 2023.

Despite the U.S. government’s stringent export controls limiting China’s access to high-performance AI chips, DeepSeek has managed to develop advanced AI models at a fraction of OpenAI’s costs. The company claims that some of its AI models were trained for as little as $5.6 million—up to 95% less than OpenAI’s reported expenditures. The secret? A strategic use of the "mixture of experts" (MoE) technology, a method that trains multiple smaller AI models simultaneously and selects the best-performing ones to generate responses. This approach not only reduces reliance on expensive semiconductors but also enhances efficiency.

Who Is Liang Wenfeng, the Man Behind DeepSeek?

A reclusive figure in China’s tech landscape, Liang Wenfeng is not your typical AI entrepreneur. Born in Guangdong in 1985, Liang’s early career revolved around investment strategies rather than artificial intelligence. After earning a master’s degree in computer vision from Zhejiang University, he ventured into AI-driven stock trading, founding hedge fund High-Flyer Quant in 2015. By 2017, his firm had fully transitioned to AI-based investment strategies, amassing assets under management (AUM) of 10 billion yuan ($1.4 billion) by 2019.

However, Liang’s ambitions stretched far beyond finance. Foreseeing the potential of AI, he began stockpiling Nvidia chips before Washington imposed restrictions. By 2023, he launched DeepSeek, entirely self-financed, with no outside investors, according to Chinese business data platform Qichacha.

Disrupting the AI Economy

DeepSeek’s business model mirrors OpenAI’s in some ways but undercuts it in pricing. While OpenAI charges $7.50 per million tokens processed, DeepSeek R1 charges a mere 14 cents. The chatbot itself is free, but developers pay to access its underlying model, allowing for customised AI applications.

This affordability has fueled its rapid adoption, sparking speculation about its long-term impact on both Western AI firms and China’s semiconductor industry. "DeepSeek doesn’t rely on vast amounts of computing power, which means it doesn’t need as many Nvidia GPUs," says Shen Meng, managing director at Beijing-based investment bank Chanson & Co. "This could signal a major shift for China’s chip developers."

The Future of AI Beyond Silicon Valley

DeepSeek’s meteoric rise challenges the long-held notion that the most cutting-edge AI developments can only emerge from U.S.-based firms with unlimited access to capital and technology. As U.S. export controls tighten, DeepSeek’s innovative approach may pave the way for other Chinese AI firms to thrive despite restrictions.

Billionaire investor Marc Andreessen is already taking notice. "DeepSeek R1 is one of the most amazing and impressive breakthroughs I’ve ever seen—and as open source, a profound gift to the world," he posted on X (formerly Twitter) on January 24.

As DeepSeek continues to push technological boundaries, one thing is clear: the AI race is no longer a one-horse competition. With China’s homegrown AI revolution gaining traction, Silicon Valley might need to rethink its long-term strategies to maintain its lead in the global AI arena.