The global AI race has taken a dramatic turn with the emergence of DeepSeek, a Chinese AI startup that has sent shockwaves through the tech world. Despite constraints imposed by U.S. export controls and limited access to the most advanced semiconductor technologies, DeepSeek has demonstrated that innovation does not always stem from the biggest players with the deepest pockets. The implications of this development stretch far beyond China’s AI ambitions, challenging the existing hierarchy of AI development and reshaping the future of technological competition.
Just 20 months old, DeepSeek was founded by Liang Wenfeng, a man whose journey embodies the resilience and ingenuity that drive technological breakthroughs. In 2022, Liang faced severe setbacks when his AI-driven stock-picking venture, Zhejiang High-Flyer Asset Management, faltered, wiping out billions in assets. Yet, what appeared to be a professional downfall laid the foundation for something far greater—DeepSeek.
The startup’s rapid ascent is not just an underdog success storey; it is a direct challenge to industry giants such as Nvidia, OpenAI, and other U.S.-based AI firms. With limited resources and using Nvidia’s less-advanced A100 chips—hardware deemed unworthy of U.S. export restrictions—DeepSeek managed to achieve AI capabilities comparable to OpenAI’s models. This raises fundamental questions about the AI industry’s dependency on high-end computing power and whether massive capital expenditure is truly the defining factor in AI innovation.
Shaking the Foundations of the AI Industry
The ripple effect of DeepSeek’s progress was almost immediate. On January 6, Nvidia unveiled its Blackwell chips, boasting enhanced efficiency for trillion-parameter language models. Its stock soared to an all-time high, only to plummet by 17% days later as investors realised that DeepSeek had made significant AI advancements without relying on Nvidia’s latest hardware.
This raised a crucial question: if DeepSeek could achieve such breakthroughs with older-generation AI chips, could others follow suit? Would AI firms start shifting away from Nvidia’s premium chips, disrupting the company’s dominance in AI computing? Nvidia’s response was an attempt at damage control, emphasising that DeepSeek’s success validated the demand for its products in China. However, the reality remains that DeepSeek’s achievement has cast doubt on whether cutting-edge AI truly requires hundreds of billions of dollars in infrastructure.
Geopolitics and the Timing of DeepSeek’s Breakthrough
The timing of DeepSeek’s release of its R1 advanced AI reasoning model on January 20, the same day Donald Trump was inaugurated as America’s 47th president, adds an intriguing geopolitical dimension. Was this a deliberate move? Trump himself acknowledged DeepSeek as a “wake-up call,” signalling the growing awareness of China’s AI capabilities.
The Chinese government wasted no time recognising DeepSeek’s success. Liang Wenfeng was invited to a closed-door symposium in Beijing, hosted by Premier Li Qiang, where discussions centred on draughting AI policies that could shape China’s technological future. This underscores the strategic significance of DeepSeek’s rise—not just as an AI company but as a symbol of China’s push to assert dominance in emerging technologies.
Lessons for the Future of AI Development
One of the most striking aspects of DeepSeek’s success is its ability to operate efficiently under constraints. AI researcher Kai-Fu Lee has long argued that while the U.S. excels in breakthrough research, China is superior in engineering and efficiency. DeepSeek’s accomplishment reinforces this notion, proving that even with limited computing power and funding, innovation can thrive with the right talent and approach.
Liang Wenfeng’s recruitment strategy also provides valuable insight into China’s AI development model. DeepSeek has largely relied on young engineering talent from top Chinese universities—doctoral candidates, interns, and Olympiad medallists. This approach prioritises high-density innovation hubs, fostering an environment where cutting-edge AI can be developed without excessive reliance on foreign expertise.
Can DeepSeek Sustain Its Momentum?
Despite its impressive debut, DeepSeek faces significant challenges. If it needed 10,000 A100 chips to achieve its current breakthrough, scaling up would require around 50,000 H100 chips—hardware it cannot legally access due to U.S. export restrictions. This presents a critical hurdle: will China be able to develop its own advanced AI chips to sustain DeepSeek’s growth?
Moreover, while DeepSeek has made a strong entry, the U.S. tech industry still holds the upper hand in the race toward Artificial General Intelligence (AGI). The infrastructure and capital backing companies like OpenAI, Google DeepMind, and Meta remain unparallelled, raising the question of whether DeepSeek can truly compete in the long run.
A Disruptor in the Making
DeepSeek’s rise serves as both a disruption and a warning. It highlights the ability of underdogs to challenge industry leaders and demonstrates that AI development is not solely dictated by access to the most advanced hardware. It also reflects the broader geopolitical competition between the U.S. and China, where technological advancements are increasingly intertwined with national security concerns.
Liang Wenfeng’s mindset—focused more on innovation than monetary gain—makes him a wildcard in this battle. His lean, analytical approach evokes historical figures who have defied expectations and changed the course of history. As the AI race continues to evolve, one thing is certain: DeepSeek has altered the narrative, and the future of AI will be shaped by its unexpected rise.
