As a leading AI application company in Hong Kong, EPC Solutions is keeping pace with industry trends, leveraging the technological strength of China's private enterprises, and making a full-scale push into the international market.
Written by | Shen Jianguang
Originally published on FT Chinese (February 12, 2025). Please credit the source when reposting.
Recently, the AI model DeepSeek-R1, released by the Chinese quantitative fund company High-Fund, has created a global sensation. DeepSeek not only significantly narrows the AI gap between China and the U.S., shattering the myth of American technological supremacy, but also offers a cost-effective solution that facilitates large-scale applications. At the same time, DeepSeek has become the latest symbol of China's smart innovation capabilities. Since its launch, confidence in China's innovation capacity and overall national competitiveness has improved both domestically and internationally, leading to a rise in the premium on Chinese assets and a notable boost in the stock market.
However, a closer examination of DeepSeek success reveals that it is a triumph of decentralized and market-driven innovation, demonstrating the critical role of capital markets and the resilience of the private sector. If China aims to foster more innovations like DeepSeek and achieve breakthroughs in new productive forces for its economy, the author believes that improvements should be made in three key areas: the economic innovation model, financial regulatory arrangements, and policies that encourage private sector development.
The significant impact of DeepSeek's release.
Since the release of DeepSeek-R1 on January 20, it has sparked a global sensation. DeepSeek not only briefly topped the free app download charts in multiple countries on Apple's App Store, but also triggered a significant adjustment in the U.S. stock market due to its challenge to the development model of advanced artificial intelligence, which traditionally relied on computational power. The world’s largest company by market capitalization, Nvidia, saw its value drop by nearly a fifth overnight. Media worldwide widely reported this breakthrough, with Americans concerned about the potential loss of technological dominance, while Japan, Europe, and India reflected on why they hadn't produced a model like DeepSeek.
From my perspective, the emergence of the DeepSeek model carries three key significances:
First, DeepSeek has greatly narrowed the AI gap between China and the U.S. By optimizing its algorithms, DeepSeek has reached a level comparable to the U.S. giant OpenAI’s inference models, significantly reducing the reliance on computational power and high-performance chips. China has not only circumvented the U.S. chip embargo and caught up in the AI field, but may also alter the market structure of large model industries, breaking the myth of U.S. technological supremacy in AI. If the effects of scale, first-mover advantage, and cutting-edge chips are no longer as pronounced, new large models could emerge in countries around the world. U.S. AI companies may find it hard to achieve the type of monopoly seen in the search engine field with Google, and Chinese companies could win a place in this critical future industry.
Second, DeepSeek significantly reduces the cost of AI models, paving the way for their broader application. Reports indicate that the training and operating costs of DeepSeek are only one-tenth of those of its U.S. counterparts. Currently, major cloud service providers have already deployed the DeepSeek model, including U.S. companies such as Microsoft and Amazon. DeepSeek's applications in finance, media, and other fields are also gradually being realized.
Third, DeepSeek may lead the world to reconsider China's innovation capacity, boosting both domestic and international confidence in the Chinese economy and enhancing the premium level of Chinese assets. DeepSeek was entirely developed by a Chinese company, and most of the research and development team is based in China. This breakthrough effectively counters doubts about China’s innovation capabilities amid the potential decoupling in the China-U.S. technology sector and even prompts a reassessment of China’s economic strength and national competitiveness. As the myth of overvaluation in the U.S. stock market fades, international investment banks have released reports urging a re-evaluation of China’s innovation capacity. The discount on Chinese assets may disappear and even turn into a premium. The A-share and Hong Kong stock markets have also significantly rebounded, with the Hang Seng Tech Index, which best represents Chinese tech companies, rising by more than 15% since DeepSeek's release.
Looking back at DeepSeek’s growth and achievements, I believe there are three key lessons to be learned:
Insight 1: DeepSeek is a success of decentralized and market-driven innovation.
Although DeepSeek's success is commendable, the reflection on China's path to technological innovation should not stop here. I have noticed that the emergence of DeepSeek is almost entirely the result of independent innovation by a market-driven company, with little to no connection to state planning or supportive policies. This may suggest that future innovation in China will need to rely more on decentralized, micro-level decision-making, and require greater respect for market forces and the curiosity of tech professionals. It cannot simply rely on centralized planning and a state-driven system.
Existing information shows that DeepSeek does not belong to any "national major science and technology project," nor has it received any policy-based funding. It did not involve researchers from public universities or research institutions. It is entirely the product of High-Fund Quantitative own efforts to recruit talent and push forward research. Even more surprising is that High-Fund Quantitative support for DeepSeek was not purely for commercial purposes. Although the company initially aimed to optimize its investment portfolio and enhance returns through artificial intelligence, after 2023, High-Fund Quantitative separated its AI team to join the global AI large model competition and promised to make most of their research findings public. A key motivation was to prevent technology from being monopolized by a few and to avoid imitating overseas tech companies' pursuit of a "historic mission." High-Fund Quantitative portfolio management does not really need DeepSeek. The result, however, is that the release of the DeepSeek-R1 model has made an immediate and significant impact.
This phenomenon seems to once again confirm the view from the classic work Why Greatness Cannot Be Planned: key innovations often come from unexpected discoveries and cannot be predetermined by plans. The discovery of penicillin happened by accident during the cultivation of staphylococci; the Wright brothers invented the airplane using bicycle technology; the first electronic computer used vacuum tubes, which were not originally invented for computers. In the case of DeepSeek, this large model, with its wide applications and significant implications, may have originated from the exploration of quantitative stock investment, which didn’t even require the application of DeepSeek itself.
Unlike in the catching-up phase, where original inventions are hard to plan or define with specific goals, Why Greatness Cannot Be Planned states, "For great enterprises, goals are misleading" because they narrow the scope of exploration. From this perspective, China's top-level design for becoming an innovative economy needs further refinement.
Insight 2: DeepSeek's success is inseparable from a developed capital market.
Although DeepSeek's training costs are significantly lower than those of its American counterparts, its success still required substantial financial investment and resource support. Forget the expensive high-performance chips; even the salaries of the personnel involved in the development of DeepSeek are not something an ordinary company can easily afford. Media reports indicate that High-Fund Quantitative offered salaries exceeding one million RMB to fresh graduates. Some Nvidia interns even chose to forgo offers from Nvidia to join DeepSeek.
It was in China’s capital market that High-Fund Quantitative accumulated enough funds and resources to support the development of DeepSeek. Established in 2015, High-Fund Quantitative reached an asset management scale of over 10 billion RMB within four years. At its peak in 2021, its asset management scale approached 100 billion RMB, with management fee revenue reaching the 1 billion RMB level. Like many foreign hedge funds, High-Fund Quantitative, after accumulating capital, invested in fundamental research with the goal of driving social progress. In addition to providing financing for tech innovation companies, this is another way the capital market supports technological innovation.
However, DeepSeek's success was also somewhat lucky. Previously, for various reasons, regulators strengthened oversight of quantitative funds, imposing restrictions on high leverage, high-frequency trading, and short-selling mechanisms, and penalizing some funds. Recently, the scope of regulation has expanded, affecting financial institutions like banks. In the past few years, thousands of hedge funds have closed, and many financial professionals have left the industry. Fortunately, High-Fund Quantitative persisted in the face of these challenges, leading to the successful development of DeepSeek. While some regulatory measures are reasonable, whether some of the blanket measures were excessive warrants further exploration.
As we know, finance and technology are the two essential driving forces for modern countries to foster innovation; both are indispensable. The success of DeepSeek is a vivid example of financial support for innovation, and regulators should carefully nurture the healthy growth of China’s capital market to avoid excessive regulation that could hinder China’s innovation capacity.
Insight 3: DeepSeek success reflects the creativity and resilience of the private economy.
Finally, DeepSeek success once again validates the creativity and resilience of China’s private enterprises. I believe that improving the business environment for private enterprises is also key to driving China’s innovation.
Despite external challenges such as chip bans and internal difficulties such as the lack of resource support and even regulatory crackdowns on quantitative funds, High-Fund Quantitative release of DeepSeek has still become a benchmark for Chinese innovation in this era. The governance structure of private enterprises is more flexible, with higher operational efficiency and a greater capacity for innovation. Whether in the field of new energy vehicles or photovoltaic cells, Chinese private enterprises have broken through numerous barriers to achieve breakthroughs, and they represent China’s competitive level internationally. In the global competition for artificial intelligence large models, apart from DeepSeek, China’s internet platform companies have also relied on their financial advantages and data accumulation to attract talent and drive innovation, enabling them to compete with global tech giants.
Currently, private enterprises still face unfair treatment in many areas. Private enterprises not only encounter market access restrictions in certain fields, but platform tech companies are also constrained in their listings. Recently, there have even been instances of local governments "hunting" private entrepreneurs. These are areas with room for improvement.
Overall, the success of DeepSeek should be fully recognized. However, in the wake of DeepSeek’s success, it highlights the fact that there is still room for improvement in China’s economic innovation models, regulatory arrangements, and business environment for private enterprises. Looking ahead, the emergence of more DeepSeek-like innovations will push China toward becoming a truly innovative economy.