China’s AI Revolution: Beyond DeepSeek
In early 2025, the tech world reacted to the release of DeepSeek-R1, an open-source large language model (LLM) from a Chinese company, with significant attention. DeepSeek emerged as a strong competitor, even surpassing the capabilities of leading LLMs like GPT-4, Llama 3.1, and Claude, while requiring less training time and data. It also offered a more attractive price point.
How did DeepSeek, a company founded only in May 2023, achieve such rapid success? Is this a temporary phenomenon, or does it signal broader trends in China’s AI development? Following the launch of DeepSeek-R1, Chinese tech giants like Tencent, Alibaba, and ByteDance announced their own LLMs, each claiming to surpass DeepSeek-R1 in performance. This raises the question: Could China be entering an era of accelerated AI development? This article examines how China has systematically built the core components of its AI ecosystem, setting the stage for significant advances.
Investing in Research: The Foundation of AI
China’s commitment to AI development is deeply rooted in substantial investment in both academic and applied research. Driven by ambitious government goals, China aims to reduce the gap that separates them from the United States. Since the 2000s, China has significantly increased financial commitments in this field. Top Chinese universities and numerous local institutions have expanded their programs to support AI research.
Government venture funds have played a crucial role, with their investments exceeding US$912 billion over the past two decades. A significant 23% of this amount was specifically allocated to AI-related firms across the country. As a result, China’s AI paper publications and patent filings have exceeded those of the United States since the 2010s. The World Intellectual Property Organisation reported that between 2014 and 2023, Chinese-led AI patent filings surpassed those of the US by a factor of six. Despite these numbers, citations of Chinese research are often less frequent than American ones. However, the implementation of cloud technology, which provides large data sets, has sped up algorithm development. China’s AI and digital ecosystem growth has been driven by strategic investment and resource allocation, which are key for maintaining a competitive advantage in the long term.
Cultivating Talent: The Human Element
Beyond research, talent is another critical component of AI development. In the past decade, Chinese tech companies have invested heavily in training and hiring programmers and data scientists for e-commerce, gaming, and marketing. China has reportedly surpassed the US in generating AI talent and plans to add another 500,000 individuals to its AI talent pool in the years to come. A recent government crackdown in the tech sector unintentionally freed up many engineers from companies like Alibaba and Tencent. This has led to a boost in the startup world. There are an estimated 1.67 million AI-related companies in China, with over 237,000 added in the first half of 2024 alone. Private investment has complemented government-backed funding, with venture capital and private equity funds turning their attention to developing new AI unicorns.
These investments, while potentially leading to some inefficiency, as seen in earlier trends, have facilitated industry growth, talent development, and increased firm success. Kaifu Li, a former executive from Microsoft and Google, is one example. He invested in the AI venture 01.ai, which quickly debuted its Yi-Lightning LLM. This matched the capabilities of models like GPT-4. Similar to DeepSeek-R1, only 2,000 GPUs and a budget of US$3 million were needed for its training, a small fraction of OpenAI’s investment in comparison.
Turning Constraints into Opportunities: Frugal Innovation
The strategy for artificial intelligence development in Silicon Valley has historically focused on advancing technologies across all fronts. Conversely, China has encountered constraints as it operates outside of the American and Western AI ecosystem. The United States government imposed an export ban on chips to China in 2022, which has gradually become more restrictive. These restrictions, coupled with the migration of top AI talents from China to the United States, present a challenge to China’s AI aspirations. However, history has shown that constraints often fuel innovation–sometimes leading to disruptions.
Several Chinese companies, including Huawei, SenseTime and Xiaomi, are actively developing AI hardware, including chips. Alibaba and Tencent have been working to encourage top AI talent to return to China. Furthermore, firms like DeepSeek have seen these constraints as opportunities to innovate. One of DeepSeek’s objectives is to create LLMs that can operate effectively on older-generation chips. Alibaba and Tencent also continue to address a critical challenge of AI: the high energy consumption of LLMs. For instance, Alibaba’s Qwen2.5-Turbo search engine operates much more efficiently than GPT-4 Turbo.
Practicing Frugal Innovation for Scaling
A key challenge lies in the scaling competition within China. With a population of over 1.4 billion, China offers an attractive market. The success of companies that prioritize innovation and reduce costs is vital to their success. Global industries have experienced the shift of wind and solar industries to low-cost, high-quality production in China, enabled by frugal innovation. AI is no exception to this trend. Rapid scaling and high competition are not without their drawbacks, as evidenced by challenges like overcapacity in China’s electric vehicle industry. However, using the frugal innovation approach to scale can be an effective strategy for achieving success both in China and globally.
Cultivating Robust Internal Competition
DeepSeek is one of many startups that have come out of intense internal competition in China’s AI sector. Interestingly, it was not even included on the 2024 Hurun China AI Enterprise Top 50 list. The list provides a glimpse into the diverse Chinese AI ecosystem. Firms operate across multiple sectors, including mobility, communications, and healthcare, with expertise ranging from hardware to data analytics and image recognition. These firms are based in different regions of China. Beijing leads the list with 20 companies, while Guangzhou and Shenzhen have a combined 12. Competition is also strong within individual industries. The mobility sector, for example, includes 11 companies focused on autonomous driving. The diversity of top AI firms in China speaks to the competitive internal market driven by the government’s vision. China has used a similar model to accelerate the growth of other nascent industries, including wind, solar, and electric vehicles. This internal competition has led to affordable, mature products that rapidly enter the global market. The American AI sector is mainly led by industry giants. Since 2015, Microsoft has set up seven industry verticals to explore AI applications with its clients, and Google has incorporated AI into its tools. Relying heavily on these major players can be a disadvantage, potentially sidelining other innovation pathways.
US stocks generally recovered from the news about DeepSeek. But the potential of China’s AI development is vast, and the next market-altering innovation could be only a matter of time. Businesses and investors should pay close attention to the growth of China’s AI ecosystem, or risk being left behind in the global AI race.
