Chinese artificial intelligence (AI) phenomenon DeepSeek is currently experiencing significant popularity, with users across China eagerly engaging with its chatbot.
During the recent eight-day Chinese New Year holiday, which concluded on February 4, many individuals turned to the start-up’s high-performance, complimentary chatbot, inputting their birth data—referred to as “bazi” or eight characters—to receive insights on matters of love, life, and financial prosperity.
Users took to social media to share their experiences, highlighting how the DeepSeek-R1 reasoning model could engage in human-like dialogues, suggest fitness routines, and even compose poetry. Additionally, some individuals utilized DeepSeek’s various AI models to prepare for the fiercely competitive civil service exams in China.
DeepSeek disrupts the status quo by providing free access to advanced AI technology for everyday users, thereby enhancing their confidence in utilizing these tools, according to Mr. Li Ruochong, a 23-year-old Master's student in AI at the Hong Kong University of Science and Technology (Guangzhou), in an interview with The Straits Times.
The phrase “po quan,” meaning breaking the circle, is commonly used online to refer to phenomena that have achieved widespread popularity beyond their initial niche audiences.
However, DeepSeek’s chatbot is not the first generative AI model to emerge in China. The start-up was established in 2023, coinciding with the launches of AI models by major Chinese tech companies, including Alibaba’s open-source Qwen, Baidu’s Ernie Bot, and ByteDance’s Doubao.
DeepSeek gained both domestic and international recognition due to its V3 large language model (LLM) and R1 reasoning model, which were introduced in the past two months. These models have demonstrated performance on par with leading global counterparts like the US’ ChatGPT, yet were developed at a significantly lower cost and without the most advanced chips.
DeepSeek's recent success may serve as a catalyst for accelerating the development of more advanced and practical AI models in China, potentially drawing increased investment and talent to the sector.
Competitors such as Alibaba, ByteDance, and Moonshot AI have already introduced updates to their models, posing significant competition to both DeepSeek and OpenAI from the United States.
The focus is now on whether China can maintain the momentum of these advancements to bolster its goal of becoming the global leader in AI by 2030.
The Competitive Landscape in China
Since the launch of OpenAI's ChatGPT, which popularized generative AI technology in late 2022, China has been striving to catch up. Numerous companies have been racing to create their own models, supported by government initiatives in the field.
The Chinese government has made substantial investments in AI firms through various funding mechanisms, including a national AI industry investment fund of 60 billion yuan (approximately S$11 billion) announced in January 2025. A US study covering the years 2000 to 2023 revealed that Chinese government venture capital funds invested around US$184 billion (S$249 billion) in over 9,600 AI companies.
Additionally, the government has established AI infrastructure, including data centers and computing clusters.
Currently, nine out of the top 20 positions in Chatbot Arena, a global benchmarking and ranking platform for large language models (LLMs), are held by Chinese firms, while 11 are occupied by US companies.
According to Rui Ma, a technology analyst and author of the Tech Buzz China newsletter, benchmarking against the latest frontier models reveals that Chinese models have significantly reduced the performance gap compared to leading international models, including those developed in the United States.
She notes the substantial progress made in China since the 2022 launch of ChatGPT, highlighting the growth from a near absence of comparable models and researchers to a robust and highly competitive talent pool actively engaged in cutting-edge large language model development.
The Chinese government's initiatives to support and finance the generative AI sector appear to have significantly narrowed the gap between China and Western nations in this field.
In China's large language model (LLM) start-up landscape, a group known as the "six little tigers" has emerged, comprising Zhipu AI, MiniMax, Baichuan AI, Moonshot AI, StepFun, and 01.ai. The founders of these companies often have backgrounds in academia or have previously worked for major tech firms like Google and Microsoft. Each of these start-ups is valued at over US$1 billion and secures funding from a combination of private venture capital, larger corporations, and occasionally state-supported organizations.
Challenging the Norm
However, the standout AI model has come from DeepSeek, which some argue achieved success not due to the government's innovation framework but rather in spite of it.
DeepSeek is regarded as an anomaly within the conventional ecosystem. It functions more like a passion-driven endeavor led by a skilled and youthful team, prioritizing innovation over commercialization and free from the profit-driven pressures that larger firms encounter.
Mr. J.S. Tan, a PhD candidate at the Massachusetts Institute of Technology focusing on innovation policies in China, highlighted on the media platform Substack that DeepSeek did not depend on state-sponsored programs or investments from established tech companies.
Instead, its founder, Liang Wenfeng, personally financed the company's operations, as noted by Mr. Tan. Mr. Liang became a billionaire after founding the hedge fund High-Flyer in 2015.
“This allowed it to operate independently of the constraints often associated with state or corporate funding,” Mr. Tan remarked in his January 29 post.
When asked why DeepSeek does not pursue both model development and potential applications simultaneously, Mr. Liang stated in a July 2024 interview with Chinese media that he believes China should transition from being a technology consumer to a contributor as its economy evolves. “Our goal is not to turn a profit quickly, but to drive the technological frontier and fundamentally promote the growth of the entire ecosystem,” he explained.
Can the success of DeepSeek be replicated?
DeepSeek's achievements have inspired Chinese AI firms, as its foundation was partially built on earlier LLM developments from China, including Alibaba's open-source Qwen, according to AI researcher Neil Zhu.
However, Mr. Zhu, who is also the founder of the Centre for Safe AGI, a Shanghai-based non-profit focused on the safe deployment of artificial general intelligence, noted that replicating DeepSeek's effective organizational structure may prove challenging for other AI companies in China. This structure was instrumental in facilitating significant breakthroughs.
He highlighted the company's agile and flat organizational model, along with Mr. Liang's proactive involvement in research.
Additionally, the start-up's ability to adopt a long-term research strategy was bolstered by its founder's resources, and it possessed greater computing power than many competitors in the sector.
Nonetheless, DeepSeek's success creates pressure on state-funded entities to evolve and innovate, while also paving the way for new opportunities in collaboration and investment, remarked Professor James Pang, who teaches AI and digital transformation at the NUS Business School.
He emphasized that this dynamic environment is likely to accelerate AI advancement in China, enhancing both its domestic industry and global competitiveness.