A Strategic Retreat from the AGI Dream
Once celebrated as China’s “Six AI Tigers,” these tech startups, initially expected to lead the nation’s rise in generative artificial intelligence (AI), are now undergoing a significant strategic shift. Instead of pursuing the expensive development of versatile, foundational Large Language Models (LLMs), they are now focusing on survival by narrowing their focus to niche markets and specific applications.
Baichuan, co-founded by Wang Xiaochuan, former CEO of Sogou, provides a telling example. On its recent second anniversary, the company announced a directional change, emphasizing the reduction of redundant operations and concentrating resources on AI in healthcare. This contrasts sharply with its original vision of becoming China’s equivalent of OpenAI.
Similarly, Zero One (01.AI), founded by Kai-Fu Lee, former president of Google China, has announced a “small but exquisite” strategy, abandoning its initial ambitions to build an AI 2.0 platform and accelerate the advent of Artificial General Intelligence (AGI).
Other “tigers” such as MiniMax are scaling back their enterprise (B2B) business to focus on overseas markets with AI video creation applications. Meanwhile, Zhipu AI, Moonshot AI (Yuezhi Anmian), and Character AI (Chinese version), while remaining active in the open-source community, are also shifting their focus to the B2B SaaS (Software as a Service) market. Some experts view this area as “the least innovative” within the AI industry.
The DeepSeek “Shock” and the Cost Burden
This synchronized strategic shift is no accident. According to tech expert Wang Wenguang, many Chinese AI companies had already abandoned independently training LLMs due to high investment costs and difficulties competing in terms of economic efficiency.
However, the real turning point came in January 2025, when startup DeepSeek launched its DeepSeek R1 model. This model caused a sensation with its powerful performance and superior cost-effectiveness (possibly due to proprietary optimization techniques). The emergence of DeepSeek R1 made most remaining AI startups, especially small and medium-sized enterprises, realize they could not keep up with the race for foundational models in terms of technology and cost. This “shock” directly prompted the “Six AI Tigers” to collectively seek new directions.
Navigating the Niche Market Challenges
The shift to markets such as AI healthcare or B2B SaaS does not guarantee easy success. The B2B SaaS market, while considered more realistic, also faces fierce competition. Wang Wenguang points out that the technical barrier to building a basic LLM platform is not too high (“It took me about half a year to develop one myself”). There are thousands of similar platforms on the market, which are easily copied, leading to price wars (Wang provides B2B services for only 40,000-50,000 RMB, a price point that large companies find difficult to compete with).
In this context, data becomes the decisive factor for competitive advantage. “To create a competitive advantage, the determining factor is what data you possess, because everyone can use the model,” emphasizes Alibaba expert Gao Peng. This is precisely the weakness of startups compared to giants like Alibaba, ByteDance, or companies that have accumulated specialized data like DeepSeek.
The Future of AI in China
Experts, including Kai-Fu Lee, seem to agree that the race to build foundational AI models in China will soon consolidate to only three main players: DeepSeek, Alibaba, and ByteDance. “Startups that continue to pursue LLM technology will all fail,” predicts expert Jiang Shao, who forecasts that the leader (possibly DeepSeek) will capture 50-80% of the market share. The ultimate race will be to see who achieves AGI first.
DeepSeek is currently highly regarded for its talented technical team, ideals, and abundant resources. Wang Wenguang even suggests that DeepSeek could rise to the number one position globally if it decides to commercialize more aggressively.
For the once-promising “Six AI Tigers,” the future looks less bright. Abandoning the LLM foundational race and seeking survival in niche or competitive B2B markets reveals a stark reality: without a significant advantage in data or truly differentiated core technology, creating miracles in the AI industry is extremely difficult. The initial dream of ‘defeating the United States and surpassing OpenAI’ is gradually giving way to the more realistic goal of survival and development in a volatile AI market.
Deep Dive into the Strategic Shifts
The initial euphoria surrounding the “Six AI Tigers” painted a picture of rapid innovation and disruption. Venture capital flowed freely, fueling ambitious projects aimed at creating general-purpose AI models capable of rivaling those developed by OpenAI and other global leaders. However, the realities of the market and the immense resources required to compete at that level soon became apparent.
Baichuan’s pivot to AI in healthcare is a prime example of this strategic recalibration. The healthcare sector presents a wealth of opportunities for AI applications, from improving diagnostics and treatment planning to accelerating drug discovery and personalizing patient care. By focusing on this specific domain, Baichuan can leverage its expertise to develop targeted solutions that address real-world challenges, potentially generating revenue and establishing a sustainable business model. This specialization allows them to become leaders in a defined area rather than spread resources thinly across a broad and competitive landscape.
Zero One’s shift towards “small but exquisite” solutions reflects a similar recognition of the need to focus on areas where the company can excel. Rather than attempting to build a broad AI platform that can do everything, Zero One is concentrating on developing specialized tools and applications that cater to specific needs. This approach allows the company to leverage its strengths and avoid spreading its resources too thin. By developing focused, high-quality applications, Zero One aims to provide significant value in targeted areas, differentiating itself from the pursuit of generalized AI.
MiniMax’s focus on overseas markets with AI video creation applications is another example of a strategic pivot aimed at finding a niche where the company can thrive. The demand for video content is growing rapidly, and AI-powered tools can help creators produce high-quality videos more efficiently. By targeting overseas markets, MiniMax can tap into new opportunities and diversify its revenue streams. This proactive approach allows MiniMax to navigate the competitive AI landscape by expanding its reach and catering to the specific needs of international markets.
Zhipu AI, Moonshot AI, and Character AI’s shift towards B2B SaaS reflects a desire to find a more stable and predictable source of revenue. The B2B SaaS market offers the potential for recurring revenue streams and long-term customer relationships. However, competition in this market is fierce, and companies need to offer compelling solutions that address specific business needs. To succeed in this competitive environment, these companies will need to demonstrate tangible value and build strong relationships with their clients, ensuring that their AI-powered solutions directly address business challenges.
Understanding the DeepSeek Advantage
DeepSeek’s emergence as a potential leader in the Chinese AI landscape is due to several factors. First, the company has assembled a highly talented technical team with expertise in areas such as deep learning, natural language processing, and computer vision. This team’s expertise forms the foundation for innovative AI solutions that can compete on a global scale. Second, DeepSeek has a strong commitment to research and development, constantly pushing the boundaries of AI technology. By investing in cutting-edge research, DeepSeek stays at the forefront of AI advancements, allowing them to develop state-of-the-art models and applications. Third, the company has access to significant resources, allowing it to invest in the infrastructure and talent needed to compete at the highest level. These resources enable DeepSeek to scale its operations, acquire top talent, and build the infrastructure required to support its ambitious AI endeavors. Finally, DeepSeek has a clear vision for the future of AI, focusing on developing general-purpose AI models that can solve a wide range of problems. This strategic vision guides DeepSeek’s research and development efforts, ensuring that the company remains focused on its long-term goals.
DeepSeek’s DeepSeek R1 model is a testament to the company’s technical prowess. The model is saidto be highly efficient and cost-effective, allowing DeepSeek to compete with larger players in the market. The model’s performance has impressed many industry observers, and it is expected to play a significant role in shaping the future of AI in China. The DeepSeek R1 model serves as a benchmark for the company’s capabilities and highlights its potential to disrupt the AI landscape.
Data: The New Oil in the AI Era
In the AI era, data is the new oil. Companies that have access to large and high-quality datasets have a significant advantage over those that do not. Data is essential for training AI models, and the more data a model is trained on, the better it is likely to perform. Access to diverse and comprehensive datasets allows AI models to learn more effectively, leading to improved accuracy, efficiency, and overall performance.
However, not all data is created equal. High-quality data is data that is accurate, complete, and relevant to the task at hand. Specialized data is data that is specific to a particular domain or industry. Companies that have access to specialized data are better positioned to develop AI solutions that address specific needs. The ability to leverage high-quality and specialized data gives companies a competitive edge in developing AI solutions that are tailored to specific industries and challenges.
This is why companies like Alibaba, ByteDance, and DeepSeek have a significant advantage over smaller AI startups. These companies have access to vast amounts of data, which they can use to train their AI models and develop innovative solutions. Their ability to access and process vast quantities of data allows them to build more powerful and accurate AI models, which in turn enables them to create innovative solutions that address a wide range of challenges.
The Road Ahead for China’s AI Startups
The strategic shifts taking place among China’s AI startups reflect a recognition of the need to adapt to the realities of the market. The race to build general-purpose AI models is expensive and competitive, and many startups have realized that they cannot compete at that level.
Instead, these startups are focusing on finding niche markets and developing specialized solutions that address specific needs. This approach allows them to leverage their expertise and avoid spreading their resources too thin. By focusing on specialized solutions, these startups can leverage their unique skills and knowledge to create innovative solutions that meet the specific needs of their target markets.
However, the road ahead for China’s AI startups is not without challenges. Competition is fierce, and companies need to offer compelling solutions that are both innovative and cost-effective. They also need to be able to attract and retain top talent, which is in high demand in the AI industry. Overcoming these challenges requires a strategic approach that emphasizes innovation, cost-effectiveness, and talent acquisition, ensuring that these startups can thrive in a highly competitive market.
Despite these challenges, there is still reason for optimism. China has a large and growing AI market, and there is plenty of room for innovation and growth. Startups that are able to adapt to the changing landscape and develop innovative solutions have the potential to thrive. The potential for growth in the Chinese AI market provides opportunities for startups that can adapt to the evolving landscape and develop innovative solutions that meet the needs of this dynamic market.
The Broader Implications
The scaling back of ambitions among some of China’s AI startups has broader implications for the global AI landscape. It suggests that the race to build general-purpose AI models is becoming increasingly concentrated among a small number of large players. This could lead to a reduction in competition and innovation, as smaller players find it increasingly difficult to compete. The concentration of resources and expertise among a few large players could limit the diversity of AI solutions and potentially stifle innovation in certain areas.
However, it also suggests that there is still plenty of opportunity for specialized AI solutions that address specific needs. Startups that are able to identify these opportunities and develop innovative solutions can still thrive, even in a market dominated by large players. The focus on specialized solutions provides opportunities for startups to differentiate themselves and address unmet needs, ensuring that innovation continues to thrive in the AI landscape.
The future of AI in China, and around the world, is likely to be shaped by a combination of factors, including technological innovation, market dynamics, and government policies. It remains to be seen how these factors will play out, but one thing is clear: the AI revolution is still in its early stages, and there is plenty of opportunity for both established players and startups to make a significant impact. The interplay of these factors will determine the trajectory of AI development and the opportunities available for both established companies and emerging startups.