Chinese AI startups are reportedly re-evaluating their strategies as they rush to keep up with DeepSeek’s success. Companies such as Baidu, O1.ai and Zhipu adopted different business models after the R1 model breakthrough.
Chinese artificial intelligence startups are changing their business models as they struggle to remain competitive after DeepSeek’s boom. Companies such as Baichuan, O1.ai, and Zhipu revealed overhauls in their strategies over the past two months.
Chinese AI companies endorse open-source licensing models
O1.ai revealed in late 2024 that it had stopped pre-training large language models (LLMs) called Yi. The company cited the rising costs of deployment, saying that its foundational model team had been transferred to Alibaba.
O1.AI announced last week that it would offer tailored solutions for businesses to deploy DeepSeek’s AI models. Lee Kai-Fu, the company’s founder said the company had shifted its business in what he called the DeepSeek age.
The founder commented that WanZhi Enterprise’s LLM platform is different from DeepSeek’s all-in-one platform. He explained that WanZhi Enterprise’s LLM model offered integrated software-hardware solutions at lower prices than its competitors.
Kai-Fu noted that the enthusiasm around DeepSeek in China mirrored the global ChatGPT craze. She added that DeepSeek’s products were not only high-performing models but also open-source systems generating international influence.
Baichuan, a Beijing-based AI startup, revealed in February that it had dissolved the finance division of its enterprise team. It added that the move was in line with the company’s shift to AI healthcare, as it planned to launch AI-driven pediatric solutions and medical assistants. The company’s leaders announced to its employees that it would focus on developing its technology for hospitals, including an AI doctor who assisted in diagnosis.
Moonshot AI, a Chinese AI startup, reportedly decided to reduce its marketing budget for its Kimi Chatbot to focus on model training. Moonshot AI explained that the adjustment was influenced by external factors and internal strategic changes. The company recognized the high deployment costs associated with its AI chatbot, noting that DeepSeek’s R1 model reduced inference costs compared to OpenAI’s O1 model.
Industry players say DeepSeek model could reduce startups’ deployment costs
Chinese AI DeepSeek was crowned China’s AI champion after its widespread adoption weeks after its launch. The company integrated its AI models across government agencies, state-owned enterprises, and financial institutions.
Wang Tiezhen, an engineer at AI research hub Hugging Face said the Chinese LLM market was rapidly consolidating around a few leaders. He added that DeepSeek promoted many companies to redirect resources to applications rather than foundational model development. Wang said by adopting top-tier models, companies can eliminate the need to invest tens of millions of dollars annually in training inferior models.
Wei Sun, principal analyst of artificial intelligence at Counterpoint Research, said that DeepSeek’s success proved that open-source strategies could lead to faster innovation and broad adoption. She noted that the R1 model is reshaping China’s landscape, with companies such as Baidu moving to open-source their LLMs as a strategic response.
Chinese startup Zhipu launched its initial public offering shortly after DeepSeek’s success to support its growth. The company reportedly aims to focus on building its enterprise sales business, selling personalized AI applications to local governments and companies.
Three investors aware of the company’s figures said the startup was burning through cash as it boosted its enterprise sales business. They said that in 2024, Zhipu made RMB 300 million ($41 million) in sales and RMB 2 billion ($276 million) in losses.
The costs sparked concerns among investors after DeepSeek introduced a new path to building AI models with lower inference costs. Zhipu employs 800 million people on its team, while DeepSeek has a smaller workforce of 160 employees.
The company still needs approval from its regulators before it can pursue a listing on the Star Innovation Board. Zhipu revealed to its investors that it was aiming to list before the end of 2025. Investors raised concerns that the government’s adoption of DeepSeek could threaten Zhipu’s business strategy of selling tailored AI solutions to government entities.
Cryptopolitan Academy: Coming Soon - A New Way to Earn Passive Income with DeFi in 2025. Learn More
