DeepSeek, a three-year-old Chinese startup developing low-cost open-source artificial intelligence models, has raised $7.4 billion in the largest private AI financing in Chinese history, valuing the company at more than $50 billion. Founder Liang Wenfeng is now preparing for a possible initial public offering in 2027 that could bring in billions of dollars in additional funding. The financing gives DeepSeek more resources to develop advanced AI services and offer them globally at prices substantially below those of US competitors such as OpenAI, the developer of proprietary AI models, and Anthropic, the AI company behind the Claude model family. Artificial Analysis estimated that completing a standardized intelligence task with DeepSeek's V4 Flash model costs about 2 cents, compared with $2.75 using Anthropic's Claude Fable 5 model. This significant pricing difference could increase pressure on US AI developers as corporate customers become more focused on controlling costs and generating returns from their AI investments.
Companies are already shifting workloads toward cheaper Chinese models. Lindy AI, a San Francisco-based company using AI services, moved from Anthropic's Claude Sonnet to DeepSeek after a six-week evaluation and is now paying approximately 10% of its previous cost, reportedly saving millions of dollars each year. TinyFish, a Palo Alto startup building enterprise infrastructure for AI web agents, reduced its inference costs by 90% after replacing frontier models with cheaper alternatives that included Chinese open-weight models. Chinese open-source providers have accounted for more than 30% of activity on the AI aggregation platform OpenRouter during some weeks this year, compared with less than 2% in 2024. Airbnb NASDAQ:ABNB, an online accommodation platform, is using a combination of Chinese open-source models and Western providers, while Pinterest NYSE:PINS, a visual discovery company, has highlighted the usefulness of Chinese models. Microsoft NASDAQ:MSFT, a major software company and one of OpenAI's largest supporters, is also considering using DeepSeek models in its Copilot Cowork product, suggesting that lower-cost Chinese technology is gaining attention even among companies closely connected to leading US AI developers.
DeepSeek's investor group also appears to connect the company with several important parts of China's broader AI infrastructure. Tencent Holdings (TCEHY), a Chinese technology company, and JD.com NASDAQ:JD, a Chinese e-commerce company, invested billions of yuan through a limited partnership controlled by Liang, accepting five-year lockups and no voting rights. China's National AI Industry Investment Fund received direct equity, voting rights and the freedom to sell its investment, making it the only investor exempted from those restrictions. Contemporary Amperex Technology, a major battery producer, is positioned within the energy layer supporting AI development, while Huawei Technologies, a Chinese technology company producing Ascend processors, competes with AI accelerators from Nvidia NASDAQ:NVDA, a semiconductor company focused on computing chips. China added 543 gigawatts of generating capacity in 2025, more than the total capacity of any country other than China and the US, potentially supporting domestic data centers and a growing market for AI tokens. For investors, DeepSeek's expanded financial resources, aggressive pricing and links to China's energy, semiconductor and technology ecosystem may accelerate global adoption of Chinese AI models while increasing competitive pressure on more expensive proprietary providers.