Chinese AI startup DeepSeek has secured about 100 billion yuan ($15 billion) in its latest fundraising round, marking one of the country’s largest financings as it prepares to compete more aggressively with global leaders such as Anthropic and OpenAI, Bloomberg reported Wednesday, citing people familiar with the matter.
The Hangzhou-based company raised its target from an initial 50 billion yuan to 80 billion yuan before settling on the final amount, according to the report.
Contemporary Amperex Technology, a leading battery maker, and Tencent Holdings were among the biggest investors. The financing also comes as DeepSeek prepares for a potential IPO, following reports that the company is considering a public listing that could become a major event for Chinese equity markets.
DeepSeek traces its origins to High-Flyer’s push into artificial general intelligence research in 2023. Liang Wenfeng, who co-founded the Chinese quantitative hedge fund in 2015, oversaw the initiative as High-Flyer moved to develop AI capabilities outside its core financial business.
DeepSeek was established as an independent research group in May 2023 and incorporated in Hangzhou later that year.
The company has since gained prominence in the global AI industry in 2025 after its low-cost models demonstrated capabilities comparable with leading US systems.
Due to its limited financial resources, it had been unable to match US frontier labs, but its latest $15 billion funding round has begun to narrow the gap. OpenAI and Anthropic have each raised well over $100 billion, although their totals include different financing structures.
DeepSeek has increased spending on data-center infrastructure and open-weight model development for use in China and overseas. Its inexpensive, high-performing releases have pressured Chinese rivals while attracting international users; the September launch of V4.1 Flash was followed by share declines at MiniMax and Z.AI.
Alibaba, ByteDance, and other Chinese technology companies are also rolling out lower-cost AI services, many based on open models.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.