DeepSeek nears a 100B yuan raise at a 500B yuan valuation ahead of a 2027 IPO

DeepSeek is nearing a funding round of about 100 billion yuan, roughly $14.9 billion, according to CNBC and Reuters, citing Bloomberg News. That is double the company's original target. The round would value DeepSeek at 500 billion yuan, about $75 billion, with Tencent and battery maker CATL among the investors. DeepSeek plans to list on Shanghai's STAR Market in early 2027.

The money is earmarked for compute. DeepSeek will deploy more than 160,000 Huawei Ascend chips at a site in Inner Mongolia, a clear commitment to domestic silicon given US export controls on NVIDIA's top accelerators. The company has already partnered with Huawei on TileLang libraries to optimize deep-learning performance on Ascend hardware. Building at that scale on Chinese chips will test whether Huawei's software stack can support frontier training, not just inference.

The raise marks a turning point for a lab that famously operated on modest budgets funded by its parent hedge fund, High-Flyer, and earned attention for training competitive models cheaply. Outside capital at this scale brings the governance and disclosure obligations of a public listing. It also arrives as competition in China intensifies: Nikkei reports Chinese labs released 16 AI models in a single month despite warnings from Anthropic about distillation of its models. Meanwhile DeepSeek's V4 is spreading globally; Microsoft Foundry hosts DeepSeek-V4.1-Flash, and legal AI firm Ivo built its open-source Sage contract model on DeepSeek V4 Flash.

Caveats: the round is reported, not finalized, and terms could change. A $75 billion valuation is a fraction of OpenAI's or Anthropic's, which may reflect both lower revenue and geopolitical discounting. US scrutiny of Chinese models in Western enterprises, as seen in debates over AWS hosting Zhipu's GLM-5.3, could limit DeepSeek's international growth. Watch for the final close, IPO filing details, and the performance of the Ascend cluster.