DeepSeek Lines Up a $12 Billion Round Ahead of a 2027 IPO
3 min readDeepSeek, the Hangzhou lab that rattled Silicon Valley with its low-cost R1 model in early 2025, is closing in on a DeepSeek funding round of at least 80 billion yuan, about $12 billion, Bloomberg reported on October 6. Tencent and battery giant CATL are among the largest backers, and CNBC reports the company is weighing whether to push the round to 100 billion yuan, roughly $15 billion.
From self-funded to oversubscribed
For most of its short life DeepSeek took no outside money at all. It was bankrolled by founder Liang Wenfeng’s quantitative hedge fund, High-Flyer, and turned away China’s biggest venture firms. That changed this year. Its first external round, completed around June, brought in about $7.4 billion from investors including Tencent and CATL at a valuation reported between $52 billion and $59 billion. When the current raise began in July the target was 50 billion yuan at a reported valuation of 500 billion yuan, about $74 billion, according to Tech Startups.
What changed
Demand came in roughly 60 percent above target. People familiar with the deal tie the surge to the September release of DeepSeek-V4.1-Flash, a model built for higher throughput and cheaper inference, and to a business that reached a $1 billion annualized revenue run rate in late September after raising its API prices. The round also sets up what could be the most watched Chinese listing in years: DeepSeek has hired CITIC Securities to prepare an IPO, potentially on Shanghai’s STAR Market, in early 2027. Timing, venue, size and valuation are not final, and DeepSeek, Tencent and CATL have not commented.
Why it matters
Twelve billion dollars is a scale of capital usually reserved for US frontier labs, and this is Chinese money for a Chinese lab that cannot buy Nvidia’s best chips. The investor mix says something too. Tencent is a platform company with obvious uses for the models; CATL is a battery maker. Industrial capital is moving into AI in a way that matches Beijing’s push to treat model building as strategic infrastructure.
It also changes the open-weight race. Mistral claimed this week that Large 4 is the strongest open model from the US or Europe while conceding that Chinese open models still lead the broad benchmarks, and South Korea’s new $3.5 billion frontier program explicitly targets parity with Chinese open models, not US labs. A DeepSeek with $12 billion to $15 billion in the bank keeps that bar moving.
What to watch: whether DeepSeek keeps releasing open weights once it answers to public shareholders, and whether the STAR Market listing lands on schedule.