DeepSeek has raised about 100 billion yuan (about $15 billion) in its latest fundraising round, Bloomberg News reported on Thursday, in one of the largest financings ever completed by a Chinese technology startup. The Hangzhou-based AI lab lifted its target at least twice while talking to investors, first from 50 billion yuan to 80 billion yuan, before settling on the final tally, according to people familiar with the matter cited by Bloomberg.
Battery maker Contemporary Amperex Technology Co. Ltd. (CATL) and Tencent Holdings, the operator of WeChat, were among the biggest investors in the round, the people said. The details are not public, and representatives of DeepSeek, Tencent and CATL did not respond to Bloomberg’s emailed requests for comment. Reuters, which also reported the raise on Thursday, said it could not immediately verify Bloomberg’s report, so the figures stand as reported rather than company-confirmed.
DeepSeek’s Funding Round: A Target That Doubled In Weeks
The size of the round grew quickly. Reporting by China’s Caijing Magazine in mid-October, sourced to people involved in the deal, said DeepSeek’s latest round was set to roughly double from a planned 50 billion yuan to about 100 billion yuan in under half a month, and described it as the company’s second funding round. Caijing put DeepSeek’s pre-money valuation at approximately 500 billion yuan, unchanged by the larger raise, and said new and existing shareholders each accounted for roughly half of the money.
Caijing also named a wider circle of new investors beyond the two biggest backers Bloomberg identified: Shixi Capital, SMIC Capital, CPE Yuanfeng, Legend Capital and Geely were among those joining, a list dominated by industrial capital rather than conventional venture funds. Bloomberg’s completion report on Thursday did not publish a valuation for the round.
What The Money Is For: Data Centres And Open Models
DeepSeek has begun to spend more aggressively to build out AI data centres and to develop more open-source models for use in China and abroad, Bloomberg reported. The lab has regularly rolled out high-performance, low-cost models that pressure its domestic competitors and win customers in the US and other overseas markets. In September it released V4.1-Flash-1017, which topped the DeepSWE benchmark and shipped alongside a new search API, a launch that set off a drop in the share prices of local rivals MiniMax Group and Z.AI.
The funding also sets the stage for a stock market debut. Bloomberg News has previously reported that DeepSeek, like rival Moonshot AI, is preparing for an initial public offering that would be an important milestone for Chinese markets. The lab has likewise been opening up its infrastructure: it recently open-sourced DeepInfer, the inference engine behind its production models, a move aimed at cutting the cost of serving its models at scale.
Why It Matters
DeepSeek’s raise matters most because of the gap it starts to close. The lab built its reputation on doing frontier-level work cheaply, but cheap research still runs on expensive compute, and Bloomberg notes DeepSeek has never had the financial resources of the US frontier labs — Anthropic and OpenAI have each raised more than $100 billion. A $15 billion war chest, much of it from industrial backers such as CATL, Tencent and Geely, gives DeepSeek the means to build data centres at a scale its earlier, leaner funding never allowed, and it signals that China’s largest industrial groups are willing to bankroll a national champion directly. For rivals at home, the pressure runs the other way: the “DeepSeek death zone” sketched in a widely shared Artificial Analysis benchmark chart describes competitors that can survive only by beating DeepSeek’s performance or undercutting its prices, and a better-funded DeepSeek makes both harder. For the open-model ecosystem that DeepSeek’s releases anchor, the raise promises more of the low-cost models that have repeatedly reset pricing across the industry.
Background
DeepSeek is the Hangzhou-based lab that triggered a roughly $1 trillion meltdown in global markets in early 2025 by showing it could compete with Silicon Valley’s leaders at a fraction of their cost, a shock that made its open-weight, low-price approach the reference point for China’s AI sector. Until now it has operated with far less outside capital than the US labs it is measured against; this round, and the IPO preparation Bloomberg has reported, mark its shift from a lean research outfit to a heavily capitalised contender.
0 Comments