DeepSeek's second funding round raises 105.5 billion yuan, valuation reaches 615.5 billion yuan
According to foreign media citing sources, DeepSeek's second funding round raised 105.5 billion yuan (equivalent to 15.8 billion US dollars), with nearly 30 investors contributing jointly, at a valuation of 615.5 billion yuan. Industrial investors include ChangXin Memory Technologies, SMIC, Tencent, Meituan, CATL and others, each investing 1 billion to 3 billion yuan. The original article also noted that in the first funding round, investors other than the National AI Fund were required to give up voting rights and lock up their shares for five years, giving Liang Wenfeng 100% of voting rights.
Why read this
AI score 76
Beyond the funding size and the list of investors, the original article also details how Liang Wenfeng maintained control through voting rights arrangements, offering a window into the funding structures of domestic large-model companies.
Topics
Body · AI translation
On October 30, according to foreign media citing sources, DeepSeek's second round of financing raised 105.5 billion yuan (equivalent to 15.8 billion US dollars), with nearly 30 investors contributing jointly, of which state-owned capital accounted for as much as 27%, venture capital 26.5%, and technology companies 41.9%, making it the largest venture investment in a Chinese AI company to date, and DeepSeek's valuation thereby reached 615.5 billion yuan.
It is worth noting that the list of industrial investors in the second round of financing reads like a map of AI infrastructure: upstream are ChangXin Memory Technologies and SMIC (memory chips), Geely, Midea, Luxshare Precision, and vivo (manufacturing and terminals), CATL (energy), as well as Tencent, Meituan, Xiaohongshu, and Ctrip (computing power and application scenarios). From chips to manufacturing to application scenarios, these technology companies each contributed 1 billion to 3 billion yuan.
Before this, in the more than two years since DeepSeek burst onto the scene, it had consistently refused external financing.
However, accepting external financing does not mean DeepSeek will thereby dilute too much equity.
At the end of June this year, DeepSeek's first round of financing had an extremely special arrangement: apart from the national AI fund, all other investors were required to put their funds into an entity held by Liang Wenfeng, give up voting rights, and lock up their shares for five years. At the same time, business registration information shows that before the financing, Liang Wenfeng had increased his personal shareholding from 1% to 34%, with voting rights reaching 100%. Together with subsequent moves such as preparations for a STAR Market listing and the appointment of a former Hillhouse partner as CFO, this is a carefully designed plan to raise funds without losing control.
DeepSeek's first round of financing exceeded 50 billion yuan. Foreign media noted that the reason DeepSeek doubled the size of its latest round of financing lies in two aspects:
First, the soaring prices of upstream Nvidia chips. It is reported that the price of a server equipped with eight B300 chips soared from 15 million yuan to about 19 million yuan, more than five times the price of similar servers in most countries.
However, judging from a series of moves by DeepSeek after it launched the relevant financing over the past six months, a lack of money may not be the problem DeepSeek is most likely to encounter next; more important is how to cope with computing power constraints while pushing forward a breakthrough in the bottleneck of domestic computing power production capacity.
Take domestic GPU maker Moore Threads as an example. In its recently announced third-quarter financial report, it achieved quarterly revenue of 1.55 billion yuan, a year-on-year surge of 1773.9%, and net profit attributable to shareholders of 49.26 million yuan, turning profitable ahead of its original target. The financial report pointed out that its intelligent computing products are being deployed at an accelerating pace among AI innovative enterprises and have achieved Day-0 support for mainstream open-source models such as DeepSeek, GLM, and MiniMax.
In the first seven months of 2026, DeepSeek's revenue was about 475 million yuan, with a gross margin of 44.6%, but its net loss still exceeded 700 million yuan. At the same time, it took only about three months for annualized revenue to double from less than 500 million US dollars to 1 billion US dollars. But in the first seven months of this year, DeepSeek's spending related to AI infra also reached 11 billion yuan. Like many large-model vendors in the industry, DeepSeek is still in a stage where its API business increases revenue but not profit.
Combined with recent differing judgments among domestic industry insiders: one side will pay more attention to DeepSeek's long-term strategic value and whether it can build a solid computing power infrastructure and secure a position in domestic computing power, while the other side will also have doubts that, like many model vendors, DeepSeek likewise has an insufficiently clear commercialization path and that AI investment may involve problems such as self-circulation.
Now that DeepSeek's financing story has reached this point, the real test has only just begun. (This article was first published in the TMTPost APP, Text | TechHorizon)