←News detail

36Kr · AIUpdated Oct 24 12:41

DeepSeek's hundred-billion financing nears completion, allies with industrial capital including Perfect World and SMIC Capital and bets on embodied intelligence

AI summary

According to Bloomberg, DeepSeek's second funding round is locked at about 100 billion yuan, with a pre-money valuation of about 500 billion yuan; the fundraising target was raised twice from the initial 50 billion yuan and then doubled. Existing shareholders Tencent and CATL each added about 10 billion yuan; new entrants include industrial capital such as Shixi Capital, SMIC Capital, CPE Yuanfeng, Legend Capital and Geely. Perfect World, Century Huatong and Litong Electronics participate indirectly through two special funds, Lisi Xingshen and Shixiang Shangshi; DeepSeek is also preparing for a STAR Market IPO and has already invested in embodied intelligence companies such as Unitree.

Why read this

SelectedAI score 82

The original text sorts out the investor structure and external fund channels of the hundred-billion financing, and gives three investment clues for the embodied intelligence layout.

Topics

Tags

Body · AI translation

It has been less than four months since DeepSeek's first funding round, which set an industry record at 51 billion yuan, closed—and the leading large-model company's second mega-round is already nearing the end of its closing process.

According to the latest Bloomberg report today (October 22), DeepSeek's current round has been locked at approximately 100 billion yuan, with a pre-money valuation held at about 500 billion yuan. People familiar with the matter said the deal underwent dramatic dynamic adjustments as it progressed: the initial target was 50 billion yuan, but after jockeying among industrial capital and existing shareholders, the target was raised twice, ultimately doubling the raise to the hundred-billion-yuan level.

In terms of how the funds are split, contributions from new and existing shareholders are roughly evenly divided. Existing shareholders Tencent and CATL each added about 10 billion yuan, becoming cornerstone investors in this round; the new entrants lean entirely toward the hard-tech camp, including Shixi Capital, SMIC Capital (Zhongxin Juyuan), CPE Yuanfeng, Legend Capital, and auto giant Geely.

If this hundred-billion-yuan round is added to the 51 billion yuan first round finalized by Liang Wenfeng and core institutions at the end of June, DeepSeek has absorbed more than 150 billion yuan in capital over a span of just over a hundred days, setting a new global record for the speed of capital intake in the artificial intelligence sector.

External liquidity is accelerating its entry

English text is being prepared.

Recently, Perfect World (002624.SZ) disclosed a series of investment progress announcements, stating that two key external special-purpose funds have completed fundraising in full: the "Tianjin Lisi Xingshen" fund reached 7.52 billion yuan, and the "Shixiang Shangshi" fund reached 3.65 billion yuan, with the two funds together absorbing 11.17 billion yuan in external capital.

Behind these two large special-purpose funds, the presence of several A-share listed companies has emerged. Perfect World subscribed 50 million yuan and 100 million yuan respectively, completing a dual-channel bet with a total contribution of 150 million yuan; Century Huatong, through its wholly owned subsidiary, subscribed 650 million yuan to Shixiang Shangshi; and computing power service provider Litong Electronics previously announced a contribution of 95 million yuan to subscribe for a share of Lisi Xingshen.

Meanwhile, information published by the Asset Management Association of China shows that Lisi Xingshen completed private fund filing on September 15, 2026, with the filing code SEX397. People familiar with the matter said that both Lisi Xingshen and Shixiang Shangshi are external asset-management channels for DeepSeek's current funding round, and with the relevant funds successively completing paid-in capital and closing their books, the channels through which external social capital flows into DeepSeek's core asset pool are essentially locked in.

The reason "Lisi Capital (the Lisi-affiliated capital)" deserves particular attention is that in DeepSeek's previous funding round, it had already become the core channel through which external SPVs (Special Purpose Vehicles/project-specific funds) took heavy positions in DeepSeek.

This hard-tech investment institution, founded in 2021 by former Sequoia China partner Cao Xi and former Boyu Capital partner Tim Wang, invests across AI large models, domestic GPUs, robotics, and AI infrastructure, with Moonshot AI, DeepSeek, and Unitree Robotics all appearing in its portfolio.

In May this year, Lisi Capital initiated and established the 2.9 billion yuan "Tianjin Lisi Xingling" fund, then subscribed for a partnership share in the holding vehicle "Hangzhou Chengli," ultimately indirectly holding approximately 0.8265% of DeepSeek's equity on a look-through basis. Now, the newly established "Lisi Xingshen" has surged to 7.52 billion yuan, nearly tripling in size, once again confirming its mature operation as a super-channel connecting the liquidity of listed companies in the secondary market with top-tier AI assets.

An even more critical capitalization signal is also emerging in parallel. Bloomberg, citing people familiar with the matter, reported that while advancing this hundred-billion-yuan round, DeepSeek is simultaneously preparing for an IPO on the STAR Market, planning to submit its listing application as early as the end of this year and aiming for an official listing in 2027. Former Hillhouse Venture Capital partner Yan Wentao has taken up the post of CFO, filling in a professional operator for this capital operation involving hundred-billion-yuan fund allocation and multi-layer look-through regulation.

Aligning with gaming, computing power, and chip industrial capital

The entry of Perfect World, Century Huatong, and Litong Electronics makes the industrial synergy of DeepSeek's current round more concrete. The two gaming leaders, Perfect World and Century Huatong, together subscribed 800 million yuan to the relevant special-purpose funds, while computing power service provider Litong Electronics subscribed 95 million yuan. The three listed companies have different businesses, yet they clearly target core links such as next-generation game content generation and computing power infrastructure.

In August this year, DeepSeek launched a round of high-salary recruitment in game development and 3D rendering, drawing widespread attention in the gaming industry. The positions included: Code Agent data engineer (game development direction) and Web 3D and rendering direction, requiring candidates to "not necessarily be proficient in model training, but to have actually made games," with emphasis on game developers' engineering methods, complex interaction, and years of accumulated industry R&D experience.

From the broader list of investors, the participation of semiconductor industry capital such as Shixi Capital and SMIC Capital injected a deeper hard-tech foundation into this funding round. Previously, DeepSeek's architecture optimization under limited computing power conditions was widely recognized by the industry, especially its close cooperation and adaptation optimization around domestic computing power such as Huawei Ascend.

With the entry of core semiconductor capital such as the SMIC group, its future moves in underlying adaptation for domestic chips, the building of customized computing power clusters, and even self-developed chips are worth continued tracking.

Overall, DeepSeek's sources of capital are further expanding horizontally from the original major internet companies and purely financial investment institutions to a broad camp of real-economy industries such as semiconductor manufacturing, the automotive industry, and energy hardware.

Embodied intelligence becomes a new variable

In addition to changes in the investor structure, founder Liang Wenfeng's recent intensive moves into embodied intelligence also reveal a new strategic direction for this super funding round.

Previously, multiple media outlets reported that Liang Wenfeng recently exchanged views with several embodied intelligence entrepreneurs, discussing directions such as robots, large models, and interaction with the real world. People familiar with the matter said DeepSeek is considering whether to further enter this field, with possible paths including building its own team for R&D or entering through investment in related companies.

On the investment side, DeepSeek has already bet on three major embodied intelligence companies.

Unitree Robotics (disclosed): DeepSeek has previously invested about 141 million yuan in the company, entering robot hardware, motion control, and general development and testing resources. The two sides also plan to jointly develop robot AI models, and Unitree provides DeepSeek with a cooperation foundation in body and motion control for exploring robot applications.

Winverse AI (planned lead investment): founded by former Tencent AI scientist Hu Han, DeepSeek plans to lead its seed round (amounting to tens of millions of dollars), with a post-investment valuation of about $400 million. Meituan's Dragon Ball Investment is also among the participants.

Yu Tao's startup (name/entity to be disclosed): the new company of Yu Tao, a former member of Tencent AI Lab's Juewu team and head of humanoid robot foundation models at LimX Dynamics, focuses on combining large models (LLM) with embodied intelligence, with valuations of about $70 million and $200 million in two rounds respectively. Media recently reported that it has received investment from DeepSeek and top-tier funds.

Embodied intelligence involves real-world data interaction, multi-physics simulation, and high-precision motion control, and places extremely high continuous demands on computing power resources, engineering testing environments, and hardware bodies.

This explains to some extent why DeepSeek's current funding round was significantly raised to the hundred-billion-yuan level in a short period, and also reveals its strategic ambition: to formally expand from model inference of virtual code in the cloud toward real intelligent entities with physical interaction capabilities.

This article is from the WeChat public account "AI Frontline" (ID: ai-front), author: Siyue, published by 36Kr with authorization.

Source

36Kr · AI