DeepSeek Secures ~100B Yuan Funding at 500B Valuation to Bet on Embodied AI

DeepSeek has completed two rounds of massive fundraising within four months, with the second round reaching 100 billion RMB at a pre-money valuation of 500 billion RMB. Existing shareholders Tencent and CATL each added approximately 10 billion RMB, while new investors focused on hard tech and the automotive supply chain. During the same period, Perfect World, an A-share listed company, raised a total of 11.17 billion RMB through two AI-specific funds under its umbrella.

The pace of fundraising and capital structure highlight how large model companies are accelerating their penetration into industrial verticals; hard tech and automotive capital have become the primary new drivers; and A-share funding channels are systematically adapting to support AI infrastructure investment.

Target Audience: AI infrastructure investors, CFOs/strategic leads at large model companies, and BD heads in the hard tech supply chain.

It has been less than four months since DeepSeek’s first round set an industry record with a 51 billion RMB raise, yet the second super-round for this leading large model company is already nearing closing.

According to the latest Bloomberg report today (October 22), DeepSeek’s current fundraising scale has locked in at approximately 100 billion RMB, maintaining a pre-money valuation of around 500 billion RMB. Sources indicate that the deal underwent intense dynamic adjustments during its progression: the initial target was 50 billion RMB, but it was raised twice amid competition from various industrial capitals and existing shareholders, ultimately doubling the final amount to the hundred-billion level.

In terms of capital allocation, contributions from old and new shareholders showed a roughly equal split. Existing shareholders Tencent and CATL each added approximately 10 billion RMB, serving as cornerstone investors for this round. Meanwhile, new entrants leaned heavily toward the hard tech camp, including Shixi Capital, SMIC Juyuan, CPE Yuanfeng, Legend Capital, and automotive giant Geely.

If this hundred-billion-scale round is combined with the 51 billion RMB first round finalized by Liang Wenfeng and core institutions at the end of June, DeepSeek has accumulated over 150 billion RMB in capital within just over 100 days, setting a new global record for capital absorption speed in the AI sector.

External Liquidity Accelerates Entry

Echoing foreign media reports, there have been dense movements in funding channels within the A-share market.

Recently, Perfect World (002624.SZ) disclosed multiple investment progress announcements, stating that two key external special-purpose funds had fully closed their raises: the “Tianjin Lisi Xingshen” fund reached a size of 7.52 billion RMB, and the “Shixiang Shangshi” fund reached 3.65 billion RMB, collectively absorbing 11.17 billion RMB in external capital.

Behind these two large special-purpose funds are several A-share listed companies. Perfect World subscribed 50 million RMB and 100 million RMB respectively, completing a dual-channel bet with a total contribution of 150 million RMB. Century Huatong subscribed 650 million RMB to Shixiang Shangshi through its wholly-owned subsidiary. Compute service provider Litong Electronics previously announced a subscription of 95 million RMB for shares in Lisi Xingshen.

Meanwhile, public information from the Asset Management Association of China shows that Lisi Xingshen completed its private equity fund registration on September 15, 2026, with registration code SEX397. According to sources, both Lisi Xingshen and Shixiang Shangshi serve as external asset management channels for DeepSeek’s current fundraising round. As these funds complete their paid-in capital closures, the channels for injecting external social capital into DeepSeek’s core asset pool are essentially locked in.

“Lisi Capital” (Lisi-affiliated capital) warrants close attention because it served as the core channel for external SPVs (Special Purpose Vehicles/project-specific funds) to heavily invest in DeepSeek during the previous round.

Founded in 2021 by former Sequoia China partner Cao Xi and former Boyu Capital partner Tim Wang, this hard tech investment firm covers areas such as AI large models, domestic GPUs, robotics, and AI infrastructure. Its portfolio includes Moonshot AI, DeepSeek, Unitree Robotics, and others.

In May of this year, Lisi Capital initiated the establishment of the “Tianjin Lisi Xingling” fund with a size of 2.9 billion RMB. It subsequently subscribed to partnership shares in the holding vehicle “Hangzhou Chengli,” ultimately indirectly holding approximately 0.8265% of DeepSeek’s equity through look-through analysis. Now, the newly established “Lisi Xingshen” has surged to 7.52 billion RMB, nearly tripling in size, further confirming its mature operation as a super-channel connecting secondary market liquidity from listed companies with top-tier AI assets.

More critical capitalization signals are emerging simultaneously. Bloomberg, citing sources, reported that while advancing this hundred-billion-level fundraising, DeepSeek is concurrently preparing for a STAR Market IPO, planning to submit its listing application as early as the end of this year, aiming for official listing in 2027. The appointment of Yan Wentao, a former Hillhouse Venture partner, as CFO adds a professional operator to this capital maneuver involving hundred-billion-level allocations and multi-layered regulatory scrutiny.

Alliance with Gaming, Compute, and Chip Industrial Capital

The entry of Perfect World, Century Huatong, and Litong Electronics makes the industrial synergy of DeepSeek’s current fundraising round more concrete. Two gaming giants, Perfect World and Century Huatong, collectively subscribed 800 million RMB to related special-purpose funds, while compute service provider Litong Electronics subscribed 95 million RMB. Although the three listed companies have different businesses, they clearly hit core segments such as next-generation game content generation and compute infrastructure.

In August of this year, DeepSeek launched a high-salary recruitment drive for game development and 3D rendering, drawing significant attention from the gaming industry. The open positions included Code Agent Data Engineer (Game Development) and Web 3D & Rendering roles. The company emphasized that candidates "do not need to be experts in model training, but must have genuine experience building games," placing heavy value on game developers' engineering methodologies, complex interaction handling, and years of accumulated R&D expertise.

Looking at the broader investor list, the participation of semiconductor industry capital such as Shixi Capital and SMIC Capital has injected a deep "hard tech" foundation into this funding round. Previously, DeepSeek’s architectural optimizations under limited compute resources were widely recognized by the industry, particularly its close collaboration and adaptation efforts with domestic computing power like Huawei Ascend.

With the entry of core semiconductor capital linked to SMIC, future moves regarding low-level adaptation for domestic chips, the construction of customized compute clusters, and even self-developed chips are worth continuous tracking.

Overall, DeepSeek’s capital sources are expanding horizontally from traditional internet giants and pure financial investment institutions to a broad array of real-economy sectors, including semiconductor manufacturing, the automotive industry, and energy hardware.

Embodied Intelligence Becomes a New Variable

Beyond changes in the investor structure, founder Liang Wenfeng’s recent intensive focus on embodied intelligence reveals a new strategic direction behind this mega-funding round.

Earlier reports indicated that Liang Wenfeng recently engaged in discussions with several embodied intelligence entrepreneurs, exploring directions such as robotics, large models, and interaction with the real world. Sources familiar with the matter revealed that DeepSeek is considering further entry into this field, with potential paths including building an in-house R&D team or entering through investments in related companies.

On the investment front, DeepSeek has already backed three major embodied intelligence enterprises:

• Unitree Robotics (Disclosed): DeepSeek previously invested approximately 141 million RMB to acquire a stake in the company, gaining access to robot hardware, motion control, and general development/testing resources. The two parties also plan to collaborate on developing robot AI models, with Unitree providing the foundational platform and motion control capabilities for DeepSeek’s exploration of robotic applications.
• Winverse AI (Lead Investor Expected): Founded by Hu Han, a former AI scientist at Tencent. DeepSeek is expected to lead its seed round (amounting to tens of millions of dollars), with a post-money valuation of approximately $400 million. Meituan’s subsidiary Longzhu Investment is also among the participants.
• Yu Tao’s Startup (Name/Entity Pending Disclosure): A new venture by Yu Tao, formerly a member of Tencent AI Lab’s Juewu team and head of the humanoid robot foundation model at LimX Dynamics. The company focuses on combining Large Language Models (LLMs) with embodied intelligence. Its valuations across two rounds are reportedly around $70 million and$200 million respectively. Recent media reports indicate that DeepSeek and top-tier funds have participated in the investment.

Embodied intelligence involves real-world data interaction, multi-physics simulation, and high-precision motion control, requiring extremely high sustained consumption of compute resources, engineering testing environments, and physical hardware bodies.

This partially explains why DeepSeek’s current funding target was rapidly raised to the hundred-billion level in a short period, revealing its strategic ambition: to formally expand beyond cloud-based virtual code model inference toward intelligent entities capable of physical interaction in the real world.