Key Points:
- DeepSeek closed its first-ever external funding round, raising over 50 billion yuan ($7.4 billion).
- The round values the Chinese AI pioneer at over $50 billion, a five-fold increase from early 2025.
- Under an unusual limited partnership structure, CEO Liang Wenfeng retains absolute voting control.
- Tech giant Tencent and battery manufacturer CATL anchored the all-domestic investment syndicate.
Chinese artificial intelligence pioneer DeepSeek closes a massive financial package this week, raising over 50 billion yuan ($7.4 billion) in its first-ever external funding round. The historic capital injection values the Hangzhou-based artificial intelligence pioneer at over $50 billion, positioning it as one of the most valuable tech startups in the world. To protect the startup’s research independence and long-term vision, the deal utilizes a highly unique and unconventional corporate structure. Instead of buying equity directly, investors must channel their capital into a limited partnership run exclusively by Chief Executive Officer Liang Wenfeng, preserving absolute founder control.
The decision to seek outside capital marks a significant strategic pivot for DeepSeek, which had previously resisted the traditional venture-capital system. Founded in July 2023 by Liang Wenfeng, the company originally operated solely on internal funding provided by its parent company, the quantitative hedge fund High-Flyer Capital Management. However, the soaring global demand for specialized computing hardware, the necessity for high-end talent retention, and the rapid pace of model scaling forced the firm to seek external financial backing. The fresh proceeds will primarily fund the expansion of its high-performance AI infrastructure, accelerate research and development, and support employee equity incentives.
The unique terms of the Series A financing round impose strict limitations on the participating investors. Aside from a single government-backed entity, the external investors do not receive any voting rights. Instead, the limited partnership structure grants them access to specific financial information and priority rights in future financing rounds. To prevent speculative trading and stabilize the company’s valuation, the agreement also mandates a strict five-year lockup period, during which investors cannot sell, transfer, or liquidate their shares. This highly restricted structure ensures that the firm can continue to focus on open-source research rather than facing immediate pressure to commercialize.
A powerful, all-domestic syndicate of industrial and platform giants anchored the record-breaking funding round. Founder Liang Wenfeng personally led the transaction, committing a massive 20 billion yuan (approximately 40% of the total round) of his own capital. Social media and gaming conglomerate Tencent Holdings emerged as the largest external investor, contributing 10 billion yuan. Meanwhile, Contemporary Amperex Technology Co. (CATL), the world’s leading electric vehicle battery manufacturer, invested 5 billion yuan, reflecting a growing convergence between artificial intelligence and high-volume industrial energy infrastructure.
The remainder of the private placement drew significant contributions from other major Chinese technology platforms and investment firms. E-commerce giant JD.com, internet pioneer NetEase, and venture capital firm IDG Capital each contributed 3 billion yuan to the round. The sole exception to the restricted LP structure involves China’s National Artificial Intelligence Industry Investment Fund. The state-backed fund invested 1 billion yuan directly into DeepSeek, securing active voting rights and escaping the five-year share lockup. This direct state involvement underscores the strategic importance that national planners place on the company’s research roadmap.
DeepSeek captured global attention and rose to national prominence in early 2025 following the release of its DeepSeek-V3 and DeepSeek-R1 models. The open-weight models shocked Silicon Valley by matching the performance of expensive, proprietary systems like OpenAI’s o1 at a fraction of the cost. The firm’s technical reports revealed that it trained its V3 model for a meager $5.6 million, utilizing highly efficient Mixture-of-Experts (MoE) architectures and custom training algorithms. This technological breakthrough shattered the prevailing industry assumption that building frontier-level AI models required hundreds of millions of dollars in computing hardware, establishing the startup as a major threat to U.S. technology monopolies.
The startup has continued to build on this cost-efficiency advantage by rapidly scaling its software architecture. Earlier this year, the company launched its highly anticipated successor, DeepSeek-V4-Pro, which extends the model’s capacity to an impressive 1.6 trillion total parameters. Despite its massive size, the model maintains a highly efficient, hardware-aligned sparse attention mechanism that activates only 49 billion parameters per token. By offering API input pricing that is up to 13 times cheaper than comparable Western models, the company has successfully driven widespread adoption among local developers and enterprise teams, creating a robust, self-sustaining data flywheel.
The successful finalization of the record-breaking $7.40 billion funding round marks a permanent turning page for the global artificial intelligence landscape. By securing a massive, long-term capital runway while retaining absolute founder control, DeepSeek has built a highly resilient framework to continue its open-source mission. As the company begins deploying its fresh capital to expand its high-performance compute clusters and recruit top-tier talent, its progress will establish a crucial precedent. The coming years will show whether this non-commercial, research-first model can successfully maintain pace with Western tech giants, or if the extreme computational demands of the AI race will require a shift toward traditional commercialization.
