China's National Chip Fund Eyes $45 Billion DeepSeek Investment
China's flagship semiconductor investment vehicle is in advanced talks to anchor DeepSeek's debut external funding round, potentially valuing the AI startup at $45 billion—a dramatic signal of Beijing's strategic pivot toward homegrown foundation models as U.S. export controls tighten around advanced chips.
The National Integrated Circuit Industry Investment Fund (commonly known as the "Big Fund") is negotiating to lead the investment alongside multiple internet giants and state-backed funds, according to sources familiar with the matter cited by Shanghai Securities News. While the final valuation remains under discussion, the involvement of China's RMB 344 billion ($47.8 billion) Phase III chip fund marks a watershed moment for DeepSeek, which has operated entirely on internal capital from its parent quantitative trading firm Phantom since its July 2023 launch.
The potential deal comes as DeepSeek confronts twin pressures: an accelerating talent drain to better-funded rivals and surging infrastructure costs tied to compute-intensive model training. At least five core researchers have departed since late 2024, including Wang Bing (first-generation LLM architect), Luo Fuli (V3 contributor), and Guo Daya (R1 lead), moving to ByteDance, Tencent, and Xiaomi with compensation packages that DeepSeek's bootstrap model has struggled to match.
Technical Breakthroughs Fuel Valuation Surge
DeepSeek's newfound leverage stems from its April 24 release of DeepSeek-V4 Preview, an open-source model featuring million-token context windows and a novel attention mechanism rivaling closed-source systems from OpenAI and Google. In agentic coding benchmarks, the V4-Pro variant achieved best-in-class performance among open models, while its STEM and competitive programming scores moved closer to proprietary leaders such as Gemini Pro 3.1.
The Hangzhou-based startup—founded by Zhejiang University alumnus Liang Wenfeng—reportedly reached 30 million daily active users faster than any previous AI application, topping app store charts in 140 countries. The growth trajectory underscores both DeepSeek's technical differentiation and the market's demand for credible alternatives to U.S.-dominated AI infrastructure, particularly as Washington tightens restrictions on Nvidia's H100/H800 GPU exports to China.
However, sustaining that edge requires capital at scale. Training frontier models now routinely costs more than $100 million per run, with leading AI labs allocating billions annually toward compute infrastructure. DeepSeek's reliance on Phantom's trading profits—while allowing it to avoid early-stage dilution—has left the company structurally disadvantaged against Alibaba, Tencent, and ByteDance, each of which commands multi-billion-dollar AI investment budgets.
Ownership Consolidation Precedes Funding Talks
Ahead of negotiations, DeepSeek executed a critical equity restructuring. On April 27, 2026, registered capital increased from RMB 10 million to RMB 15 million ($1.39 million to $2.08 million), with Liang boosting his direct stake from 1% to 34%. Combined with indirect holdings through Ningbo Cheng'en Enterprise Management Consulting Partnership (which now holds 66%), Liang controls approximately 84.29% of voting rights—a defensive maneuver signaling his intent to retain decision-making authority after the funding round.
Sources indicate Liang also plans to participate personally in the current round, a move interpreted as reinforcing his long-term commitment to DeepSeek while partially offsetting dilution from incoming strategic investors.
Earlier financing rumors in February 2025—claiming Alibaba planned a $10 billion investment at a $100 billion valuation—were swiftly denied by Alibaba Vice President Yan Qiao. Yet persistent speculation throughout April 2025 suggested Alibaba and Tencent separately explored potential investments at valuations exceeding $20 billion, highlighting intensifying competition for strategic AI assets amid China's push for technological self-reliance.
Strategic Implications of State Capital Injection
The Big Fund's potential entry carries significance beyond financing alone. Established under the State Council's 2014 guidelines to strengthen China's semiconductor supply chain, the fund historically concentrated on fabrication equipment, materials, and chip manufacturing infrastructure. A shift toward AI model developers reflects Beijing's growing recognition that software-layer sovereignty matters as much as silicon, particularly as Huawei's Ascend chips and domestic GPU alternatives continue attempting to narrow the performance gap with Nvidia.
Industry analysts note that combining state-backed capital with market-oriented co-investors—reportedly including Tencent and sovereign wealth vehicles—could provide DeepSeek with the financial resources needed for hyperscale compute procurement and talent retention while preserving operational flexibility. At the same time, the arrangement raises broader governance questions surrounding model deployment, data sourcing, and international collaboration under increased state involvement.
Navigating a Crowded, Capital-Intensive Arena
DeepSeek's fundraising push unfolds against a backdrop of fierce domestic and global competition. Within China, ByteDance's Doubao, Baidu's Ernie, Alibaba's Qwen, and Tencent's Hunyuan all benefit from parent ecosystems spanning cloud infrastructure, e-commerce, and social platforms—distribution advantages DeepSeek lacks. Zhipu AI and MiniMax, meanwhile, have each secured more than $1 billion in venture backing, accelerating their commercialization efforts.
Globally, OpenAI's GPT-4 successor and Anthropic's Claude series continue to lead in advanced reasoning capabilities, while Google's Gemini Ultra benefits from deep integration across search and productivity products. DeepSeek's open-source positioning differentiates it philosophically, but also complicates monetization because enterprise customers often prefer commercially supported and liability-protected licensing structures.
The startup's next challenge lies in converting technical momentum into sustainable revenue streams—whether through enterprise API licensing, industry-specific fine-tuned models, or partnerships with Chinese cloud providers seeking alternatives for inference workloads. A $45 billion valuation implies expectations of eventual decacorn-scale revenue generation, a threshold only a handful of AI-native companies globally have approached.
As negotiations continue, DeepSeek's trajectory may become a test case for whether China's hybrid state-market investment model can foster globally competitive AI champions—or whether capital alone is insufficient against rivals controlling end-to-end computing ecosystems. If completed, the transaction would represent the largest AI investment in Chinese history, reshaping both domestic innovation dynamics and the geopolitical balance surrounding foundation model development.
Related Coverage：
DeepSeek Raises Capital as Founder Liang Wenfeng Increases Direct Stake to 34%
DeepSeek Seeks $300M at $10B Valuation as Nvidia Dependence Becomes Existential Liability