DeepSeek Made Anti-Poaching a Condition for First-Round Investors

DeepSeek's debut external funding round came with an unusual admission: the Chinese AI lab trusted investors enough to accept their capital but not enough to let them near its people. According to CNBC's reporting, citing 36Kr, founder Liang Wenfeng laid out a non-negotiable condition during a four-hour virtual meeting with prospective backers in May: invest, and you must not recruit DeepSeek's engineers or encourage them to start their own companies. The round reportedly closed at a valuation exceeding $50 billion, raising approximately $7.4 billion and making DeepSeek China's most valuable AI-only startup.

The anxiety behind that clause has a specific origin. Luo Fuli, reportedly a core contributor to DeepSeek's V3 model, left late last year to lead Xiaomi's MiMo AI team, which has since released models that outperformed DeepSeek on several benchmarks. That loss appears to be part of what finally pushed DeepSeek, which had declined outside funding since its founding to prioritize research over commercialization, to open itself to external capital at all.

The rest of the deal's structure reinforces how carefully DeepSeek's leadership is guarding its operational independence. According to 36Kr, China's state AI fund retained both voting rights and an exemption from lock-up, while commercial investors, reportedly including Tencent and CATL, surrendered all voting rights and face a five-year lock-up. Founder Liang contributed approximately $3 billion of the round personally, cementing his position as its single largest contributor. The anti-poaching clause, read alongside that structure, is less a conventional legal instrument than a statement of expectations: investors are capital sources, not talent pipelines, and DeepSeek intends to keep those roles separate.

What the reporting does not address is how enforceable these clauses actually are under Chinese law, what penalty would apply to a breach, and whether Luo Fuli's departure reflects a broader pattern of researcher attrition or an isolated case. Anti-poaching agreements can be difficult to enforce even in jurisdictions with robust contract law; in practice, the clause may function more as deterrent than binding constraint.

For other Chinese AI labs competing for researchers who can train competitive large language models, the situation confirms something already visible in hiring patterns: the market for that talent is active enough that a lab of DeepSeek's stature felt compelled to write talent protection into its term sheet. Whether capital, contractual terms, or researcher autonomy ultimately determines who keeps the best people is the live question this round leaves open.