DeepSeek Breaks Silence on Capital, Seeks $300M in First-Ever External Raise at $10B+ Valuation

DeepSeek, the Chinese AI lab that spent two years turning down capital from Tencent, Alibaba, and top venture firms, is now in talks to raise at least $300 million at a valuation exceeding $10 billion, according to a report from The Information first published on April 17, 2026. The deal would mark the Hangzhou-based startup's first-ever external funding round and a decisive break from its founder's long-standing research-first posture.

Key Highlights

• At least $300 million targeted at a valuation of $10 billion or more
• First external raise ever — previously funded solely by parent hedge fund High-Flyer Quant
• Round reportedly drawing heavy interest, primarily from domestic Chinese investors
• Comes as V4 training costs reportedly exceed $500 million per run and five senior researchers have departed for Xiaomi and ByteDance

A Significant Strategic Shift

Founder and CEO Liang Wenfeng had repeatedly turned down offers from Tencent, Alibaba, and multiple top-tier venture firms since DeepSeek's 2023 founding, relying instead on capital from High-Flyer, the quantitative hedge fund he co-founded in 2015. High-Flyer reportedly posted 56.6% returns in 2025 and generated roughly $700 million in revenue, funding DeepSeek's training runs without outside dilution.

That stance has now reversed. According to The Information, DeepSeek is engaging with investors specifically to cover high model development costs, expand compute capacity, and raise compensation to retain talent amid fierce poaching by Chinese tech giants.

Why Now

Three pressures appear to be converging on the lab simultaneously.

First, the next-generation V4 model has slipped multiple times — from Lunar New Year to February, then March, and now reportedly late April — due to engineering challenges including adaptation for Huawei Ascend chips. Industry reports peg the cost of a single V4 training run at over $500 million.

Second, DeepSeek lost several core researchers over recent months. Luo Fuli, a core contributor to V3, joined Xiaomi, while V3 co-author Guo Daya moved to ByteDance in what reports describe as a high-eight-figure package. Market rumors place competing offers at up to 100 million yuan (roughly $14.7 million) annually for top AI talent.

Third, an 11-hour outage in late March 2026 exposed operational fragility and underscored the need for deeper investment in cloud infrastructure and on-call engineering.

Investor Landscape

The Information notes that U.S. regulatory constraints and national-security scrutiny have effectively shut American venture firms out of Chinese frontier-AI deals, leaving the round to be contested primarily among domestic investors. One Chinese state-backed investor reportedly told The Information they "couldn't get in" despite having connections.

Impact on the AI Race

A $10 billion valuation places DeepSeek well below U.S. peers — Anthropic closed a Series G at a $380 billion valuation earlier this year, and OpenAI was last valued at $852 billion — but it vaults DeepSeek into the top tier of Chinese frontier labs alongside Moonshot, Zhipu, and Minimax. For a company that released R1 for a reported training cost of $5.6 million and triggered a global rethink of AI economics in January 2025, the round is less about capital scarcity and more about what the next chapter requires.

What's Next

Observers are watching for three signals: the final valuation and lead investor when the round closes, the V4 release (reportedly due by the end of April), and whether DeepSeek's Huawei Ascend migration proves production-ready. All three feed the same question — can China sustain an open-source frontier lab that competes with Western closed-model incumbents on both performance and unit economics?

DeepSeek has not officially commented on the funding report.

Source: The Information