DeepSeek Goes External for $10B, AGI in Sights

DeepSeek founder Liang Wenfeng has told prospective investors that the lab's primary goal is artificial general intelligence (AGI), and it will keep releasing open-source models rather than chase short-term revenue. Bloomberg reported the messaging on Friday, covering DeepSeek's first-ever external funding round, which targets a valuation of 70 billion yuan (~$10 billion). The startup is seeking at least $300 million in this round, with longer-term ambitions of up to $7 billion as it moves toward recurring revenue, according to The Information.

From Self-Funded to Outside Capital

Until now, DeepSeek was entirely funded by High-Flyer Quant, the quantitative trading firm Liang founded. Liang previously framed the lack of outside investors as insulation from product-roadmap pressure. The shift to external funding is partly due to scale: training runs at DeepSeek's current size have outgrown what even a profitable hedge fund can self-finance.

AGI as the North Star

The AGI framing distinguishes DeepSeek from many AI startups that prioritize enterprise revenue. Liang has committed to open-source, frontier-oriented research. The lab's V4-Pro and V4-Flash models, released in April 2025, are a 1.6-trillion-parameter Mixture-of-Experts system and a 284-billion-parameter variant, both under permissive licensing. The V4 family is optimized for Huawei Ascend and Cambricon silicon in addition to Nvidia, signaling a deliberate focus on the domestic Chinese market increasingly cut off from high-end US accelerators.

R1's Market Impact

DeepSeek's R1 model, released in January 2025, erased roughly $600 billion from Nvidia's market cap in a single trading session. The claim that frontier reasoning models could be trained for a fraction of US labs' spending sparked debate, but the strategic point landed: a Chinese lab could keep pace at the frontier and do it in the open.

Open Questions

Investor identities, the formal valuation, and the close date have not been announced. DeepSeek has historically declined press inquiries and did not provide on-the-record comment for Bloomberg's report. A raise of this size with this stated goal will also draw attention from Chinese regulators, who have spent the last two years developing oversight for foundation-model developers.

What This Means for Developers

DeepSeek's continued open-source releases give developers access to state-of-the-art models without vendor lock-in. The V4 family's support for domestic Chinese hardware also matters for developers in markets affected by export controls. The AGI focus signals that DeepSeek will invest in long-term research, not just API monetization.

Technical Details at a Glance

  • V4-Pro: 1.6 trillion parameters, Mixture-of-Experts architecture
  • V4-Flash: 284 billion parameters, smaller variant
  • Both released under permissive open-source license
  • Optimized for Huawei Ascend, Cambricon, and Nvidia hardware

Summary

DeepSeek's first external funding round marks a strategic shift from self-funding to outside capital, while maintaining its commitment to open-source AI and AGI research. The outcome will signal whether investors buy into a research-first, revenue-later approach.