The founder of DeepSeek announces that AGI is the objective as the $10 billion funding round progresses.

      Liang Wenfeng has informed potential investors that the Hangzhou lab will focus on cutting-edge research rather than generating revenue in its inaugural external funding round. As the founder of DeepSeek, Liang has expressed that the lab's main objective is to work towards artificial general intelligence, continuing to release open-source models instead of pursuing immediate commercialization. Bloomberg highlighted this messaging on Friday amid coverage of a substantial 70 billion yuan (approximately $10 billion) fundraising effort.

      This amount reflects the valuation DeepSeek aims to achieve, not the actual investment amount. The startup is looking to raise at least $300 million in this round. According to The Information, the long-term goal could exceed $7 billion as the lab shifts towards generating regular revenue. This is the first instance in which Liang has accepted external investment.

      Historically, DeepSeek was entirely funded by High-Flyer Quant, the quantitative trading firm founded by Liang, which continues to support the lab. Liang has previously characterized the lack of outside investment as a way to prevent pressure on the product roadmap.

      The current fundraising round reflects a change in scale, as training runs at DeepSeek’s level have surpassed what even a profitable hedge fund can self-finance. The focus on AGI also indicates how the company positions itself within the industry. In April, DeepSeek introduced its V4-Pro and V4-Flash models, a 1.6-trillion-parameter Mixture-of-Experts system and a smaller 284-billion-parameter version, both released open-source under liberal licensing.

      The V4 models are optimized to function on Huawei Ascend and Cambricon hardware, in addition to Nvidia, signaling an intention to cater to a domestic Chinese market that is becoming increasingly distanced from top-tier US accelerators.

      According to Bloomberg’s interpretation, the investor pitch heavily emphasizes the research-first approach that led to the creation of the R1 model in January 2025. The release of R1 reportedly wiped around $600 billion off Nvidia’s market capitalization in a single trading day, based on the notion that frontier reasoning models could be trained for a fraction of the costs incurred by US labs. Whether this pricing assertion stands upon scrutiny became a topic of debate, but the strategic message was clear: a Chinese lab could maintain competitiveness at the forefront of the field while operating transparently.

      Details that are yet to be made public are the most crucial. The identities of investors have not been disclosed; the company's valuation remains unconfirmed; and the closing date has not been established. DeepSeek has typically refrained from responding to press inquiries, and the lab did not provide any official comments for Bloomberg’s article. A fundraising effort of this magnitude, at this valuation, with the outlined goals, is likely to attract scrutiny from Chinese regulators, who have been determining what state oversight of foundation-model creators should entail over the past two years.

      What Liang has publicly committed to is a clear direction: open-source, research-focused, and with AGI as the emphasis instead of enterprise revenue. When the round concludes, it will mark the initial occasion that external capital has accepted these terms.