(Refile to fix typos in second bullet point and in paragraph 6)
May 6 (Reuters) – Chinese AI startup DeepSeek could be valued at up to $50 billion in its maiden fundraising drive, three sources said, as the large language model builder seeks to reverse its years-long strategy of rejecting outside funding.
China’s 60-billion-yuan ($8.8 billion) national artificial intelligence fund, founded in January last year, is in talks to be a lead investor in DeepSeek’s fundraising, said one of the sources familiar with the matter.
The startup could raise $3 billion to $4 billion from the funding round to fuel its computing capabilities and improve employee benefits, said separate sources with knowledge of the matter.
Chinese tech giant Tencent Holdings has also been in talks to invest in DeepSeek, said the sources, who all declined to be named as the information is confidential.
DeepSeek did not immediately respond to Reuters requests for comment.
The China Integrated Circuit Industry Investment Fund, which is the main backer of the national AI fund, declined to comment. Tencent also declined to comment.
The Financial Times first reported on Wednesday, citing sources, that China Integrated Circuit Industry Investment Fund is in talks to lead DeepSeek’s first fundraising that could value the frontier AI lab at about $45 billion.
LOSING GROUND TO RIVALS
The maiden fundraising comes at a time when DeepSeek is losing ground to domestic competitors with deep pockets from tech giants such as ByteDance and Alibaba to upstarts like MiniMax and Moonshot AI that have raised billions from either private or public markets.
DeepSeek’s founder Liang Wenfeng, who sources said is directly involved in the fundraising talks, has for years stood out in China’s AI industry for preferring to fund and operate the company like a research laboratory, with the budget coming from his quant hedge fund High-Flyer, rather than a Chinese tech conglomerate or an initial public offering.
Liang could not be reached for comment.
Some of DeepSeek’s rivals have poached DeepSeek researchers, such as Luo Fuli, who left the startup last year to lead Xiaomi’s AI model team MiMo.
The fast-moving AI industry has also largely moved on from and recreated the low-cost, highly efficient open-source chatbot models that fueled DeepSeek’s globally viral breakthrough moment early last year.
The focus has shifted to agents that can do far more complex tasks with much less human intervention than chatbots, while requiring a lot more computing power to run.
DeepSeek claimed last month that its next-generation, agent-suited V4 model “redefined the state-of-the-art” for open-source models but third-party evaluations suggested it lagged behind the best models from some U.S. and Chinese competitors.
Tellingly, V4’s release did not repeat the global tech share selloff triggered last year by its predecessors V3 and R1.
($1 = 6.8110 Chinese yuan renminbi)
(Reporting by Reuters Staff; Editing by Sumeet Chatterjee, Louise Heavens and Elaine Hardcastle)
Comments