DeepSeek Secures USD 7 Billion in China's Largest AI Fundraise
- tech360.tv
- Jul 2
- 2 min read
DeepSeek, a Chinese artificial intelligence startup, completed its initial external fundraising round, securing approximately USD 7 billion. This funding, the largest single AI raise in China's history, results in a valuation nearing 400 billion yuan and signals a shift from the company's previous operational principles.
This substantial capital injection marks a departure from DeepSeek's long held stance against external fundraising, public offerings, or commercialization efforts since its inception. Previously, High-Flyer Quant, a quantitative hedge fund managed by DeepSeek founder Liang Wenfeng, provided all backing. High-Flyer Quant oversees more than 70 billion yuan and reported an annualised return of 56.55 per cent in 2025. The shift in strategy indicates a significant change in the organisation's financial approach.
But, according to Pandaily, the decision to seek outside investment was primarily influenced by the need to retain talent. Competitors, such as Zhipu AI and MiniMax, have seen their valuations increase significantly, with Zhipu AI now valued near HKD 1 trillion and MiniMax exceeding HKD 130 billion. These developments created incentive gaps for DeepSeek employees holding unvested options. An industry insider noted that without fundraising, the valuation would not increase, meaning employee options would not appreciate.
Following the funding announcement, DeepSeek advertised 33 vacant positions across seven distinct categories. These included roles in algorithms, research and development, operations, product management, data engineering, human resources, legal affairs, and finance. The organisation also commenced work on establishing an Internet Data Centre team in Ulanqab, Inner Mongolia, indicating a move towards substantial infrastructure investment. This expansion targets various core functions within the company.
And DeepSeek's infrastructure development proceeds under existing export controls concerning advanced semiconductor chips. This means its computing clusters will rely on domestically produced Chinese semiconductors. Huawei proposed its 'Tau (τ) Law' framework towards the end of May. This full stack collaborative optimization framework covers devices, chips, and systems, with the stated aim of mitigating the ongoing slowdown of Moore's Law.
On the technological front, the company continues to publish academic research. DeepSeek and Peking University jointly released a paper in late June detailing 'DSpark'. This new speculative decoding framework is reported to enhance inference speed by 60 to 85 per cent without altering the fundamental model architecture. The technique is slated for deployment on DeepSeek's V4 production system.
So DeepSeek's official V4 version is scheduled for release in mid-July. This upcoming version will introduce a new peak valley pricing mechanism for its application programming interface. The company's transition from an exclusive research laboratory to a commercial enterprise presents the challenge of preserving its previously lean and agile team culture, which was instrumental in the creation of DeepSeek R1, whilst simultaneously expanding to compete in the race for Artificial General Intelligence.
DeepSeek secured USD 7 billion in its first external funding round, marking China's largest single AI fundraising.
The funding signals a shift from the company's previous policy against external capital and commercialisation.
Talent retention challenges and competitor valuations influenced the decision to raise funds.
DeepSeek plans significant infrastructure investment, including an Internet Data Centre, using domestic Chinese semiconductors.
The company continues to advance its technology, with plans to launch its V4 system and a new pricing mechanism.
Source: Pandaily
