Deepseek is finalising a funding round that could reach $15 billion, significantly exceeding its initial target of $7.5 billion. Battery manufacturer CATL and tech giant Tencent are providing the largest portions of this capital. The investor interest follows the release of the V4-Flash model, which demonstrated superior cost efficiency and performance compared to offerings from Anthropic and OpenAI. After concluding the round, the company intends to restructure for an initial public offering in early 2027. Founder Liang Wenfeng aims to maintain a focus on open models with freely available weights while expanding infrastructure.
The company is constructing a new data centre equipped with at least 160,000 AI chips sourced from Huawei. This move addresses previous concerns regarding reliance on Nvidia hardware, a topic that caused the round to pause temporarily after Liang’s comments went viral. Deepseek has also begun developing its own inference chip to further reduce external dependencies. The startup previously raised $7.4 billion in June with a valuation above $50 billion before releasing preview versions of V4-Pro and V4-Flash in April.
- Valuation potentially approaches $15 billion
- Targeting an IPO in early 2027
- Building a data centre with 160,000 Huawei chips
