Deepseek targets $12 to $15 billion with CATL and Tencent

The Chinese AI startup is preparing a funding round of at least $12 billion, potentially increasing to $15 billion. At the same time, it is building a data center with 160,000 Huawei chips and plans an IPO in early 2027. The V4-Flash model has set new benchmarks against Anthropic and OpenAI.

Infrastructure, Models, and Technological Capabilities Showcased by Deepseek

Deepseek is constructing a data center that integrates at least 160,000 AI chips supplied by Huawei. The founder, Liang Wenfeng, aims to continue developing open models with freely accessible weights. The company is also working on designing its own inference chip to reduce its dependence on Nvidia and Huawei. The funding round had been put on hold after Liang's comments about dependence on Nvidia chips circulated widely. In April, Deepseek launched the preliminary versions of V4-Pro and V4-Flash.

Inflated Funding Round, with CATL and Tencent as Major Contributors

Deepseek is set to raise at least $12 billion in a new funding round, a figure that could reach $15 billion according to sources close to the matter. The initial target was around $7.5 billion for a valuation estimated at about $75 billion. CATL and Tencent are among the main investors in this round; CATL declined to comment, while Tencent and Deepseek did not respond to requests for comments. In June, Deepseek had already completed an initial external funding round of approximately $7.4 billion, with a valuation exceeding $50 billion.

IPO Timeline and Claimed Performance of V4-Flash

Following this fundraising, Deepseek is considering a reorganization in preparation for an IPO expected in early 2027. Deepseek's V4-Flash model has set new standards in terms of cost and efficiency compared to Anthropic and OpenAI.