DeepSeek nears at least $12 billion before its IPO

Bloomberg reported on October 6 that DeepSeek is close to raising at least 80 billion yuan, about $11.9 billion, with Tencent and battery maker CATL committing the largest amounts. The lab first sought around 50 billion yuan, but demand jumped after its latest model launched, and signed term sheets suggest the total could approach 100 billion yuan, about $14.9 billion.

The round is seen as the last big financing before an IPO planned for early 2027. DeepSeek had grown almost entirely on funds from High-Flyer, the quant fund run by founder Liang Wenfeng, and shook markets in January 2025 with a low-cost model that rivaled far pricier ones.

Tencent already uses DeepSeek models in WeChat and its cloud. CATL's stake points to the link between energy and AI, as data centers need storage and manufacturers invest in AI-driven production.

AI funding hit records this year, led by OpenAI's $122 billion round and Anthropic's $30 billion. Chinese labs, including DeepSeek, Qwen and GLM, lead in open-weight models, and last month Chinese open models were reported to hold a majority of usage on developer platforms.

Reports focus on the amount; the valuation has not been disclosed, and it will shape the IPO price. DeepSeek has released many models as free open weights and earns mainly from API use, so how it balances openness and revenue as a listed company will be closely watched.

The round is expected to close in the coming weeks. The listing venue is unclear, and US export curbs on advanced chips will shape what hardware the money can buy.

DeepSeek's edge is cost: it builds strong models cheaply, which means cheaper services and a wider market. That efficiency, more than any single model, is what investors are paying for ahead of the listing.

More money behind low-cost open models means cheaper AI. Revisit automations that looked too expensive a year ago, build products that can switch models, and know where customer data is processed whichever model you choose.