DeepSeek, the Chinese AI lab, has more than doubled its annualised revenue to $1 billion in a few months, the tech news site The Information reported.
The run rate, which projects current sales over a full year, stood at under $500 million in July.
The jump followed an increase in the prices DeepSeek charges developers in August.
The company is finalising a 50 billion yuan (about $7.5 billion) fundraising at a valuation of 500 billion yuan (about $75 billion).
It aims to close the round by the end of October.
That compares with a $52 billion valuation in its first outside funding round earlier this year.
DeepSeek is also preparing for a stock market listing in Shanghai.
Its revenue sits between two Hong Kong-listed Chinese rivals: Zhipu AI, which has reported $1.8 billion, and MiniMax, which passed $800 million in August.
It remains a small fraction of the tens of billions generated by US leaders OpenAI and Anthropic.
DeepSeek makes gross margins above 80% on its cloud model services, despite charging some of the lowest prices in the market.
Investors credit its ability to run models on far fewer chips and less memory than its rivals need.
Founder Liang Wenfeng told investors at a recent closed-door meeting that revenue is not the company's main focus.
More than 70% of its computing capacity goes on training new models rather than serving customers.
It is prioritising Chinese chips for future training and expects Huawei to start delivering them in the fourth quarter.
