DeepSeek’s Revenue Reaches $70 Million as of July, Tenfold Jump from 2025

DeepSeek generated about 475 million yuan ($70.7 million) in revenue in the first seven months of this year, roughly tenfold its full-year 2025 revenue, as the Chinese AI lab goes full steam ahead in its second round of funding, according to two people with knowledge of the company’s financial data.

DeepSeek recorded a net loss of about 715 million yuan between January and July, compared to 935 million yuan for the entire 2025, the two people added. The company is having ongoing discussions with existing and new investors about the funding round, which aims to raise 50 billion yuan at a valuation of 500 billion yuan.

DeepSeek didn’t immediately respond to an emailed request for comment.

While DeepSeek’s revenue is still minuscule compared to U.S. AI model developers such as Anthropic and OpenAI, the company is achieving a healthy gross profit margin. That’s because it has managed to keep the costs of running its models low, such as improving the efficiency of its AI infrastructure so that the AI systems can perform tasks using fewer chips.

DeepSeek’s gross margin for the first seven months was 44.6%, and the gross margin for selling access to its models through application programming interface was 82.9%. By comparison, OpenAI generated $5.7 billion in revenue and a 39% gross margin in the first quarter, while Anthropic recorded $11.5 billion in revenue in the second quarter and projected that its gross margin in 2026 would rise to 63% from 40% in 2025.

The revenue and other numbers, which haven’t previously been reported, show an accelerated pace in just the last few months. The Information last month reported that DeepSeek’s annualized revenue had reached between $400 million and $500 million.

DeepSeek’s revenue growth could give a boost to the funding talks and a potential initial public offering down the line. The company raised 50 billion yuan in its first-ever funding round closed in June and almost immediately started its second funding round. It has hired investment banks to help prepare a listing in Shanghai next year.

The growing revenue highlights the greater adoption of DeepSeek’s open-source AI models. Its flagship V4 series, especially the small-size V4-Flash, is popular among developers around the world due to their low costs and solid capabilities.

DeepSeek has recently taken steps to further increase its revenue. Earlier this month, the company increased its API prices. The price of its flagship V4-Pro model, for example, rose to $3.96 for output per 1 million tokens at peak hours and $1.98 at non-peak hours—up sharply from the current price of $0.87.

But even after the price hike, DeepSeek’ models are still much cheaper than other popular Chinese and American models. Moonshot’s Kimi K3 charges $15 for output, while Anthropic’s Claude Opus 5 costs $25 for output.

DeepSeek’s expenses are growing fast, too, albeit still tiny compared to the U.S. labs. In the first seven months of this year, DeepSeek spent about 11 billion yuan on AI infrastructure-related expenses, including renting servers equipped with AI chips or purchasing chips and other computing equipment. In 2025, the company spent only about 1.2 billion yuan on such expenses.