According to two sources familiar with the financial data, this Chinese AI lab is actively pushing forward its second round of funding. In the first seven months of 2026, DeepSeek achieved revenue of about 475 million yuan (about 707 million US dollars), which is ten times its full-year revenue in 2025, and its net loss narrowed from 935 million yuan last year to 715 million yuan.

Open-source models drive high gross margins

Strong growth reflects a significant acceleration in business growth over recent months, paving the way for a potential IPO. The company just completed its first round of funding of 5 billion yuan in June, and immediately launched the second round, aiming to raise 5 billion yuan with a target valuation of 50 billion yuan, and has already hired investment banks to prepare for an IPO next year.

The confidence behind rising revenue comes from the continuous expansion of the deployment scale of open-source models. Its flagship V4 series, especially the lightweight version V4-Flash, has been popular among global developers due to its low cost and solid performance. The overall gross margin for the first seven months was 44.6%, while the gross margin from API calls was as high as 82.9%, far exceeding OpenAI's 39% and Anthropic's estimated 63%.

DeepSeek has also recently raised its revenue. Early this month, it increased API pricing. For example, the peak time price for the flagship V4-Pro model went up to 3.96 dollars per million tokens, and 1.98 dollars during off-peak hours, significantly higher than the previous 0.87 dollars level. Advancing the second round of funding at the moment when both revenue and gross profit are rising, this lab, which was founded on open-source, is turning its story of "low-cost and high-performance" into a scalable commercial narrative.