DeepSeek’s Annualized Revenue Reaches $400–500 Million
DeepSeek’s annualized revenue run rate has reached $400 million to $500 million, roughly double the estimates that circulated earlier this year, as the company prepares to raise fresh capital at a valuation of approximately 500 billion yuan ($74 billion) and files for a listing on Shanghai’s STAR Market as early as the second quarter of 2027. The figures, reported by Reuters on July 15, citing people familiar with the matter, offer the first substantive public picture of DeepSeek’s financial scale since the company emerged as a global AI force in early 2026.
The revenue trajectory reflects the commercial momentum that has built since DeepSeek released its R1 and V3 models at the start of the year. The company’s API pricing, among the lowest in the industry, drove rapid adoption among developers and enterprises seeking capable models at a fraction of the cost of US frontier models. Earlier estimates had placed daily revenue at approximately $562,000 against daily costs of roughly $87,000, implying gross margins well above 500%. The new annualized figure of $400–500 million suggests that API volume has grown substantially since those estimates were made.
New Round Targets $74 Billion Valuation — a Step Up from the June Close
The new fundraising round is being planned at a pre-money valuation of approximately 500 billion yuan ($74 billion), up from the 450 billion yuan post-money valuation (approximately $66.5 billion) at which the company’s first-ever external round closed in June 2026. That round raised $7.4 billion in total, with Tencent committing 10 billion yuan, CATL contributing 5 billion yuan, and the national AI fund participating alongside NetEase, JD.com, IDG Capital, Loyal Valley Capital, Monolith Management, and Shixiang Capital. Founder Liang Wenfeng personally committed 20 billion yuan to the round, a signal of his confidence in the company’s trajectory and a mechanism for maintaining control over its strategic direction.
The new round is expected to raise as much as 50 billion yuan, which would bring total external capital raised to approximately 100 billion yuan. The valuation step-up from the June round to the new round, from $66.5 billion post-money to $74 billion pre-money, reflects both the revenue growth and the broader re-rating of Chinese AI assets that has accompanied the sector’s strong performance in the first half of 2026.
IPO Filing Targeted This Year; Debut Expected in 2027
The IPO preparations are at an early stage. DeepSeek is targeting completion of its STAR Market filing this year, with a debut as early as Q2 2027. The company has been expanding rapidly to support its growth: it announced plans in late June to double staff across all departments, a hiring push that spans model research, infrastructure engineering, and commercial operations. It is also developing its own AI inference chip to reduce dependence on Nvidia hardware, a strategic priority that has become more urgent as US export controls on advanced semiconductors have tightened.
DeepSeek’s financial profile is unusual among AI companies of its scale. Unlike most frontier AI labs, which have operated at significant losses while building out infrastructure, DeepSeek has maintained high gross margins from the outset of its commercial life. The company’s origins in Liang Wenfeng’s High-Flyer quantitative hedge fund gave it both the capital and the computational expertise to build AI systems with an efficiency orientation that has become a competitive advantage as the industry grapples with the cost of inference at scale.
The combination of strong revenue growth, a committed investor base, and a clear path to public markets positions DeepSeek as one of the most consequential IPO candidates in China’s technology sector since Alibaba. Whether the STAR Market can accommodate a listing at the scale implied by a $74 billion valuation, and whether DeepSeek’s open-source model strategy is compatible with the disclosure requirements of a public company, are questions that will shape the IPO process as it develops.
The Open-Source Tension: Can DeepSeek Give Away Models and Still Go Public?
The open-source question is particularly pointed. DeepSeek has released model weights for its R1 and V3 series under permissive licenses, a strategy that has driven developer adoption and established the company’s technical reputation globally. But a public company faces pressure to monetize its intellectual property rather than give it away, and the tension between open-source community building and shareholder value maximization has no clean resolution. DeepSeek’s management will need to articulate a model for how open-source releases and commercial API revenue coexist, and convince public market investors that the strategy is sustainable — before the IPO can proceed on the terms currently being discussed.
For now, the revenue trajectory and the valuation step-up from the June round to the new round tell a story of a company that has moved faster than almost anyone expected, from a research lab to a commercial AI provider. The $400–500 million in annualized revenue, if it holds, would place DeepSeek among the fastest-growing AI companies in history by the time it reaches the public markets.