DeepSeek Readies Mainland China IPO as Second Funding Round Pushes Valuation to $71 Billion

IPO preparation

DeepSeek has begun preparations for an initial public offering on mainland Chinese stock exchanges and is targeting a 2027 trading debut, according to a Bloomberg report republished in mid-July 2026. The company aims to submit its listing application by the end of 2026 and is engaging accounting firms and investment bank advisors to complete the financial-report work that is a prerequisite for an A-share filing, with Shanghai identified as the likely listing venue. The Wall Street Journal and other outlets have corroborated the move, while DeepSeek had not publicly responded to the reports as of mid-July 2026.

Second funding round

The Financial Times reported that DeepSeek has opened a second funding round at a pre-money valuation of $71 billion, equivalent to roughly 480 billion yuan, a 37% increase from the $52 billion mark reached only six weeks after its first round closed. The IPO and second-round reports come amid intense primary-market demand, with multiple accounts describing high-profile funds struggling to secure allocations in DeepSeek's rounds.

First funding round and state backing

DeepSeek's first external funding round in its three-year history is nearing completion, with the National Integrated Circuit Industry Investment Fund, often referred to as China's "Big Fund," leading the round, Caijing reported. The round is sized at about 700 billion yuan, or roughly $10 billion, at a $45 billion pre-money valuation, which would make it the largest first-round financing for a Chinese AI foundation-model company. Market investors including Monolith砺思 Capital and Hillhouse Capital are in talks, and final allocations are largely set to be locked in by the end of July 2026. Earlier reporting from 36Kr on the investor lineup named Tencent, CATL-related entities, JD.com, NetEase, IDG Capital and the National AI Industry Investment Fund as participants, alongside a 20 billion yuan personal contribution from founder Liang Wenfeng.

Commercial metrics

QuestMobile data cited in coverage showed DeepSeek reached 129 million monthly active users in June 2026, ranking among the top three AI-native applications in China, while average monthly usage per user rose about 90% and session time rose about 110% year-on-year. The Information reported the company has reached $400 million to $500 million in annualised revenue, primarily from enterprise and developer API consumption, with the V4 API gross margin exceeding 50% even at price points well below OpenAI and Anthropic.

Technology and infrastructure

DeepSeek's April release of the V4 family added support for Huawei's next-generation Ascend 950PR AI chip, reducing dependence on the Nvidia CUDA ecosystem, according to Caijing. The V4-Pro variant is priced at 0.025 yuan per token, described as among the lowest globally, with single-token inference compute at 27% of the previous V3.2 generation when handling million-token contexts. The company is also expanding self-built gigawatt-scale data centres, in-house inference chips and research headcount as it transitions from a pure model lab toward a full-stack AI infrastructure operator.

Capital structure

Founder Liang Wenfeng directly and indirectly holds 84.29% of DeepSeek, retaining near-total voting rights, while all investor shares are subject to a five-year lock-up. In investor briefings, Liang has framed DeepSeek's mission as achieving artificial general intelligence through open-source models, with short-term commercialisation a secondary priority. Investors cited in the coverage noted that even after a state-led listing, DeepSeek's day-to-day operations and technology roadmap are not necessarily bound to public-market pressures, given the founder's controlling stake and locked-up investor base.