DeepSeek's $74 Billion Valuation Primed for IPO as Revenue Surges 10x

DeepSeek is raising fresh capital at $74 billion ahead of a potential initial public offering (IPO) on Shanghai's STAR Market (Science and Technology Advanced Registration Board).

The Hangzhou-based startup seeks up to 50 billion yuan in the new round. Three people with knowledge of the matter confirmed the plan. It comes weeks after a $7.4 billion financing in June valued the company at 450 billion yuan post-money.

I get the sense this functions as a valuation calibration ahead of the listing. Back-to-back rounds, six weeks apart, at escalating prices suggest DeepSeek's bankers are building a comps narrative for the roadshow.

Revenue changes the conversation. DeepSeek reported 475 million yuan, roughly $70.7 million, in revenue for the first seven months of 2026. That is approximately 10x year-on-year growth. The figure landed after the June round closed. It probably influenced the step-up to 500 billion yuan.

I think the revenue multiple is hard to pin down from outside. If annualized run rate sits around 800 million yuan, the 500 billion yuan valuation implies a multiple north of 600x. That is aggressive even by AI standards. Public-market Chinese AI comps are thin and the scarcity premium is real.

The June round included a 20 billion yuan personal commitment from founder Liang Wenfeng. Tencent Holdings contributed 10 billion yuan. Contemporary Amperex Technology, the battery giant known as CATL (Contemporary Amperex Technology Co. Limited), added 5 billion yuan. Those three became the largest external shareholders.

Other investors ranged from China's national artificial intelligence fund to NetEase, JD.com, IDG Capital, Loyal Valley Capital, Monolith Management, and Shixiang Capital.

Why the money matters less than the filing

DeepSeek has started early deliberations on the STAR Market IPO. The internal target is to complete the filing this year, according to people briefed on the plan.

I think the STAR Market choice is the telling detail. A mainland listing sidesteps regulatory friction of a Hong Kong or US listing. That matters given ongoing US-China technology tensions. It also taps a domestic investor base willing to pay premium multiples for AI names.

The recent CXMT (JHXT, China's largest memory chipmaker) semiconductor IPO priced at $8.45 billion showed that appetite. But the mainland route limits the investor pool to Chinese institutional and retail capital. For a company that drew global attention with low-cost AI models last year, that narrowing is worth watching.

The competitive backdrop

ByteDance, Alibaba, and well-funded startups including Z.ai, Moonshot, and MiniMax are all spending heavily on computing infrastructure and engineering talent. Liang recently told staff the company plans to double headcount across departments. That includes data centers and AI agents, systems that perform tasks with limited prompting.

I get the sense the hiring targets are the real burn rate. DeepSeek has discreetly increased recruitment of chip-design engineers. The project aims to develop its own AI inference chip, Reuters reported. Custom silicon is capital-intensive and risky. If the chip effort stalls, it drags on an otherwise clean model-focused narrative.

The company's cost advantage shocked global technology markets. Its open-weight models appeared to rival leading US systems at lower training and operating cost. Competitors are closing the gap on efficiency every quarter.

The unresolved edge

The fundraising and IPO plans remain at early stages. Terms and timetable may change, the sources said. DeepSeek did not immediately respond to a request for comment.

I think the unverified piece is the revenue trajectory. The 475 million yuan figure covers seven months. If the back half of 2026 decelerates, the 500 billion yuan valuation looks stretched. If it accelerates, the round may be priced too low. The next quarterly revenue disclosure will be the first real data point outside of fundraising narratives.

Source