DeepSeek is preparing to take a step towards the BagThe Chinese artificial intelligence startup has chosen Citic Securities to prepare a possible Ipo on the Star Market of Shanghai, the technology listing of the Chinese stock exchange. The goal is to initiate the process by the end of 2026, as the company seeks new capital to support model development, computing infrastructure, and the competition for talent. This was reported by Reuters, based on information provided by two people familiar with the negotiations.
The choice of investment bank is a sign that the IPO project is entering a more concrete phase. There are not but still one data for the offer, one figure official to collect or a review company's final statement. DeepSeek and Citic Securities did not respond to requests for comment. ReutersThe investment bank's role, however, is significant: in China, companies intending to list on the mainland market typically entrust financial intermediaries with the preparation and support phase that precedes the application submission.
DeepSeek: From Liang Wenfeng's Funding to the Stock Market Rush
Behind the potential IPO is a company that has changed its face in just three years. DeepSeek was founded in Hangzhou in July 2023 on the initiative of Liang Wenfeng, also the founder of High-Flyer, the quantitative fund through which the startup was financed in its early years. Then came the turning point: from a founder-funded gamble to opening up the capital to large Chinese investors.
In June DeepSeek has raised approximately 7,4 billion of dollars, reaching a post-money valuation of more than $50 billion, according to Reuters. Liang Wenfeng personally invested 20 billion yuan, while Tencent and the battery manufacturer catl contributed 10 billion and 5 billion yuan respectively, becoming among the main external shareholders. Other important Chinese groups also participated in the round, including NetEase, JD.com and the Chinese National Fund for Artificial Intelligence.
And the rush for capital doesn't seem to be over. Reuters had reported in July of a new round of financing that could bring the review DeepSeek's up to 500 billion yuan, about 75 billion of dollars. A figure that also helps to understand the growing interest surrounding the future listing. Meanwhile, according to the Financial Times, the latest collection would be pushing the valuation around 71 billion of dollars, also fueling an informal secondary market in the shares.
The price of the race to artificial intelligence
Why do we need so much money? Because the'artificial intelligence, especially at the levels at which DeepSeek wants to compete, is a race which requires a huge amount of resources. Processors, computing power, data centers, energy, network infrastructure, and researchers have become the main items on the bill.
DeepSeek has built its fame on its ability to obtain results Competitive with more efficient use of resources. But efficiency doesn't eliminate the need for investment. According to information reported by the FT, the company will have spent approximately $1,6 billion on AI infrastructure in 2026, ten times its investment in 2025, while annualized recurring revenue will reach approximately $500 million.
There's another resource that's become difficult to acquire: talent. DeepSeek has recently lost some researchers to better-funded Chinese groups, including ByteDance e XiaomiThe upcoming IPO could also serve this purpose: offering stronger incentives to retain key personnel and using company shares as part of their compensation. In a sector where a small group of researchers can directly impact the quality of a model, human capital is almost as valuable as financial capital.
The story of DeepSeek, however, became global even before it became financial. At the beginning of 2025, the R1 model had challenged one of the dominant beliefs in Silicon Valley: developing competitive AI systems wouldn't necessarily require an unlimited supply of chips and capital. The impact was immediate. On January 27, 2025, shares Nvidia they ended up losing about 10%, while Wall Street began to question the consequences of more efficient models on the entire AI economy. DeepSeek had shown that the technological gap between the United States and China it could be less extensive than it seemed.
Since then, the startup has continued to update its models. On August 13, 2026, it presented DeepSeek V4 Pro, primarily focused on AI agents, programming, and the execution of complex tasks, along with a new business model for API access. The challenge, therefore, is no longer simply proving that the technology works: it's transforming it into a business capable of growth and financing.
Chinese AI's Rush to the Stock Market
DeepSeek isn't alone. During 2026 Z.AI e MiniMax they landed at the Stock Exchange Hong Kong, while Moonshot AI, the company that develops the Kimi family of models, has confidentially filed documentation for a potential IPO, again in Hong Kong. The rush to the stock markets has thus become a new chapter in the competition among Chinese AI companies.
Across the Pacific, however, the numbers tell a different story. anthropic, developer of the Claude and Mythos models, could be valued at up to $2.000 trillion when it IPOs, investors said. OpenAI It could instead aim for a valuation of up to $1.000 trillion in a future listing.
The gap isn't just a matter of investor enthusiasm. It also reflects a difference in revenues and highlights the difficulty Chinese companies face in turning increasingly competitive technology into profits. For DeepSeek, a listing could therefore become a tool to at least partially bridge this gap, funding research and growth without relying exclusively on large private investors. But the potential listing comes at a time when a much larger game is also being played out around DeepSeek.
DeepSeek on the new front of the technological war
Just a day before the news about the appointment at Citic Securities, US officials have accused DeepSeek, Moonshot AI, Alibaba and other Chinese companies to have used it on a large scaledistillation techniques, that is, having exploited the responses generated by American AI models to train or improve their own systems. According to US authorities, some Chinese companies would have used variants of models developed by anthropic, OpenAI, Google e SpaceX.
Washington argues that these practices were likely made known to Chinese authorities and that, in addition to lowering development costs, they could strengthen China's military and cyber capabilities. Beijing has rejected the accusations, arguing that China's AI progress stems from its own scientific and technological capabilities and calling for greater cooperation between the two countries.
The timing makes the story even more significant. DeepSeek's potential listing comes as the United States and China are battling over access to advanced chips, intellectual property, and computing power. An IPO on the Shanghai Star Market would therefore not only be a financial step for a startup that has become a major player: it would also demonstrate that one of China's flagship AI companies can find the resources to continue its momentum in its own capital markets.
For now, of course, it is still too early to talk about stock market debutThe IPO process has been launched, but the timing, size of the offering, and final valuation are still unclear. Citic Securities' decision, however, indicates that DeepSeek has begun to concretely prepare for the next chapter of its story: after challenging the American giants in the field of AI models, it now aims to compete with them in the equally crucial field of capital.
