DeepSeek aims for Shanghai IPO at around $75 billion

Chinese AI company DeepSeek is preparing an initial public offering (IPO) in Shanghai, according to several media outlets. It is one of the biggest tests for Chinese AI on the stock market to date. Advisory firm Nelson Advisors has written an analysis of the case and of what cheap and efficient AI models could mean for healthcare. I have not been able to find the analysis itself, so this article is based on media coverage of the listing plans.

Valuation and money

DeepSeek is in the middle of a new funding round that, according to reports, could value the company at around 500 billion yuan, equivalent to about $75 billion. Other sources cite $71 to 74 billion before the new money is counted. The company reportedly raised more than $7 billion in June 2026 at a valuation of over $50 billion.

According to one of the sources, founder Liang Wenfeng has personally pledged 20 billion yuan. Tencent and battery giant CATL are also considering investing 10 and 5 billion yuan, respectively. These figures have been reported by the media and not confirmed by the company.

Timeline

According to reports, DeepSeek has chosen CITIC Securities as one of several banks to prepare a listing on Shanghai's STAR Market, which resembles the US Nasdaq and is for technology companies. The company could file an application before the end of 2026, and the listing itself could take place in 2027. That depends on regulatory approval and market conditions.

Why efficient models matter

DeepSeek became known because the company's R1 model delivered strong answers at a fraction of the price of Western competitors. R1 was reportedly trained for about $294,000. That figure is the company's own claim. The newest model family, V4, can read up to one million tokens (pieces of text) at a time, meaning very long documents.

For ordinary readers, the point is that cheaper AI can make it economically realistic to use it in areas with tight budgets, such as healthcare. That could be for writing medical records or reading long patient histories. This is, however, a general consideration. I have not been able to confirm what Nelson Advisors specifically concludes.

Caveats

According to coverage, investors will demand clarity on revenue, profit and computing power. DeepSeek has also been banned in several countries, and competitors have raised allegations about how the company obtained access to data. That could also affect whether healthcare organizations dare to use the models, since patient data is particularly sensitive.