DeepSeek nears pre-IPO funding round, eyes 2027 stock listing in China

DeepSeek, a Chinese artificial intelligence startup based in Hangzhou, is close to closing a new funding round that would value it at about 500 billion yuan (roughly $74 billion) pre-money, moving toward a possible listing on Shanghai's Star Market. The information comes from the South China Morning Post, which spoke with sources familiar with the negotiations.

What's on the table

According to SCMP, the company was seeking to raise about 50 billion yuan in the round, with completion expected before the end of August. Investors reportedly include names that took part in the previous funding round, such as local venture capital firms Monolith and Shixiang Capital, as well as Chinese battery giant CATL (Contemporary Amperex Technology Limited).

Among the new investors under discussion, the sources named CPE, Legend Capital and Stony Creek Capital, a semiconductor-focused private equity firm historically linked to Zhu Yiming, chairman of chipmaker ChangXin Memory Technologies and founder of memory chip designer GigaDevice. Funds backed by GigaDevice and by state-owned investment vehicles from Hefei are also said to be involved.

What the investor list signals goes beyond the money: there is a strong presence of capital tied to China's semiconductor and battery supply chains, a signal about the domestic bet on building an AI infrastructure less dependent on foreign suppliers.

Toward the Star Market

The round comes as DeepSeek begins preparing for a possible IPO on the Star Market, the Shanghai Stock Exchange's technology board, often compared to the Nasdaq. According to two sources who spoke with the newspaper, the company could file for its stock listing as early as the end of this year, targeting a 2027 debut.

It's worth setting the context: this information comes from anonymous sources, who asked not to be named because they were discussing private data. Neither the valuation nor the timeline has been officially confirmed by DeepSeek. Even so, the move fits a trajectory that had been building since early 2025.

Why DeepSeek is on the radar

DeepSeek gained global attention in early 2025 by releasing language models that rivaled OpenAI's and Anthropic's in performance, but with reported training costs far lower, which shook the thesis that the AI frontier would necessarily depend on billion-dollar investments in computing. The company also adopted a relatively open stance, making model weights available, which made it popular among developers outside China.

An IPO on the Star Market, rather than on U.S. exchanges, fits the regulatory environment: listings of Chinese tech companies on American markets have become politically sensitive in recent years, and the Shanghai market has become the natural destination for strategic companies in the sector.

What changes for developers in Brazil

For Brazilian developers, the most direct effect isn't the listing itself but what it represents: a DeepSeek with more capital and the status of a publicly listed company tends to accelerate the pace of model releases and sustain an aggressive pricing strategy for longer. This matters because DeepSeek is today one of the lowest-cost-per-token alternatives in the LLM API market, and much of the Brazilian teams integrating AI into products look at the cost-per-call ratio as a decisive factor.

In practice, a few points deserve attention from those building software here:

• Supplier diversification. Having a strong player besides OpenAI, Anthropic and Google reduces lock-in risk and gives more bargaining power on inference costs.
• Open-weight models. DeepSeek's approach favors those who want to run models on their own infrastructure or on regional clouds, a relevant path for cases requiring data to stay within national territory.
• Compliance considerations. Adopting a Chinese-origin model in products that handle sensitive data requires evaluating data residency and the requirements of the LGPD, Brazil's data protection law, especially when inference runs on endpoints hosted outside Brazil.

What remains open

Several pieces are still not settled. The final size of the round, the definitive list of investors, the IPO filing date, and even the confirmation of the $74 billion valuation remain in the realm of anonymous sources. There's also no indication in the report of how the additional capital will translate into new models or the expansion of international-facing infrastructure.

What can already be said is that the race for AI capital is no longer a phenomenon confined to Silicon Valley. A potential DeepSeek stock market debut in 2027 would, for the first time, put a Chinese frontier-model company under the scrutiny of a public market, with financial statements and growth targets in plain view, something that helps gauge how sustainable the low-cost AI thesis the company has been championing really is.