DeepSeek is preparing its first funding round and is expected to raise around 50 billion yuan, equivalent to US$7.4 billion. The operation could value the Chinese artificial intelligence (AI) startup at between 350 billion and 400 billion yuan, or between US$52 billion and US$59 billion.
The numbers show that DeepSeek has a valuation and fundraising volume that are completely different from those of its Western competitors at the same stage of development.
In 2019 and 2021, when OpenAI and Anthropic conducted their first funding rounds, the generative artificial intelligence market was still nascent and considered high-risk.
The first to raise capital was OpenAI, with Microsoft, in a funding round and valuation of US$1 billion. Two years later, Anthropic raised US$124 million for a valuation of US$550 million.
Among the investors negotiating with DeepSeek, according to Reuters , are some of the most prominent names in the Chinese economy. Tencent is considering investing 10 billion yuan, while CATL, a global giant in batteries for electric vehicles, is studying an investment of 5 billion yuan.
If confirmed, the two groups will be the largest external investors in DeepSeek. NetEase, JD.com, and China's national artificial intelligence fund are also in final talks.
The expectation is that the funding round will be completed in the coming weeks, although the financial terms and the final list of investors could still change. So far, DeepSeek has made no indication of plans for an IPO, unlike its Western competitors.
The funding round is expected to be among the largest technology funding rounds ever made in China. And it marks a change of course for the company founded by Liang Wenfeng.
Until now, DeepSeek had been avoiding third-party capital, relying instead on the resources of its founder and the profits of High-Flyer, a quantitative hedge fund created by Liang. In the new round, he reportedly committed to investing 20 billion yuan.
The movement didn't start now. In April, DeepSeek was negotiating its first external funding round, then designed to raise at least US$300 million and seek a valuation exceeding US$20 billion. The appetite of strategic Chinese investors, such as Alibaba and Tencent, had already raised expectations surrounding the round. Now, the ambition has reached a new level.
The irony is that DeepSeek gained worldwide recognition precisely by challenging the capital-intensive logic of AI. In January 2025, when the startup's rise gained the spotlight, its R1 chatbot had reached the top of the Apple Store downloads and entered the top 10 of Chatbot Arena, a popular ranking of AI model performance.
At that moment, shares of companies linked to the chip supply chain fell in the American pre-market, with Nvidia dropping as much as 11% and ASML as 9%. The fear was that DeepSeek could deliver performance close to that of American rivals at a fraction of the cost, thus reducing the need for billions of dollars in spending on semiconductors and infrastructure.
DeepSeek itself helped build this narrative. The company claimed its latest models had cost $5.6 million to train. The comparison was powerful: Dario Amodei , CEO of Anthropic, had previously estimated costs between $100 million and $1 billion for cutting-edge AI tools.
The startup's new investment round, however, shows that the equation has become more complex. Even when training is more efficient, the competition for distribution, talent, infrastructure, and new models continues to demand capital on a large scale.
The industry itself has also shifted. According to Reuters , the race has moved away from low-cost, open-source chatbots, which propelled DeepSeek, and towards AI agents capable of performing more complex tasks with less human intervention—and which demand significantly more computing power.
The billion-dollar funding round sought by the company shows that it still carries the image of being able to do cutting-edge AI more efficiently. But, to continue competing with OpenAI, Anthropic, Alibaba, ByteDance, and Tencent, it needs to enter the same capital race that it once seemed to challenge.
Anthropic raised $65 billion last month, while OpenAI raised $122 billion in March.
CATL's presence in the round also helps explain the new architecture of the Chinese AI competition. Best known for its leadership in the global supply chain of batteries for electric vehicles, the company has been advancing in artificial intelligence data centers, with energy and storage solutions for intensive workloads. In practice, the investment connects two critical fronts of the race: AI models and energy infrastructure.
Tencent, for its part, is trying to close the gap with its local rivals. The company has been promoting its own model, Hunyuan, but is still lagging behind domestic leaders such as ByteDance 's Doubao and DeepSeek itself. A closer relationship with the startup could help Tencent gain traction in AI, a market in which Alibaba has invested heavily in Qwen.
More than just a venture capital round, the deal adds a new dimension to the technological rivalry between China and the United States. DeepSeek has become a sort of national champion of Chinese AI after its V3 and R1 models gained global recognition and were hailed in Silicon Valley as a sign that China could advance in AI even under restrictions on access to the most advanced American chips.
For DeepSeek, there is also a structural limitation. Western restrictions on chip exports prevent the startup from accessing the most advanced American semiconductors. This limits its hardware strategy and helps explain why funding tends to be concentrated within China. At the same time, it reinforces the company's role within Beijing's technological self-sufficiency agenda.
