According to multiple media reports, the Chinese AI company DeepSeek has completed a new funding round of over 7 billion dollars. This brings the company’s valuation to over 50 billion dollars, making DeepSeek China’s highest-valued AI startup overnight.
According to reports by The Information and The Wall Street Journal, more than 50 billion yuan was raised from investors. Founder and CEO Liang Wenfeng reportedly contributed the equivalent of approximately 3 billion dollars himself, according to SiliconANGLE. Reuters had previously reported that Tencent was also considering an investment of around 1.48 billion dollars.
DeepSeek entered the international spotlight in early 2025 with the launch of its open-source model, R1. That model attracted significant attention because it demonstrated performance on par with leading U.S. AI systems, while requiring considerably less computing power. At the time, the launch caused unease among investors in the chip sector, as more efficient AI models might require less specialized hardware.
New model aims to further reduce costs
In April, DeepSeek unveiled a successor to R1 called DeepSeek-V4-Pro. The model has 1.6 trillion parameters and uses a mixture-of-experts architecture. Under this architecture, only a portion of the total model is activated when answering questions, which limits the required computing power.
In addition, DeepSeek claims significant improvements in memory usage during inference. The model is said to be able to process prompts of up to one million tokens with considerably less cache memory than previous generations. This should further reduce the operational costs of AI applications.
This focus on efficiency sets DeepSeek apart from many competitors, who primarily rely on ever-larger models and more computing power. For organizations looking to deploy AI on a large scale, lower infrastructure costs can be a significant advantage.
DeepSeek’s technology now appears to be attracting attention outside of China as well. According to Axios, Microsoft is exploring the possibility of incorporating a customized version of a DeepSeek model into its new Cowork Copilot platform.
Microsoft is looking at DeepSeek
Microsoft reportedly aims to offer a more affordable alternative to the models from OpenAI and Anthropic currently used in various AI services. The company is considering a so-called fine-tuned version of DeepSeek V4 or a comparable open-source model.
If Microsoft does indeed choose DeepSeek, it would be a remarkable move. The software giant has been investing billions in OpenAI for years, but at the same time is looking for ways to reduce the costs of generative AI services. A more efficient model could be an attractive alternative in this regard.
With the new capital injection, DeepSeek now has sufficient resources to further intensify its competition with both Chinese and American AI developers. The company is thus emerging as one of the key players in the global AI market.