DeepSeek has agreed to steer Winverse AI's first external raise, described by people familiar as a seed round worth tens of millions of dollars. It expects to tap a portion of its own $12 billion-plus financing for the effort. Those people said the round would value Winverse at roughly $400 million once completed.
The transaction is not yet final, and specifics such as the valuation and the roster of investors could still shift, according to the people, who requested anonymity because the talks are private. Winverse has additionally secured support from other investors, among them Meituan's Long-Z Investments.
Winverse was set up in 2026 by Hu Han, who most recently served as lead scientist at Tencent, running the company's vision language and embodied foundation model groups. His exit from Tencent came as the WeChat operator reorganized its AI research unit in a push to keep pace with rivals such as DeepSeek and Moonshot.
DeepSeek was founded by hedge fund manager Liang Wenfeng in 2023 in Hangzhou and vaulted to industry prominence two years later. Before teaming with Winverse, it acquired a 2% stake linked to Unitree Robotics's Shanghai listing in order to collaborate on embodied-intelligence models. Hu did not respond to questions sent by email seeking comment. Representatives for DeepSeek and Long-Z likewise declined to respond to emailed inquiries.
Who funds robotics startups shapes which approach wins. Market Briefs covers AI capital free every weekday.
Most of that money is earmarked for training models and building data centers, but it could also support co-developing startups like Winverse.
But making those machines follow voice commands, solve physical tasks, and move beyond showy maneuvers requires new types of AI models that differ from systems popularized by OpenAI's ChatGPT or by DeepSeek. Capital is tighter in China, where Unitree and related players are continually seeking funds for their next research steps.
If it closes, this would be one of DeepSeek's largest startup investments and a clear move from lab work into factory-floor use cases. It also spotlights a broader trend in China, where top researchers are leaving tech giants to build specialized labs targeting world models for self-driving, interactive video, or robotics.
For everyday investors, the takeaway is simple: embodied AI is drawing serious money on both sides of the Pacific. That does not mean chase the flash. It does mean the bridge from algorithms to real machines is being funded, which could ripple into hardware suppliers, industrial automation, and cloud spend as these systems scale.
A model lab leading a hardware round is a notable crossover. Join Market Briefs free and follow the money.
