- Published on
DeepSeek Reportedly Plans to Lead Winverse AI Seed Round at $400 Million Valuation
One tap, and our reporting appears more often in your Google Top Stories. It makes a real difference to a small newsroom.


Your guide
Rich Robinson
Jee Chee Ren · Executive tour
See China’s AI and robotics ecosystem first-hand.
Six days exploring AI, robotics and advanced manufacturing.
With Rich Robinson, Clinical Professor of Entrepreneurship at Peking University
Beijing · Guangzhou · Shenzhen
29 November–4 December 2026
Referral partnership · We earn a commission on referred bookings.
DeepSeek reportedly plans to lead a seed financing round for Winverse AI, a startup developing AI for robots founded by former Tencent scientist Han Hu. The proposed deal would value the company at approximately $400 million after the investment, according to Bloomberg.
- Bloomberg reports that DeepSeek has agreed to lead a seed round of tens of millions of dollars for Winverse AI.
- The roughly $400 million figure is the proposed post-investment valuation, not the amount being raised.
- The transaction has not been finalized, and its terms and investor participation could change.
- Founder Han Hu previously led Tencent’s vision-language and embodied foundation-model teams.

A proposed bet on robot intelligence
Citing people familiar with the matter, Bloomberg reports that Winverse has also attracted commitments from other investors, including Meituan’s Long-Z Investments. The months-old company remains in stealth, and the report does not detail a product launch or technical architecture.
The distinction between the round size and valuation is significant: investors would collectively provide tens of millions of dollars, while valuing the business at around $400 million once that investment is included. Bloomberg says the transaction is still subject to change. Hu, DeepSeek and Long-Z did not respond to its requests for comment.
For the robotics industry, the potential investment is notable because it would bring backing from a major AI model developer to a company focused on robot intelligence. It does not, by itself, establish how DeepSeek’s models might be used by Winverse or whether the companies would jointly develop technology.
The weekly humanoid robotics briefing
One email a week: the launches, funding and research that mattered, with context from Humanoids Daily.
Read recent issuesFrom computer vision to embodied AI
Hu’s personal research page documents a career spanning computer vision, language models and embodied AI. It identifies him as the head of Tencent’s vision-language and embodied foundation-model teams, with earlier roles at Microsoft Research Asia and Baidu Research. He holds a doctorate from Tsinghua University and was an author of Swin Transformer, which won the ICCV 2021 best paper award.
The page uses the name Han Hu, while Bloomberg refers to him as Hu Han. It still lists his Tencent position; the account of his departure and creation of Winverse in 2026 comes from Bloomberg’s reporting.
Winverse would join a growing field of companies developing the intelligence needed to make robots useful across tasks and environments. The commercial challenge extends beyond understanding an instruction to carrying it out reliably on physical hardware. For now, the news concerns the capital and research talent behind that effort: Winverse’s capabilities and deployment plans remain to be detailed publicly.
Share this article
Read next
- Published on
- Reading time
- 4 min read
Archon Robotics Shares Debut Demo of Its Whole-Body Humanoid Intelligence
- Published on
- Reading time
- 3 min read
MindOn Claims 142 Garments per Hour in Hour-Long Mind-1 Folding Demo
- Published on
- Reading time
- 3 min read
Unitree’s Wang Xingxing Says GPT-6 Tests Strengthen His Optimism for Embodied AI
- Published on
- Reading time
- 3 min read
China’s Humanoid Robot Rental Prices Fall as Operators Face the Cost of Service
- Published on
- Reading time
- 3 min read
XPENG Targets H2 2027 European Rollout for IRON Humanoid Robot
- Published on
- Reading time
- 3 min read
Robotera’s VPP2 Tops RoboDojo Simulation Benchmark by Linking Video Prediction to Robot Actions
The weekly humanoid robotics briefing
One email a week: the launches, funding and research that mattered, with context from Humanoids Daily.
Read recent issues









