- Tencent joins HongShan and Gaorong in backing the stealth startup
- The former Alibaba executive is betting on world models and embodied AI
Former Alibaba executive and Qwen model leader Justin Lin Junyang (林俊旸) has raised several hundred million dollars for a new AI startup focused on world models and embodied intelligence, according to a report by The Information.
The deal gave the company a post-money valuation of about $2 billion despite having no public product, revenue or official website.
The financing, which closed recently, was led by Gaorong Capital and HongShan with commitments of about $100 million each, while Tencent invested roughly $20 million.
The team is already preparing for a follow-on fundraising round, the report said.
The deal sets a new benchmark for first-round valuations among Chinese AI startups and highlights investor appetite for founders with proven track records in foundation models.
From linguistics to LLM
Lin, born in 1993, studied English at the University of International Relations in Beijing before earning a master’s degree from Beijing Foreign Studies University.
He joined Alibaba’s research arm DAMO Academy in 2019 as a senior algorithm engineer and rose four levels in six years to become one of the company’s youngest P10 technical leaders.
Alibaba’s internal employee grading system traditionally uses P (Professional) levels for technical and business staff, while senior executives may also hold management titles.
He led development of the Qwen family of large language models, which emerged as one of China’s most prominent open-source AI projects and helped Alibaba gain traction among global developers.
Focus on physical AI
According to The Information, the startup’s focus is not on predicting the next word but on predicting what happens next in the physical world — a core objective of world-model research that many in the industry see as a prerequisite for advanced robotics and embodied AI systems.
The move is not entirely new for Lin. He quietly assembled a robotics research team inside Alibaba in late 2025, signaling an early shift from language models toward embodied intelligence.
Following his departure from the company in March, he published an essay on X titled From Reasoning Thinking to Agentic Thinking, arguing that the next frontier of AI should center on “agentic thinking,” a concept widely interpreted as his entrepreneurial manifesto.
Multiple entities
Corporate registration records show that between May and June, Lin established several entities in Shanghai, including wholly owned Yuyong (Shanghai) Technology and a near-wholly owned company called Shanghai Bulage Technology, while also controlling a management partnership through a more complex ownership structure.
The company names appear to reference “pragmatics,” a branch of linguistics, reflecting Lin’s academic background before entering the field of natural language processing.
The investment also extends Tencent’s strategy of backing multiple AI model developers beyond its in-house Hunyuan model.
According to The Information, Tencent has previously invested in Moonshot AI, MiniMax and Zhipu AI, and is reportedly participating in the latest financing round for DeepSeek.
Transition to embodied AI
The funding underscores a broader shift in China’s AI investment landscape. After two years dominated by large language models, capital is increasingly flowing toward world models and embodied AI.
Investors are betting that future breakthroughs will come not from chatbots, but from systems capable of understanding and acting in the physical world.
Data from multiple sources shows that China’s physical AI, embodied intelligence, and world model startups have raised over 46 billion yuan ($6.8 billion) in more than 288 disclosed funding rounds during the first half of 2026 — marking an unprecedented surge in capital flowing into the sector.
