Alibaba Leads $290 Million Investment in ShengShu for Next-Generation AI Models Beyond LLMs

Apr 12, 2026

Alibaba Cloud led a 2 billion yuan ($290 million) series B investment in ShengShu, a three-year-old Chinese AI startup developing world models that aim to surpass traditional large language models by connecting perception and action. ShengShu, founded by Zhu Jun, previously raised 600 million yuan two months earlier and created the Vidu AI video generation tool that launched globally before OpenAI's Sora and ranks among the top 10 models according to Artificial Analysis. The funding will support development of a general world model using multimodal data like vision, audio, and touch to bridge digital applications with physical world uses in autonomous driving and robotics. This investment reflects a broader industry shift toward world models built on videos and physical scenarios rather than text-based language models, with Alibaba also backing similar startups Tripo AI and PixVerse while developing its own AI models for video generation and robotics.

Alibaba's $290 Million Investment in ShengShu and Vidu

▪Alibaba and Baidu Ventures led a $50 million investment in Tripo AI last month.
▪ShengShu raised 600 million yuan from Qiming Venture Partners and other backers about two months before the Alibaba-led investment.
▪TAL Education and Baidu Ventures participated in ShengShu's series B funding round.
▪ShengShu launched Vidu globally months before OpenAI made its Sora tool for AI video generation widely available.
▪ShengShu is the startup behind the AI video generation tool Vidu.
▪Alibaba led a $60 million investment in PixVerse in September.
▪ShengShu's Vidu Q3 Pro model ranks among the top 10 AI models for generating videos from text and images according to Artificial Analysis.
▪ShengShu's latest Vidu Q3 Pro model was released in January.
▪Alibaba Cloud led a 2 billion yuan ($290 million) investment in ShengShu in a series B funding round announced Friday.
▪ShengShu declined to disclose its valuation after the series B funding round.

Shift from Large Language Models to World Models

▪Tripo AI is moving away from techniques used by language models toward AI tools grounded in physical space and is developing its own world model.
▪Alibaba released free, open-source AI models for video generation.
▪ShengShu believes a general world model built on multimodal data such as vision, audio, and touch more naturally captures how the physical world works than large language models.
▪Alibaba launched an AI model for powering robots in February.
▪PixVerse released an AI world model earlier this year that allows users to direct how a video unfolds while it is being generated.

World Models as Critical Infrastructure for Robotics and Embodied AI

▪Kevin Kelly wrote that LLM-powered chatbots have created the knowledge element of AI, making world models a key area requiring a breakthrough.
▪ShengShu has strategic partnerships with companies developing embodied AI systems such as humanoid robots that interact with the physical world for use across industrial, commercial and home settings.
▪Kevin Kelly wrote that world models are critical for robotics because the technology needs more than LLMs to work.
▪Kevin Kelly wrote that to replicate human intelligence, AI will need three things: reasoning, an understanding of the physical world, and continuous learning.