Alibaba’s cloud division led a 2 billion yuan round, approximately $290 million, with some outlets reporting $293 million due to currency conversion differences, in ShengShu Technology, the Chinese AI company behind Vidu. According to Reuters reporting on the announcement, ShengShu stated the funding will support development of a “general world model.” Co-investors in the round include Investment Fund, TAL Education Group, and Luminous Ventures.
The phrase “general world model” is doing real work here. ShengShu’s position, stated consistently across funding materials and secondary coverage, is that world models, AI systems trained on multimodal data including vision, audio, and touch, are better suited than large language models for AI operating in physical environments. LLMs process language well. ShengShu contends that autonomous vehicles, robotic systems, and any AI that needs to act in the physical world require a different architectural foundation: one that understands spatial relationships, physics, and sensory input together, not as separate tasks.
Whether that thesis holds against the weight of LLM-first development at every major lab is an open question. It’s ShengShu’s advocacy position, not an established technical consensus. But the fact that Alibaba’s cloud division is leading the round suggests someone with significant resources finds the argument credible enough to back at scale.
ShengShu’s existing product, Vidu, gives the company a concrete track record in video generation. Per Yahoo Finance’s coverage of the round, Vidu Q3 Pro launched in January 2026 and currently ranks ninth on Artificial Analysis’ AI video generation leaderboard. That’s not a dominant position, but it’s a verifiable foothold in a competitive benchmark environment where the top spots change frequently.
The applications ShengShu intends to pursue with its world model approach, autonomous driving and robotics, are the two sectors where the gap between language model capability and physical-world deployment requirements has been most visible. Both sectors have consumed enormous investment across multiple funding cycles. Both remain commercially difficult at scale. ShengShu’s bet is that the architectural problem, not the engineering problem, is what’s actually blocking progress.
For practitioners watching the Chinese AI ecosystem, this round is notable for two reasons. First, the Alibaba cloud division’s involvement isn’t a passive financial stake, it puts ShengShu’s infrastructure on Alibaba’s cloud roadmap, with implications for how Vidu and future world model products get deployed. Second, the “general world model” framing positions ShengShu in a small but growing cohort of companies, alongside efforts from DeepMind, Google DeepMind, and select research labs, arguing that the next significant architectural advance in AI runs through world models, not through larger LLMs.
What to watch: Whether ShengShu publishes any technical documentation for its world model approach, and whether Vidu’s leaderboard position improves or holds as competing video generation tools continue to release updates. The Artificial Analysis ranking is a concrete, trackable signal, ninth today doesn’t mean ninth in 90 days.
The broader read: Alibaba is making a directional bet, not just a financial one. A $290 million Series B in a company explicitly positioning itself against the LLM-first consensus is a meaningful signal about where at least one major Chinese cloud player thinks the architectural frontier is. That’s worth tracking regardless of whether you’re building with Chinese AI tools.