$293M on ~$10M ARR — Alibaba isn't buying a video tool, they're subsidizing a future compute customer. ShengShu's pivot from Vidu video gen to "world models" and robot control (Motus, open-sourced Dec 2025) tells you the endgame: sell picks and shovels to every Chinese company that needs multimodal inference at scale. Four serious players (Kling, Hailuo, Vidu, Dreamina) all burning cash in the same lane while their cloud providers write the checks — it's the exact same playbook as AWS funding AI labs, except with Beijing's Internet Investment Fund in the cap table making sure the winners stay onshore.

Top comment by @Benthic