Sand AI Raises $200 Million A-Plus Round and Reaches Unicorn Status After $300 Million in Ten Months

Sand AI, a Beijing-based video generation startup, announced a $200 million A-plus round co-led by Cornerstone Capital and Wangsu Science and Technology, per a report from Chinese outlet Kechuang Board Daily. The round includes participation from Times Capital, Shanghai Guofang, Panlin Capital, Dongzheng Capital, Tianchuang Capital and several Shandong state-backed funds, with existing investors Matrix Partners China, Huayetiancheng and CAS Star adding again.

It is the company's third round this year and, by the company's own account, brings cumulative funding to roughly $300 million within ten months — enough to cross the unicorn threshold. The capital is earmarked for video foundation model research, training and inference infrastructure, product iteration and international expansion. Company-reported figures have not been independently audited, though the listed-company angle gives this round unusually verifiable paper trails.

That angle is Wangsu Science and Technology, a Shenzhen-listed CDN and edge computing provider. Wangsu disclosed on September 29 that it planned to invest 300 million yuan of its own funds into Sand AI's A-plus round, then announced on October 19 that agreements were signed on October 16 — structured as a convertible bond plus warrants over the holding entity, converting into equity. The two companies are also committing to commercial cooperation on audio-video inference and low-latency edge inference, with Wangsu contributing compute, network nodes, scheduling and security infrastructure.

Sand AI was founded in 2024 by Cao Yue, who earned both his bachelor's and PhD degrees at Tsinghua University, worked as a lead researcher at Microsoft Research Asia, headed a research center at the Beijing Academy of Artificial Intelligence, and co-founded the startup Light Years Beyond. He is one of the co-authors of Swin Transformer, the hierarchical vision transformer that won the ICCV 2021 best paper award — a credential that in the Chinese AI scene functions as both reputation and recruiting magnet.

The company's product line centers on the MAGI series of video foundation models, which generate visuals, sound and human motion together, and VidMuse, a creator-facing tool for ideation, storyboarding and video production. Sand AI sits in the same lane as Kuaishou's Kling, PixVerse and Shengshu Technology, all fighting for the video foundation model market that Sora and Google's Gemini Omni have pushed toward commodity pricing.

The Wangsu deal matters beyond the money. Video models live or die on inference economics — thousands of users calling expensive GPU time concurrently — and Wangsu's edge network gives Sand AI a distribution partner for the cost side of that equation. The company said the collaboration targets exactly that: serving video generation at scale with acceptable wait times and compute costs.

For China's AI funding market, the round is another signal that capital is concentrating around teams with top-tier research pedigrees and clear commercial wedges. Sand AI's bet is that video generation becomes infrastructure — and that the winners will be decided as much by inference cost per second as by model quality.