Video generation is proving the priciest AI category to compete in, and Sand.ai has just topped up its war chest accordingly.
The Chinese startup closed a $200M Series A+ on 5 November, a round it says pushes it into unicorn territory and takes its 2026 fundraising to $300M across three separate raises. Wangsu Science & Technology and Cornerstone Capital put in the lead cheques.
Where the money goes before the model ships
Cash from the raise is already committed to four things: research on video models, training and inference infrastructure, building out products, and international expansion. Alongside that, the full release of MAGI-2 is expected shortly. It is a bigger video-generation model, and it runs on a mixture-of-experts architecture rather than a single dense network.
Raising this heavily ahead of a flagship launch is the norm in Chinese video AI, where compute commitments lock in years ahead and competitors release on a monthly rhythm. Three rounds in a single calendar year also tells you something about how quickly the company has been repricing itself.
What investors are actually underwriting
Wangsu is a content delivery and cloud infrastructure provider, which makes its participation strategically interesting rather than purely financial, given how much inference capacity video models consume. Cornerstone Capital brings the growth-stage cheque.
Sand.ai has not disclosed the revenue picture, customer numbers or the valuation attached to the unicorn claim. For founders watching the sector, the more useful signal is the deal structure: capital is still flowing to video generation in China, but it is concentrating in a small number of names that can credibly promise frontier-scale training runs.
