Chinese video-generation startup Sand AI announced a $200 million Series A+ funding round on November 5, a raise that the company says pushes it into the unicorn club and brings its total fundraising this year to $300 million across three rounds. The announcement, reported by TechNode and covered by outlets including DealStreetAsia, TNGlobal and Ventureburn, lands as the company prepares to ship the full version of MAGI-2, a larger video-generation model built on a mixture-of-experts architecture.
The round was jointly led by Cornerstone Capital and Wangsu Science & Technology, a publicly listed Chinese cloud and communications company that has emerged as one of Sand AI's most committed backers. According to TechNode, which cited a report from the Chinese outlet Jiemian, the new capital will fund video-model research, training and inference infrastructure, product development, and international expansion.
Why Investors Are Betting on Open-Source Video AI
Sand AI's rise has been unusually fast. Coverage of the round points out that the company has closed three funding rounds and accumulated $300 million in only about ten months, a pace that few Chinese AI startups have matched this year. A BigGo Finance analysis of the round framed it as Wangsu Technology making a deliberate bet of 300 million yuan on Sand AI's approach to video generation.
Series A+ rounds are an extension mechanism that is common in venture funding: rather than opening a new lettered round, a company tops up its existing Series A with fresh capital, typically after early momentum has validated the original investment thesis. For Sand AI, closing an extension at this scale — instead of jumping straight to a Series B — suggests its existing backers were willing to double down on an approach that is already attracting attention, even before the full MAGI-2 release gives them commercial results to point to.
The company's pitch is also distinctive. While most of the biggest names in AI video — from Kuaishou's Kling to ByteDance's Seedance — operate closed, invitation-style platforms, Ventureburn's coverage of the round describes Sand AI as challenging those incumbents with an open-source approach to AI video. That positioning has made Sand AI one of the most closely watched of the new wave of Chinese video-generation companies, and it gives developers and researchers direct access to model weights that competitors keep behind APIs.
Open-source approaches are reshaping many corners of the AI industry, and you can follow our breaking AI news coverage as the competition between open and closed models intensifies.MAGI-2: A Mixture-of-Experts Bet
The timing of the funding round is not accidental. Sand AI says the full version of MAGI-2 is coming soon, and TechNode reports that the new model is a larger video-generation system built on a mixture-of-experts, or MoE, architecture.
Mixture-of-experts designs split a model into many specialized sub-networks, activating only the relevant ones for a given task. In video generation, that matters because producing coherent, high-resolution footage is one of the most compute-hungry workloads in generative AI. An MoE architecture can, in principle, deliver the quality of a much larger model without paying the full inference cost of running every parameter on every frame — an advantage for a startup that needs to serve video generation economically at scale.
Sand AI has not announced an exact release date for the full MAGI-2 model, and the company has not published benchmark results for it yet. What is clear from the funding announcement is that the raise is meant to bankroll exactly that kind of heavy lifting: the stated use of proceeds includes model research and the training and inference infrastructure that a frontier video model requires.
What the Money Means for the Video AI Race
The round arrives at a moment when AI video generation has become one of the most competitive segments of the entire industry. Chinese platforms like Kling and Seedance have built large user bases with high-quality output, while Western labs continue to iterate on their own video systems. Into that landscape, Sand AI's combination of an open-weight strategy and fresh unicorn-scale capital is a notable differentiator.
For enterprise watchers, Wangsu Science & Technology's involvement is arguably the most interesting detail. A listed infrastructure and cloud company co-leading a $200 million round signals that established Chinese tech firms see commercial value in owning a piece of the video-generation stack rather than merely renting it from hyperscalers. If MAGI-2 performs as promised, Sand AI's models could end up embedded in media workflows, advertising pipelines and app features far beyond its own products.
There are open questions, of course. Open-source video models raise the same dual-use concerns that have followed open-weight language models, and it remains to be seen what safeguards Sand AI ships alongside the full MAGI-2 release. The company's international expansion plans will also test how far an open-source Chinese video model can travel with Western developers and businesses.
For now, the funding gives Sand AI the resources to answer those questions with shipping products rather than promises. The next milestone to watch is the full MAGI-2 launch — and with $200 million in new backing, the company has little reason to delay it.
Stay Ahead of AIThe AI industry moves fast, and funding rounds like this one reshape the competitive landscape overnight.
Read more AI news →
