Sand.ai valued above $1 billion in Series A extension
Beijing-based video-generation startup Sand.ai has completed a Series A extension that valued it at more than $1 billion, DealStreetAsia reported. The round was jointly led by CoStone Capital and Wangsu Science & Technology.

Key takeaways · 4
- 01
The Series A extension valued Sand.ai above $1 billion, and CoStone Capital and Wangsu Science & Technology jointly led it.
- 02
Sand.ai’s fundraising over 10 months reached $300 million; the extension was its third funding round of 2026.
- 03
Planned uses for the funding include inference infrastructure and video foundation-model research and training.
- 04
Sand.ai distinguishes its open-source MAGI-2 Preview research release from an official version it plans to introduce in the near term.
A broad investor group joined the round
CoStone Capital and Shenzhen-listed Wangsu Science & Technology jointly led the extension.[1] DealStreetAsia reported participation from industry, government funding, Chinese yuan funds and US dollar investors.[1] Pudong Venture Capital Group and an investment arm of Shanghai International Group also participated.[1] Existing investors iMPC, CAS Star and HyT Capital reinvested; iMPC was formerly known as Matrix Partners China.[1] The range of participants gives professionals a view of who backed this financing, without indicating how much each investor contributed.
Funding follows a rapid fundraising period
The extension was Sand.ai’s third funding round of 2026, and its fundraising over 10 months totaled $300 million.[1] Less than five months earlier, the company had raised more than $100 million across two rounds.[1] The evidence does not state the size of this extension, so the reported cumulative fundraising and valuation should not be read as the amount raised in this round.
Models and a paid creation service
Sand.ai’s model lineup includes Magi-1, an autoregressive video-generation model, and GAGA-1, which is designed for simultaneous audio-visual generation.[1] MAGI-2 Preview is an open-source research release; Sand.ai describes it as a 114-billion-parameter unified audio-video model that activates six billion parameters per token.[1][2] Separately, the company offers a paid AI video-creation service with tools for music videos, product ads and anime dramas.[1] VidMuse exceeded $10 million in annual recurring revenue within two months of its January launch.[1]
The next steps are research and expansion
Sand.ai plans to use the new funding for video foundation-model research and training, inference infrastructure, product iteration and global expansion.[1] It also plans to introduce the official version of MAGI-2 in the near term.[1] The company says its goals are to lower costs, accelerate turnaround times and democratize high-quality video creation.[1] For teams assessing the products, the distinction between a paid service and a research release is relevant: the evidence identifies MAGI-2 Preview as research, not as the official version.[1]
The financing gives Sand.ai resources it says it will direct toward model development, inference capacity and global expansion. Professionals assessing AI video tools can distinguish its paid creation service from its open-source research release and watch for the planned official MAGI-2 version.
Why it matters
Bring this to your team — one session a day, built for your industry.
kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.
For teamsHow this developed
8 November 2026
Sand.ai valued above $1 billion in Series A extension