Image: Sand.ai — official site visual (sand.ai).
📌 KEY HIGHLIGHTS
- Sand.ai closed a $200 million Series A+ round on 5 November 2026, co-led by CoStone Capital and Wangsu Science & Technology.
- The round takes the company's fundraising in 2026 to about $300 million across three rounds, and the company says it is now valued above $1 billion.
- Sand.ai builds MAGI, a family of video-generation models that generate video chunk by chunk, like a language model predicts words — instead of the diffusion approach most rivals use.
- Its MAGI-2 Preview is a 114-billion-parameter mixture-of-experts model that generates video and audio together; the full MAGI-2 release is expected soon.
- The money goes to model research, training and inference infrastructure, product development and international expansion.
What happened?
Beijing-based Sand.ai announced on 5 November 2026 that it had completed a $200 million Series A+ financing round. CoStone Capital and Shenzhen-listed Wangsu Science & Technology jointly led the round, with Times Capital, Panlin Capital, Orient Securities Capital, Pudong Venture Capital and Shanghai Guofang among the other participants, according to TechNode and DealStreetAsia.
This is Sand.ai's third funding round of 2026, and the company says it brings its total fundraising over the past ten months to about $300 million. The company also says the round pushed its valuation above $1 billion, making it a unicorn. Both the valuation and the cumulative fundraising figures are company-supplied and have not been independently verified — no exact valuation was disclosed.
Sand.ai was founded in early 2024 by Cao Yue, a Tsinghua PhD and former principal researcher at Microsoft Research Asia who co-authored the Swin Transformer, a widely used computer-vision architecture. In under three years the startup has gone from a research project to one of China's most talked-about AI video companies.
| Detail | Sand.ai Series A+ |
|---|---|
| Amount raised | $200 million |
| Round | Series A+ (third round of 2026) |
| Lead investors | CoStone Capital, Wangsu Science & Technology |
| Valuation | Above $1 billion (company-reported, not independently verified) |
| Total raised in 2026 | About $300 million across three rounds |
| Founded | Early 2024, Beijing |
| Founder | Cao Yue, ex-Microsoft Research Asia |
Sand.ai $200 million funding: what the money is for
Sand.ai says the proceeds will fund video foundation-model research, training and inference infrastructure, product development and international expansion. The most concrete near-term item on that list is the full release of MAGI-2, the company's next video model, which is expected soon.
Wangsu's part of the deal was disclosed before the round closed. In a 29 September filing with the Shenzhen Stock Exchange, the cloud and network infrastructure company said it planned to invest RMB 300 million (about $44.5 million) through a convertible note and warrants, for roughly 4.4% of Sand.ai once the other investors completed their subscriptions. Wangsu also plans to work with Sand.ai on low-latency video inference at the edge.
The other side of the ledger is worth stating plainly. The same Wangsu filing shows Sand.ai Group recorded revenue of 17.4 million yuan (about $2.6 million) in the first half of 2026, against a net loss of about 152 million yuan (about $22.8 million), both unaudited. That is the classic shape of a high-valuation, high-burn AI startup: the $1 billion-plus price tag is a bet on the technology and the market, not on current earnings.
Illustrative image: a video production camera on set — Wikimedia Commons / Don Ramey Logan.
What is MAGI-2 — and why investors bet on it
Sand.ai's pitch is a different way of generating video. Most leading video models use diffusion: they start from noise and refine it into a clip. Sand.ai's MAGI models are autoregressive — they generate video chunk by chunk, each piece conditioned on what came before, much like a language model predicts the next word. Founder Cao Yue has argued this makes long, continuous video easier to build and extend.
The first model, MAGI-1, was released as an open-source project in 2025 and won the Physics-IQ video continuation benchmark. MAGI-2 Preview followed in August 2026: a 114-billion-parameter mixture-of-experts model that activates only around 6 billion parameters per token, so it costs far less to run than its size suggests. It generates video and synchronised audio together in a single pass, producing 10-second clips upscaled to 1088×1920, and its weights, inference code and technical report are public.
Sand.ai claims a distilled MAGI-2 can produce a 10-second 1080p clip for about 0.5 yuan — roughly a tenth of the cost of mainstream models. That cost claim is the company's own and has not been independently audited, but it is the heart of the investment case: if video generation gets cheap enough, it stops being a demo and becomes a production tool.
On the product side, Sand.ai runs VidMuse, a paid AI video creation service for music videos, advertisements and animated content, launched in January 2026. The company says VidMuse passed $10 million in annual recurring revenue within two months of launch; that figure is also company-reported and unaudited.
Why it matters for readers in India
AI video is becoming a creator tool, not just a research demo, and India has one of the world's largest creator economies. Cheaper, open-weight video models mean Indian YouTubers, advertisers and small studios can generate product ads, music videos and explainers without a shoot — or remix open models like MAGI-2 Preview for their own workflows. The full MAGI-2 release, when it lands, is worth watching for that reason.
The competitive picture matters too. Sand.ai is chasing Kling from Kuaishou and Seedance from ByteDance, both backed by far larger companies. Its answer is openness and cost: open weights, a public technical report, and a model designed to run cheaply. If that strategy works, the next wave of AI video tools may come from a startup, not a giant — and Indian developers can build on the open releases today, without waiting for an invite or an enterprise contract.
The funding also shows where the money is flowing in late 2026: not only to the companies that train the biggest models, but to the ones that run them. Fal's $1.3 billion Series F at a $16 billion valuation, covered here earlier this week, is the same bet — inference, the business of running AI models for other people, is where investors are writing the largest cheques.
Should you care?
FAQs
What is Sand.ai?
Sand.ai is a Beijing-based AI startup founded in early 2024 by Cao Yue, a former Microsoft Research Asia principal researcher. It builds the MAGI family of video-generation models and VidMuse, a paid AI video creation service for music videos, advertisements and animated content.
How much funding has Sand.ai raised?
Sand.ai announced a $200 million Series A+ round on 5 November 2026, its third round of the year. The company says its total fundraising over the past ten months is about $300 million, and that the round took its valuation above $1 billion. Both figures are company-supplied.
What is MAGI-2?
MAGI-2 is Sand.ai's video-generation model. The preview released in August 2026 is a 114-billion-parameter mixture-of-experts model that activates about 6 billion parameters per token and generates video and synchronised audio together. Its weights, inference code and technical report are public, and the full version is expected soon.
Who invested in Sand.ai's Series A+ round?
CoStone Capital and Wangsu Science & Technology jointly led the round. Other participants included Times Capital, Panlin Capital, Orient Securities Capital, Pudong Venture Capital and Shanghai Guofang, alongside existing investors such as Matrix Partners China.
Takeaway: Sand.ai turned $200 million of new funding into unicorn status in under three years — but the number to watch is not the valuation, it is MAGI-2. If the full release keeps the preview's quality at the claimed cost, AI video gets meaningfully cheaper for everyone.
Developers can try the open MAGI-2 Preview weights today from Sand.ai's site, and creators should watch for the full MAGI-2 release in the coming weeks — it is the test of whether this $200 million bet pays off.
Sources
- Sand.ai raises $200 million ahead of MAGI-2 video model release — TechNode
- Chinese AI video generation startup Sand.ai enters unicorn club after $200m round — DealStreetAsia
- Sand.ai Raises $200 Million To Rival Kling, Seedance AI Video — VentureBurn
- Startup Funding News Today, November 5, 2026 — TechStartups
- Sand.ai — official website
Comments
Post a Comment