PixVerse hits $2B on a bet that AI video labeling beats raw data

Singapore-based PixVerse just locked in a $2 billion-plus valuation, closing an extended Series C at $439 million. Alibaba, Lollapalooza Capital, and Mirae Asset led the round, signaling that even as AI video gets crowded, serious money still sees room for another winner—if the technical foundation is right.

The two-year-old startup, founded by ex-ByteDance engineer Wang Changhu and Jaden Xie, claims 150 million registered users and 15 million monthly actives across its video generation, film production, and world modeling tools. Co-founder Xie didn’t mince words with TechCrunch, arguing only a handful of companies can actually deliver the quality needed. He pointed to OpenAI shelving Sora 2 and Meta and Tencent failing to build high-quality video models as proof the field is thinner than it looks.

What investors are really buying here is Wang’s background. He built the visual AI tech that powered TikTok’s recommendation engine at ByteDance—a system that didn’t just hoover up data but labeled it with the kind of precision that fed eerily accurate recommendations. That labeling philosophy now underpins PixVerse’s approach to video generation. In a space where everyone’s scrambling for training data, the thesis is that how you structure that data matters more than how much you have.

Whether that edge holds is an open question. ByteDance itself, Midjourney, Google, Runway, and Luma are all pushing video and world models forward at roughly the same clip. A $2 billion valuation says PixVerse has something real. But in a field this fast, today’s differentiator can become tomorrow’s table stakes before the ink dries on the term sheet.