Manycore's spatial-intelligence world model enters internal beta
The company behind Kujiale — one of China's biggest interior-design platforms — just made its clearest declaration yet that it wants to be infrastructure for the spatial-AI era, not just a design-software vendor. Its first interim results since listing show the AI business is now the only part of the company growing fast.
Manycore Tech said its self-developed world model has entered internal beta, anchoring a pivot the company itself frames as becoming the "water seller" of spatial intelligence — selling the picks and shovels to whoever builds the AI that understands physical space. The world model takes what the company calls an explicit 3D approach: instead of predicting the next frame of a 2D video like many rivals, it generates full 3D geometry, materials, and spatial relationships, a bet that a model grounded in real 3D structure will reason about the physical world more reliably than one that only learns pixel patterns.
The financials explain why the pivot is accelerating. In the first half of 2026, Manycore posted revenue of roughly 405 million yuan (about $60.4 million), up only 1.5 percent year over year, while adjusted net profit surged 210.9 percent to about $8.3 million. The standout line is the AI segment: revenue from new AI applications and products jumped 177 percent to about $4.6 million. Its LuxReal AI-video platform, launched in May, grew monthly registered users more than 270 percent month over month and monthly revenue more than 470 percent — early, but real adoption. The company says its GPU cluster now processes about 12.9 million computing tasks a day and that its ecosystem consumed roughly 2.4 billion tokens in July.
The world-model bet is where this gets interesting beyond the earnings math. Manycore has mapped a staircase from spatial reconstruction through understanding and generation toward action, and it has shipped the pieces — the Lux3D generation model, the SpatialVerse synthetic-data business, the Aholo platform. Its chief scientist argues world models are now splitting into competing technical paths, and that 3D-explicit generation is the more defensible one. The risk is that the company has not yet published independent benchmarks for Lux3D, so "enters internal beta" is still a promise until outside researchers or paying customers can measure the quality. Chinese media also reported the shares could enter Hong Kong's Stock Connect program in early September, which would broaden the investor base watching this story.
What to watch: whether Manycore releases public benchmarks for Lux3D and the world model — and whether the second half turns that 177 percent AI growth into a meaningfully larger share of revenue.
Do you buy the argument that world models need explicit 3D structure rather than just more video data? Tell us in the comments.
