Delta Intelligence Secures 500 Million Yuan to Accelerate Humanoid Foundation Models

Founded in January 2026 by a trio of UCLA Ph.D.s, the startup is building a "full-body intelligence" foundation model for humanoids. The check-writers are corporate venture arms of listed Chinese firms and top-tier financial investors. At this velocity, the cap-table math is the real story.

Following the money

The investor bench is thick. Prior rounds have pulled in three of China's better-known humanoid players — Zhiyuan Robotics, Leju Robot, and Galaxea AI — alongside established financial shops GL Ventures, Oriza Holdings, Fosun RZ Capital, Meridian Capital, and Legend Capital. Industrial capital from the automotive and semiconductor sectors rounds out the syndicate.

Six rounds in six months means dilution is compounding fast. No valuation is disclosed in the announcement, which is itself a tell. At this cadence, the founders are likely trading equity for runway in a structure that resembles a rolling venture facility more than a conventional angel-to-Series-C arc. The question is not whether Delta raised money; it is how much of the company still belongs to the founders.

The hardware play

Bundled with the raise: the Delta D1, a head-mounted panoramic data-collection rig built on a pure-vision architecture. The pitch is straightforward — capture first-person footage and full-body joint trajectories from human wearers, then feed that data into the foundation model without a physical robot in the loop.

If the unit economics hold, this is the more interesting side of the bet. The bottleneck in embodied AI is not algorithms; it is training data, and data collection has historically required fleets of expensive hardware. A wearable that decouples data generation from the robot itself is a real cost-structure advantage — if, and only if, the resulting dataset translates into model performance. That translation has not been independently verified.

The sobering reality check

Three UCLA Ph.D.s, academic roots at Tsinghua and Peking University, and industry stints at Google Robotics, NVIDIA Research, DeepMind, and Meta. DARPA, ONR, and NSF projects on the resume. The pedigree is not in question.

The market structure is. Delta's named co-investors include direct competitors in the Chinese humanoid space — a textbook portfolio-overlap problem. Backers writing checks on both sides of a competitive race rarely see clean exits. At ~$74 million per disclosed round and six rounds already booked, the burn trajectory is steep, and the next institutional raise will need to be priced for a step-up — or force existing backers to defend their positions with pro-rata capital they may not want to deploy.

For the cap table, the clock is now ticking on whether the Delta D1 bet converts into revenue before the next round resets the ownership math.