LatentVerse Raises Several Hundred Million RMB Seed Round With Hillhouse
The AMW Read
The reported seed round adds a new software-first Chinese entrant to embodied-model development, but undisclosed valuation and performance evidence limit its near-term sector-wide significance.
Robotics / Embodied AI
LatentVerse Raises Several Hundred Million RMB Seed Round With Hillhouse
LatentVerse, a Beijing-based embodied foundation-model startup, has reportedly completed a seed round worth several hundred million RMB, according to the source article. Hillhouse Venture, ClearVue Capital, Inno Angel Fund, and embodied-AI companies AgiBot and Robot Era were named as investors. The company was founded less than two months ago by a team from Tsinghua University's Institute for Interdisciplinary Information Sciences, led by doctoral researcher Hu Yucheng. The article says several first-tier VCs are already lined up for a next round, but discloses no valuation, ownership terms, or closing date.
LatentVerse is making a software-first bet on physical AI. Its proposed UTAM, or Unified Tactile-Action Model, would train intent understanding, world-state prediction, robot action generation, and tactile feedback inside one embodied-native model. This departs from treating a robot as a fixed hardware platform with separate perception and control stacks. It does not yet establish technical leadership: the source itself notes that world-model definitions, benchmarks, architectures, and data practices remain unsettled. What the financing does show is another China-based entrant seeking to make robotics systems generalize beyond narrowly scripted industrial tasks.
For builders and investors, the concrete test is whether LatentVerse can turn its stated data recipe—Internet-scale data, human-hand data, non-robot capture, and tactile signals—into repeatable real-world task performance and recovery from slips or object deformation. Backing from two robotics companies matters only if it creates a defined route to data, hardware integration, or deployments; the article discloses none. Until that evidence appears, diligence should emphasize task success in the physical world, data provenance, and the cost of collecting the tactile data its approach requires.