# HKUST and CalmCar Launch Physical AI Innovation Center > Source: > Published: 2026-06-05 07:54:38.749935+00:00 # HKUST and CalmCar Launch Physical AI Innovation Center HKUST and Suzhou CalmCar Electronics Technology (CalmCar) signed a strategic agreement on June 3, 2026 to establish the **Physical AI Innovation Center** in Hong Kong (HKUST; ACN Newswire). The center pursues full-stack physical AI research spanning chips and systems, foundational models and data, privacy and safety governance, and applications in **autonomous driving**, **robotics**, and **smart manufacturing**. Prof. Guo Song, a Chair Professor in HKUST's Department of Computer Science and Engineering, will direct it; HKUST highlights a Physical Alignment approach that upgrades generative video models into interactive world models, shifting AI from seeing the world to acting on it. CalmCar serves as an initial industry test bed, and the launch created an Industry-Research Alliance and a Capital Alliance with academic, chip, and investment partners (Gasgoo; Hong Kong Business). ### What happened HKUST and Suzhou CalmCar Electronics Technology (CalmCar) signed a strategic agreement on June 3, 2026 to establish the Physical AI Innovation Center in Hong Kong (HKUST; ACN Newswire). The center is framed as a full-stack research and translation hub for physical AI across autonomous driving, robotics, and smart manufacturing, and its inauguration created an Industry-Research Alliance and a Capital Alliance. ### Technical focus Per HKUST, ACN Newswire, and Gasgoo, the center will pursue research spanning chips and systems, foundational models and data, privacy protection and safety governance, and core application scenarios. HKUST emphasizes a Physical Alignment technique that upgrades generative video models into interactive world models - in the university's framing, a shift from AI that sees the world to AI that acts on the world - trained in high-fidelity virtual environments before transfer to the real world to improve generalizability, interpretability, and reliability in open settings. ### Organization and partners Gasgoo and Hong Kong Business report that Prof. Guo Song, a Chair Professor in HKUST's Department of Computer Science and Engineering, will direct the center, with a strategic committee including academic leaders such as Zheng Weimin and HKUST Provost Guo Yike. Named collaborators include **Tsinghua University**, **Microsoft Research Asia**, **51WORLD**, **MetaX Integrated Circuits**, **Arm China**, and **Huixi Technology**; listed investors include **HSBC**, **Lion X Ventures**, **Dragonstone Capital**, and **Delian Capital**, among others (Gasgoo; Hong Kong Business). ### Industry context Editorial analysis: The center fits a growing pattern in which universities and specialized startups co-locate research, hardware, and commercialization to accelerate end-to-end stacks for robotic AI and vehicle autonomy. Multi-stakeholder hubs with industry test beds can reduce friction between simulator-driven research and on-road or on-robot validation. The emphasis on world models and physical alignment signals continued interest in simulation-first training that encodes physical causality rather than relying purely on large perception datasets, which tends to shift engineering effort toward realistic simulators, domain randomization, and system-level evaluation. ### What to watch Editorial analysis: Track concrete outputs - published benchmarks, open-source simulators or model checkpoints, and safety-governance toolkits - and case studies showing transfer from virtual training to on-vehicle or on-robot deployment, as well as any interoperable datasets or reference architectures the Industry-Research Alliance produces for others to reproduce and validate. ## Scoring Rationale A well-backed university-industry center brings a regional hub and serious partners (Microsoft Research Asia, Arm China, Tsinghua, major investors) to physical-AI and world-model research relevant to simulation-to-reality transfer and autonomy. It is notable but is a center launch rather than a frontier model or benchmark release, so it sits in the mid-notable band. Practice interview problems based on real data 1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with. [Try 250 free problems](/problems)