On November 13, Stanford Professor Fei-Fei Li published a comprehensive article positioning spatial intelligence as "the next frontier of AI," highlighting its role as the key driver in advancing AI from "language intelligence" to "physical intelligence." This assertion has further heightened market interest in physical AI. According to a November 14 research report by Shenwan Hongyuan, physical AI—a systematic integration of world models, physics simulation engines, and embodied intelligence—is becoming the cornerstone for digital twins and embodied intelligence applications.
Among the most promising applications of physical AI are autonomous driving and embodied intelligence. Smart warehousing, with its mature commercialization, strong demand, and clear value realization, has emerged as a breakthrough for scaling embodied intelligence and physical AI technologies. Consequently, Shenwan Hongyuan identifies GEEKPLUS-W (02590), a global leader in warehouse robotics, as a core player in the physical AI industry chain, particularly as an "equipment manufacturer" that bridges industry expertise with physical execution logic to drive real-world applications and value creation.
**Commercialization of Physical AI** Shenwan Hongyuan emphasizes that physical AI relies on the synergy of three technologies: world models, physics simulation engines, and embodied intelligence controllers. The latter serves as the critical link between virtual reasoning and physical execution. GEEKPLUS-W aligns with this trend through its technological breakthroughs. On October 28, the company unveiled the world’s first embodied intelligence-powered "unmanned picking workstation" and a fully automated picking solution, marking a leap toward "truly unmanned" smart warehousing.
Powered by its proprietary embodied intelligence base model, Geek+ Brain, the workstation adopts a "perception-strategy" hierarchical architecture. It accurately identifies objects and environmental data via perception models while generating optimal grasping solutions in real time through strategy models, achieving a closed-loop "virtual decision-physical execution" process that aligns with physical AI’s systemic logic. The model, pre-trained on millions of real-world warehouse data points and validated through multi-layered testing (internet data, simulation data, and real-machine data), offers three core advantages: - Supports tens of thousands of SKUs without post-training, enabling rapid deployment (48 hours) for complex categories. - Delivers millisecond-level inference for 24/7 operations, doubling picking efficiency over manual labor, with ROI achieved in 1–2 years. - Ensures data security compliance for industries like pharmaceuticals and retail through localized data processing.
To date, GEEKPLUS-W has served over 850 global clients across 40+ countries, maintaining its position as the world’s largest warehouse fulfillment robotics provider for six consecutive years.
**Commercial Strength Solidifies Industry Leadership** As a core equipment manufacturer in the physical AI chain, GEEKPLUS-W’s technological prowess and market position are widely recognized. Its R&D team comprises top-tier academic talent, with patents covering robotics hardware/software and AI algorithms. Globally, the company has established local teams in key markets (Europe, the Americas, and Asia-Pacific), backed by 52 service sites and 12 spare parts centers offering 24/7 support—a network that underpins its commercial scalability.
This global-local strategy has translated into robust financial performance. In 2024, GEEKPLUS-W reported revenue of RMB 2.41 billion, leading Hong Kong’s robotics sector. Growth continued in H1 2025, with revenue reaching RMB 1.025 billion (up 31.0% YoY), while adjusted EBITDA turned positive for the first time at RMB 11.621 million—a milestone unmatched by peers. Customer loyalty further underscores its market strength: the 2024 repurchase rate stood at 74.6% (84.3% for key clients), rising to over 80% in H1 2025.
Looking ahead, founder and CEO Yong Zheng stated that after achieving full unmanned picking, the next goal is to automate robotic packing, advancing toward a "fully unmanned warehouse."
**Conclusion: The "Warehousing Pathway" for Physical AI Commercialization** From enabling embodied intelligence to upgrading smart warehousing, the industry trend underscores "technology deployment + scenario refinement." GEEKPLUS-W’s embodied intelligence-driven, globally supported solutions exemplify the "warehousing pathway" for physical AI. As technology evolves, the realization of fully unmanned warehouses will unlock greater industry value, with GEEKPLUS-W at the forefront of this transformation.
Comments