> **来源：[研报客](https://pc.yanbaoke.cn)** # Summary of "Physical AI and the Future of Robotics" ## Core Content This document explores the evolving landscape of **Physical AI** and its implications for the future of robotics, particularly in **warehouse automation** and **supply chain** applications. It highlights how **intelligent hardware** is becoming a central driver of innovation, reshaping investment trends, and altering global trade dynamics. --- ## Main Points ### **Physical AI and Robotics: A New Era** - **Intelligence in Hardware**: Robots are now capable of generalizing in novel environments, marking a shift from traditional, task-specific automation. - **Market Expansion**: The rise of Physical AI is expanding the addressable market for robotics, with the potential for nearly every warehouse, factory, and logistics hub to adopt automation. - **VC Investment Trends**: - US VC investment in hardware is expected to reach **\$120B in 2026**, which is **one-third of all investment**. - **22%** of global VC firms now have at least **10% of their investments** in US hardware companies, up from **9% in 2022**. - **Edge computing and data centers** are seeing record investment, with **\$3.4B** in edge compute and **\$12B** in data center tech projected for 2026. ### **Supply Chain Challenges and Shifts** - **Supply Chain Pressure**: The **Global Supply Chain Pressure Index (GSCPI)** is at its highest in four years due to trade disruptions, notably the closure of the **Strait of Hormuz**. - **Oil and Semiconductor Impact**: The **oil price surge** has increased shipping costs and strained the semiconductor supply, which is critical for AI growth. - **Component Shortages**: **Japan** provides **57%** of inputs for AI accelerator chips, while **Taiwan, South Korea, Germany, and others** are also key suppliers. - **HBM Shortage**: **High Bandwidth Memory (HBM)** is a critical bottleneck, with **90% of HBM** consumed by the **top four chipmakers**. This shortage impacts both **data centers** and **physical AI devices** like **humanoid robots**. ### **AI and Compute Economics** - **Token Cost and Efficiency**: AI inference tasks vary in cost and token usage, with **simple tasks** like drafting an email costing **\$0.05** and **complex tasks** like analyzing a 10-K filing costing **\$15**. - **Latency and On-Device Compute**: Physical AI systems require **low-latency processing**, often necessitating **onboard compute hardware**, which is expensive and in high demand. - **Scaling Challenges**: Physical AI systems must handle **variability and exceptions**, which are common in real-world logistics environments. ### **Warehouse Automation Landscape** - **Current Adoption**: A **median of 40%** of large warehouses have some level of automation, while **one-third** have less than **25%**. - **AI Adoption**: AI is being adopted in **lower-friction workflows** like reporting and labor planning, but **robotic picking, palletizing, and flexible robots** are still in **pilot or evaluation** phases. - **Key Technologies**: - **Conveyor and Sortation**: **90%** of executives report full scaling. - **AS/RS or Goods-to-Person**: **75%** of executives report full scaling. - **AMRs**: **66%** of executives report scaling. - **Robotic Picking and Palletizing**: **66%** of executives are piloting. - **Computer Vision and Scanning**: **75%** of executives are evaluating or piloting. ### **Challenges in Automation Adoption** - **Operational Complexity**: Automation systems must integrate into **live logistics environments**, handling **variability, exceptions, and peak periods**. - **Reliability and Payback**: While **reliability and savings** often exceed expectations, **ease of implementation and flexibility** remain major challenges. - **Humanoid Robots**: Despite hype, **humanoid robots** are still **overhyped**, with **limited current operational value**. --- ## Key Information - **Global Supply Chain Pressures**: The closure of the **Strait of Hormuz** has had a ripple effect on **global trade**, increasing **shipping costs** and **inflation**. - **VC Investment Trends**: The **US hardware VC investment** is expected to **double in 2026**, reaching **\$120B**. - **HBM Shortage**: **HBM** is a **primary chokepoint** in chip production, with **90% of HBM** consumed by the **top four chipmakers**. - **Physical AI Models**: Emerging models like **Vision-Language-Action (VLA)** and **Neural World Models** are enabling **generalized robotics** and **real-world adaptability**. - **Automation Barriers**: **Cost and payback** are the **biggest hurdles** to automation scaling, even though **reliability and efficiency** are often met or exceeded. --- ## Conclusion The convergence of **AI, robotics, and edge computing** is redefining the **hardware landscape**, with **Physical AI** at the forefront. While **warehouse automation** is still in its early stages, **VC investment** is growing rapidly, signaling a **shift in focus** toward **intelligent, adaptable hardware**. However, **real-world implementation** remains challenging due to **supply chain constraints**, **latency requirements**, and the **complexity of integration**. The future of robotics lies in **generalized, intelligent systems** that can **operate in diverse environments** and **deliver tangible economic benefits**.