產業脈動｜具身智能的技術突破與臺灣機器人產業的機會

在模型、算力與資料逐漸成熟下，具身智能正從研究與示範階段邁向可大規模部署的應用階段。 國際具身智能發展有三個方向，首先是核心技術的突破，包括多模態感知、世界模型、強化學習等技 術，使機器人具備更高程度的自主決策與運動控制能力。其次是AI 平台企業建立機器人基礎設施，例如 NVIDIA 的邊緣AI 晶片、機器人開發平台與數位模擬環境。最後是通用機器人的發展，尤其是人型機器 人成為全球科技與新創企業競逐的焦點。臺灣具備半導體與AI 晶片、精密機械與機器人關鍵零組件製造 等基礎，有機會在下一波全球具身智能產業競合中扮演關鍵角色。

With the maturation of AI models, computing power, and data, embodied AI is transitioning from research and demonstration toward large-scale deployment. Globally, its development is driven by three key directions: breakthroughs in core technologies such as multimodal perception, world models, and reinforcement learning, which enable robots to achieve higher levels of autonomous decision-making and motion control; the emergence of AI platform companies building foundational infrastructure for robotics, including edge AI chips, platforms, and simulation, exemplified by NVIDIA; and the rapid advancement of general-purpose robots, particularly humanoid robots, which have become a major focus for global technology firms and startups. Taiwan possesses strong capabilities in semiconductors, AI chips, machinery, and key robotic components manufacturing, positioning it well to play a critical role in the next wave of global embodied AI competition and collaboration.