Embodied intelligence refers to AI systems equipped with a physical body to enhance real-world interaction capabilities. However, researchers have long avoided the topic of “machine consciousness” to sidestep controversy. With the advent of large language models (e.g., ChatGPT and DeepSeek) showing a strong desire for environmental and bodily interaction, machine consciousness becomes an unavoidable frontier issue. This article posits that machine consciousness, an AI system’s awareness of its own internal state and external environment, is crucial for autonomous adaptive decision-making. We explore the intrinsic relationship between machine consciousness and embodied intelligence, drawing on biological insights into the modular nature of consciousness and unconscious processes. A brain-inspired “consciousness space” framework is proposed to integrate multi-modal sensory information for internal simulation and contextual reasoning. We further discuss how consciousness supports the formation of decisions and how the execution of decisions validates the value of conscious processing. Finally, the paper outlines future research directions, including unified multi-modal consciousness-space modeling, active inference-based learning, and considerations of machine ethics and control boundaries.