At the 2026 World Robot Conference Forum held in Beijing from August 19 to 22, Gao Jiyang, founder and CEO of Star Chart AI Technology Co., Ltd., stated that as intelligence becomes more universal—expanding from single-scenario utility to multi-scenario and cross-task applications—the core business model of the embodied intelligence industry will undergo another transformation. This shift will center on the sale of physical world tokens as the primary revenue stream.

Throughout this evolution, embodied intelligence encompasses both complete machines and the intelligence itself, with overall value gradually migrating from hardware to the intelligence layer. Gao noted that over the past year and into this year, gross margins for selling various complete machine products across the industry have ranged from 40% to 60%. In the second phase—the solution subscription stage—gross margins could decline to around 20%. By the time true token sales become the norm, complete machines may even operate at negative gross margins.

"We need more terminals deployed in the physical world to generate greater token consumption for embodied intelligence foundation models," Gao explained. "This represents the entire industrialization process of embodied intelligence, driven by model capabilities."