Unitree, a frontrunner in global humanoid robot shipments, had its founder, Wang Xingxing, highlight at the 2026 World Physical AI Conference that the embodied AI industry is grappling with fundamental challenges. Despite promising advancements in physical AI, the seamless integration of AI model inputs and outputs with real-world robots remains elusive. The much-anticipated 'ChatGPT moment' for embodied AI is projected to materialize within the next one to three years, consequently postponing its deployment in both household and industrial settings. Presently, only commercial service scenarios with a high tolerance for errors are poised to adopt these technologies initially.

Moreover, the industry is significantly underinvesting in the development of the 'brains' behind embodied AI, with a substantial share of investments being funneled into hardware instead. Consequently, products are predominantly confined to low-generalization scenarios, such as demonstrations, and have yet to achieve widespread, scalable applications. Following the initial public offerings (IPOs) of leading companies, their market capitalizations have witnessed notable declines, prompting a reassessment of industry-wide valuations. Investments previously funneled into the primary market, driven by expectations of swift commercialization, are now confronted with considerably extended return periods.

Indicators of an industry downturn are becoming apparent, with numerous companies at risk of collapse before the anticipated surge in applications. Only those firms with ample cash reserves and the capability to establish genuine, revenue-generating closed-loop operations will endure until the technological breakthrough materializes.