By 2026, the robotics sector is poised to see several isolated breakthroughs, though the overall trajectory of progress remains somewhat unpredictable. Following in-depth research on the U.S. robotics industry, Citrini Research has highlighted that the sector is not characterized by a single, monumental breakthrough. Instead, it is the accumulation of numerous intensive signals (assuming "dense" here refers to "intensive signals" or "concentrated indicators") that are collectively forming a tipping point. Presently, the robotics industry can be likened to the AI landscape prior to the advent of ChatGPT.

From a hardware perspective, robotics solutions span a wide spectrum, from basic robotic arms to sophisticated humanoid robots. The fundamental hardware for locomotion has reached maturity, with anticipated reductions in costs. Nevertheless, the design of dexterous hands remains a pivotal challenge, with approaches such as tendon-driven and direct-drive systems still under exploration, and haptic perception technology yet to achieve a significant breakthrough.

At the intelligent system level, physical AI presents even greater hurdles due to the intricacies of the real world. Its "nervous system" comprises visual language models, world models, visual language action models, and low-level control systems. The synergy between world models and action strategies poses a challenge (assuming "difficulty" here means "challenge" or "obstacle"), and the involvement of large model companies could provide fresh perspectives.

A major obstacle confronting the industry is the scarcity of training data. Currently, data is gathered through methods like capturing videos of human activities, UMI demonstrations, and teleoperation. Some firms are ramping up their investments in data and computational power, opting to initially mass-produce robots with limited capabilities to amass data.

In terms of commercialization, robots have made their way into the market through routine tasks such as folding clothes and mixed palletizing in warehouses, achieving deployment by incrementally expanding their capabilities. In the initial stages, there are predominantly combinations of models with diverse hardware, and there exists a disparity between the actual progress and reports of humanoid robots entering factories.

In the rivalry between China and the U.S., Chinese companies hold advantages in hardware manufacturing and production volume expansion, while the U.S. grapples with challenges in accumulating hardware manufacturing capabilities and supply chain uncertainties. However, firms like Tesla and Figure still possess the potential to bolster their production capabilities.

The report posits that the foundational locomotive capabilities of humanoid robots have matured, hardware costs will decline, there are multiple pathways for complex operations, simple tasks can already be executed, and progress in complex tasks may surpass expectations. Nevertheless, development hinges on the smoothness of the capital cycle. Analysts, upon witnessing the autonomously walking Figure 03 outside the Figure headquarters, were profoundly convinced that transformative changes are underway in the industry.