BofA: Humanoid Robotics Moves Toward Real-World Deployment as WAM and Lower-Cost Data Collection Gain Ground
A BofA Securities report on China's industrial sector says humanoid-robot applications in 2026 are expanding beyond education and research and development into industrial, logistics, retail and household settings.
A BofA Securities report on China's industrial sector says humanoid-robot applications in 2026 are expanding beyond education and research and development into industrial, logistics, retail and household settings.
World action models, or WAMs, have become the mainstream architecture because they incorporate learning of physical laws. Competition in data is shifting toward real-world collection, with UMI and egocentric technologies significantly lowering costs. Leading companies are targeting the collection of more than 1 million hours of high-quality data.
The cost of dexterous hands is falling rapidly, while high-end models with more than 20 degrees of freedom are becoming more widely available. Tactile sensors have become a key input for VTLA models.
BofA's one-sentence conclusion is that humanoid robots are moving from proof of concept toward deployment across multiple settings. The report identifies WAM architectures, lower-cost data collection and advances in dexterous-hand technology as the three core drivers, and says the industry is approaching commercialization.
The report identifies China's robotics supply chain—including dexterous-hand, sensor and actuator companies—as a potential beneficiary. It also says market attention on humanoid robots is currently extremely high, while the pace and scale of actual commercial deployment remain points of disagreement in the market.
Potential catalysts identified by the report are production timelines from leading humanoid-robot companies, progress in the scale of real-world data collection and continued declines in dexterous-hand costs. Indicators to watch include robot shipments, hours of data collected and changes in dexterous-hand prices.
World action models, or WAMs, have become the mainstream architecture because they incorporate learning of physical laws. Competition in data is shifting toward real-world collection, with UMI and egocentric technologies significantly lowering costs. Leading companies are targeting the collection of more than 1 million hours of high-quality data.
The cost of dexterous hands is falling rapidly, while high-end models with more than 20 degrees of freedom are becoming more widely available. Tactile sensors have become a key input for VTLA models.
BofA's one-sentence conclusion is that humanoid robots are moving from proof of concept toward deployment across multiple settings. The report identifies WAM architectures, lower-cost data collection and advances in dexterous-hand technology as the three core drivers, and says the industry is approaching commercialization.
The report identifies China's robotics supply chain—including dexterous-hand, sensor and actuator companies—as a potential beneficiary. It also says market attention on humanoid robots is currently extremely high, while the pace and scale of actual commercial deployment remain points of disagreement in the market.
Potential catalysts identified by the report are production timelines from leading humanoid-robot companies, progress in the scale of real-world data collection and continued declines in dexterous-hand costs. Indicators to watch include robot shipments, hours of data collected and changes in dexterous-hand prices.