Physical AI opens a new era in robotics, according to research

According to the analysis sent to MTI, physical artificial intelligence includes AI systems that enable machines to independently sense and interpret the physical environment and then react to changes in real time. These capabilities are now appearing not only in software applications, but also in new-generation robots, autonomous vehicles and advanced sensor systems. Unlike previous rule-based solutions, systems using physical AI learn from experience and continuously adapt their operations to current data and situations, they said.

The rapid development of the technology is being accelerated by several breakthroughs. Multimodal, vision-language-action (VLA) models allow robots to interpret complex environments and select the most appropriate actions for a given situation – in a way similar to human thinking. Built-in neural processing units (NPUs) ensure fast, low-latency processing directly on the machines, reducing dependence on cloud-based systems. All this is further strengthened by the development of robotic hardware, such as more advanced computer vision and more sophisticated sensors. Together with increasingly favorable economic conditions, these factors are bringing physical AI closer to industrial scale.

However, the report says that the widespread adoption of physical AI remains hampered by a number of challenges. One of the most important technical issues is addressing the so-called “reality gap“: ensuring that systems trained in simulated environments also operate reliably in the real world. In addition, organizations must navigate a complex regulatory environment, manage large amounts of different types of data, and address issues of human acceptance – especially addressing concerns about the future of work.

Despite the challenges, physical AI is increasingly being used in more areas: in healthcare, AI-driven robotic surgery solutions and autonomous imaging devices are helping to address labor shortages and increase the accuracy of care. In the energy sector, drones are being used to monitor the grid, and in the longer term, robots are planned to be deployed in dangerous field work to protect human lives. In public services, AI-based drones and autonomous vehicles contribute to assessing the condition of infrastructure, they wrote.

According to the analysis, humanoid robots represent the next major development step in physical AI. Their human-like structure and mobility allow them to operate effectively in existing environments designed for people – such as production halls or even homes – without significant infrastructural transformations.

According to analyst forecasts, the spread of humanoid robots may accelerate significantly in the next decade: according to estimates, they may appear in workplaces in the millions, while the market value may increase to several thousand billion dollars by 2050. Warehousing and logistics are already a prime testing ground, as these sectors are experiencing both a persistent labor shortage and the need to automate high-precision, physical tasks, according to the Deloitte Tech Trends 2026 research.