By Anjali Bastianpillai, senior client portfolio manager in Pictet’s thematic equities team
AI has already transformed the digital economy and is now starting to change the physical one, too. The technology is being embedded in an increasing number of machines that can sense, move, and operate in a physical space.
Physical AI combines advanced computing models with sophisticated hardware: sensors, actuators (which convert electrical signals into physical motion), processors and machine vision. Together, they create robots and autonomous systems that can operate safely and productively in real environments.
Humanoid robots and autonomous vehicles are the most visible examples of this tech in action. The autonomous vehicle market (including public transport) is estimated to grow to $1.2trn by 2040, thanks to a compound annual growth rate (CAGR) of 44%1.
As the technology continues to mature, the rate of adoption is accelerating: the proportion of US rideshare miles completed by robotaxis is forecast to increase from 0.4% in 2025 to 7.5% in 2030.
We also expect to see autonomous trucks and public transport, as well as vehicles used in industry, delivery, mining, defence, farming and logistics. That, in turn, will boost the growth of related industries such as mobility as a service (MaaS), software licensing and fleet management.
Physical AI applications: Robots in healthcare and industry
Many other applications of the technology are already commercially available and highly scalable. According to consultancy PwC, humanoids and autonomous driving will account for only half of the global physical AI market by 2030.
In healthcare, robotic surgery systems illustrate how AI-enhanced robotics can deliver precision, consistency and improved patient outcomes at scale. Robots built by Intuitive Surgical, one of the leading companies in this field, treated more than 3.1 million people worldwide in 2025. With its da Vinci systems surgeons can use 3D vision and a magnified view to control scissors, scalpels or forceps in robotic hands and perform minimally invasive procedures in areas such as urology.
See also: Funding as the new AI bottleneck: What Alphabet’s move reveals
Applications in industrial settings may be less delicate in nature, but they are equally impactful. Companies such as Teradyne are driving the adoption of collaborative robots on factory floors, which can work safely alongside humans. Cobots uncomplainingly take on tasks that are dangerous, dirty or simply dull. Because they do not need light or warmth, they can also help reduce energy use.
Elsewhere, in advanced industrial automation, machine vision is a particularly powerful aspect of physical AI. Suppliers such as Cognex and KEYENCE provide the “vision and perception” layer that allows robots to understand and navigate complex environments. Their AI-based 2D and 3D vision enables robots to recognise and localise specific parts, inspect quality, and perform pick-and-place and assembly tasks in logistics, warehouses or factories.
Why physical AI matters: Labour, reshoring and productivity
These advances in physical AI come at a very opportune time. Many industries face rising labour shortages due to deep-rooted demographic trends such as ageing populations. The US, for example, faces a labour shortage of some 2 million people by the end of this decade2.
As workforces age and labour participation rates plateau, demand for automation rises, with robot density already climbing rapidly in countries such as China. Its factories now “employ” over 2 million robots – twice as many as in 2021.
Rising labour costs since the early 2000s, particularly in China, further tilt the economics in favour of collaborative and humanoid robots, which can deliver higher productivity and lower long-term unit costs. In markets where productivity growth has stalled, physical AI offers a credible path to re-acceleration.
At the same time, robotics can support the reshoring and regionalisation of supply chains in response to geopolitical tensions and the desire for resilience.
Opportunities and bottlenecks in robotic advancement
The development of physical AI is not without obstacles. Two bottlenecks stand out: technology and regulation.
Regulators and the public must be convinced that these systems are cost-effective and meet rigorous safety standards.
Hardware and software must become cheaper, more reliable and safer to support mass deployment, particularly in safety-critical applications like driverless cars or healthcare.
For progress to continue, the industry also needs plentiful supply of semiconductors. While this is a potential bottleneck, it’s also a great opportunity for innovation and growth.
See also: Paddy Ponders: How ‘against the grain’ is contrarianism?
To build more autonomous vehicles, ever more powerful and efficient chips are needed to process vast sensor data streams and make real-time decisions. We believe the automotive semiconductor market has the potential to reach hundreds of billions of dollars over the coming decades.
Analogue and power semiconductor leaders such as Infineon and Microchip provide power management, motor control, sensing and edge-AI chips, which are essential for real-time motion, safety and perception in humanoids and autonomous machines.
Despite all these challenges, the direction of travel is clear: as long as the structural drivers of labour scarcity, reshoring, and rising costs persist, demand for physical AI solutions should continue to build.
1 Bank of America
2 OECD, Deloitte, FRED, Goldman Sachs Research
