AI is leaving the data centers and coming into the real world. Robots learn, see and act – but the winners on the stock market may be somewhere else entirely.
Dusseldorf. Enthusiasm for artificial intelligence (AI) is slowly returning to the stock market. Last Tuesday, the Nasdaq 100 technology index in the USA reached its highest level in its history.
The operators of large data centers, so-called hyperscalers, and associated equipment and suppliers, which include almost the entire chip industry, have particularly benefited from the AI rally so far. But the next AI winners could come from other industries.
Ulrike Hoffmann-Burchardi, investment strategist at asset management at major bank UBS, says: “We have officially entered the age of physical AI.” Handelsblatt explains what this means, which regions are leading and how investors can invest.
What is physical AI?
Physical AI refers to artificial intelligence that is integrated into machines and perceives, decides and acts in the real world. For example, a robotic system can grab, sort or move objects.
A central technological component is so-called in-context learning. Traditionally, robots had to be extensively programmed for new tasks or adapted through long training and simulation processes. “You had to carry out millions of digital simulations, train a model for weeks and then program the exact parameters into the machine,” explains Hoffmann-Burchardi.
In-context learning now turns this approach on its head. Modern robots can observe how a human performs a task and then immediately adapt their behavior. “You grasp the task immediately and solve it precisely in this context,” says the investment strategist.
The Feri Cognitive Finance Institute, the strategic research center of the investment house Feri, has come to the conclusion in a joint study with the Fraunhofer Institute that physical AI can fundamentally revolutionize automation.
“We are seeing the beginning of a rapid transformation in robotics that will have enormous consequences worldwide,” says Heinz-Werner Rapp, founder and head of the Bad Homburg think tank.
What are the benefits?
The most important advantage of physical AI is its flexibility and ability to learn. “In-context learning has given machines the agility they need to learn directly on the job,” says Hoffmann-Burchardi. Physical AI can learn new tasks through observation, adapt more quickly to changing environments, respond immediately to real-world situations, perform different activities without complete reprogramming, and interact precisely with its environment.
The technology is intended to create economic added value, particularly in areas such as logistics, manufacturing and healthcare.
Rapp also points out the importance of physical AI in light of demographic change. The supply of skilled workers and workers is likely to decline as the population ages. Rising wages and ancillary wage costs also make the use of the new technology interesting, as does increasing international competitive pressure.
"These forces are forcing massive investments in automation worldwide. Robotics will ultimately become an indispensable pillar of social sovereignty and livelihood security," says Rapp.
What sub-areas are there in the robotics market?
To date, industrial robotics is the largest and most economically important submarket. However, due to advances in AI, the global robotics market is in a phase of structural change.
The global market for service robots is growing dynamically. You operate in environments that change frequently, such as in transport and logistics. However, humanoid robots trigger the greatest growth fantasies among investors. Last year, venture capitalists invested around $6.1 billion in this area, according to financial data provider Pitchbook.
In contrast to industrial and service robotics, humanoid robots are still in an early phase of commercialization, the experts from the Feri Institute write in their study: “In addition to initial pilot projects and industrial field tests, research and development activities as well as investments in future production capacities dominate.”
Several investment houses see significant growth potential in the coming decades. The investment bank Goldman Sachs Research predicts a global market volume of around $38 billion by 2035, while the major bank Morgan Stanley estimates the global addressable market to be around five trillion dollars by 2050.
However, Morgan Stanley expressly points out that there are significant uncertainties. Feri also warns against excessive expectations. From their perspective, it is currently unlikely that humanoid robot systems will replace existing robotics solutions in the short term: “Stationary industrial robots, collaborative systems and autonomous mobile robots remain superior in terms of productivity, reliability and cost structure in many standardized applications.”
How can investors invest?
If you are an investor who wants to invest in physical AI, Hoffmann-Burchardi recommends taking the advice of André Kostolany to heart. The legendary investor, who died in 1999, once said: “In a gold rush, don’t invest in the gold diggers, invest in shovels!”
Applied to the current situation, this means that investors should focus more on suppliers and infrastructure companies. Nico Baum, head of AI at Berenberg Bank until August, warned investors in May not to narrow their focus on the robotics stars. “The first winners of the physical AI wave are less the robots themselves than the infrastructure behind them,” he told Handelsblatt.
Hoffmann-Burchardi explains: "Physical AI requires a new hardware supply chain. These advanced robots require highly specialized physical bodies in order to be able to carry out what they learn." The UBS expert names three crucial levels of the supply chain: actuators and precision gears, sensors and perception as well as edge computing and energy infrastructure.
Rapp also believes that investing in upstream value creation stages is a better approach compared to individual robotics providers. He emphasizes: “No matter which specific robot system prevails – suppliers of sensor systems, simulation software and AI infrastructure will benefit in any case.”
Which stocks could benefit from physical AI?
Following the advice of the experts is not that easy. In order to invest in suitable stocks, investors have to identify companies where physical AI either already makes up a large part of the business, is an important sub-business or is a fundamental growth market. Investors will then find what they are looking for primarily in China, South Korea, Japan and the USA.
Cognex from the USA is a provider of image processing solutions for manufacturing and logistics. The Chinese companies Hesai and Robosense as well as Ouster from the USA each develop and produce lidar sensors for three-dimensional recording of the environment. Harmonic Drive Systems from Japan, in turn, produces high-precision drive technology that is important in automation.
It should be noted, however, that some of the companies are still relatively small and are difficult for German investors to trade in this country. This can lead to price increases when buying and discounts when selling. Alternative trading via foreign exchanges may involve higher costs.
In addition, it is currently uncertain which companies will prevail in the respective field in the long term. So stocks are a bet on the future. Baum therefore urges caution: “Not every robotics fantasy ends up being a profitable business model.”
