Humanoid robots will become the next explosive track: predicted to surge 70 times in the next 10 years

Kuai Technology News on July 6, Goldman Sachs released the latest industry report, making a new prediction on humanoid robots and AI underlying technology. It is believed that the humanoid robot industry will usher in rapid growth in the next ten years, with global shipments soaring 70 times compared to the current scale. The organization classifies robots, autonomous driving, and industrial intelligent equipment as physical AI. Compared with large models that simply generate text and pictures, the research and development threshold for this type of physical equipment is higher. Machines not only have to understand pictures and languages, but also deal with real-life physical conditions such as gravity, friction, temperature, and movement safety. Humanoid robots are the hottest track among them. The data calculation is very intuitive. In 2025, global humanoid robot shipments will only be 20,000 units, and will increase to 1.4 million units by 2035, a 70-fold increase in scale in 10 years. The core reason for the explosion in demand is the continued global labor shortage. Looking only at the U.S. manufacturing industry, there are more than one million material handling job vacancies, and repetitive assembly line work is very suitable for robot replacement. However, the industry will not be fully popularized immediately. Large-scale commercial implementation will have to wait until 2027 to 2029. It is still in the technical verification stage. The report also proposes that world models will become the second growth engine for AI computing power demand. Traditional large models only deal with text images, while world models can simulate complex causal relationships such as physical rules, supply chain changes, and policy impacts. It is divided into two major application directions. The physical category supports robots, autonomous driving, and industrial simulation; the social category is used for corporate strategy deduction, investment calculation, and policy simulation. The world model will not replace the existing large language model, but will add new computing power consumption. Simulation requires continuous and large amounts of calculations. If the technology is implemented faster than expected, the current market investment estimates for servers and power supplies will be low. Overall, AI development is shifting from online text to the physical world, and humanoid robots are the core carrier of this route. The labor gap brings long-term rigid demand, and the superimposed world model technology iteration will jointly support the industry's large-scale growth in the next 10 years. The upstream and downstream hardware and computing power industry chains will simultaneously usher in growth opportunities.