“Science and Technology Innovation Board Daily”, December 30 — Competition for AI chips is becoming increasingly intense. According to media reports, Nvidia, which dominates computing power, is betting on robotics, hoping it will become the company's next major growth driver.

“The ChatGPT moment for physical artificial intelligence and robotics is coming.” Deepu Talla, vice president of Nvidia's robotics business, said.

Currently, robots, especially humanoid robots, are still in the early stages of scale, and the industry is still working to reduce costs, improve product accuracy, and expand scale.

Tella pointed out that the robot market's transformation is mainly driven by two technological breakthroughs: the explosive growth of AI models and the ability to train robots on these models through simulated environments. The latter is an extremely important development that helps bridge the “simulation-to-reality gap” and ensures that robots trained in virtual environments operate effectively in the real world.

“Over the past 12 months, this gap has matured enough (gradually bridged). Now we can combine generative AI to perform simulation experiments, something we couldn't do two years ago.”

In the field of robotics, Nvidia positions itself as a supplier of robotics technology platforms, supplying full-stack solutions covering all layers of software and robot chips for training robots. The company plans to launch the latest generation Jetson Thor, a small computer for humanoid robots, in the first half of 2025.

Meanwhile, Goldman Sachs has drastically raised its humanoid robot market size forecast in its latest research report. By 2035, the humanoid robot industry market will reach 38 billion US dollars, which is more than 6 times the previous forecast value (6 billion US dollars).

When explaining the reason for the increase in expectations, like Nvidia's Deepu Talla, Goldman Sachs also indicated the important role of AI models in robot training. “The accelerated evolution of AI, technological breakthroughs, and increased capital expenditure investments are all core drivers for us to adjust our forecasts. End-to-end AI has now made significant progress. Models can train themselves through this AI, and no longer require human engineers to write all the code by hand. This has accelerated the development of robots, enabling these devices to complete more tasks faster and adapt to new situations (such as working outside the factory).”

Goldman Sachs predicts that humanoid robots are expected to be first used in factories between 2024 and 2027, while the time to enter the consumer market is expected to be between 2028 and 2031.

Overall, many companies at home and abroad have recently accelerated the pace of robot deployment.

After OpenAI has already invested in Figure and 1X, the company has restarted the robot team; GAC Group released its intelligent humanoid robot GoMate; the Leju humanoid robot production line was officially launched, which is expected to produce 200 humanoid robots per year; Shanghai Matrix Superintelligent Integrated System Co., Ltd. released a self-developed full-size general-purpose humanoid robot prototype MATRIX-1; Eurich Technology's new general-purpose humanoid robot Wanda 2.0 was released...

Great Wall Securities pointed out that the humanoid robot industry chain is currently at an accelerated stage from “0-1” to commercialization and implementation of “1.” Since this year, it can be clearly seen that the number of entrants has increased dramatically, and it has become a clear trend for large companies to enter the market. Analysts are optimistic about humanoid robots as one of the important directions for the implementation of general artificial intelligence, a cutting-edge direction of new productivity, and an important branch of the startup economy. Driven by policies, the commercialization process continues to accelerate, and it is expected that the first year of mass production will arrive in 2025.