The Completion of AI: The Dawn of the Humanoid Era
Physical AI, Humanoids, Hardware, Software, and Applications
A humanoid robot marathon held in China has captured global attention. Tiangong Ultra, developed by the Beijing Humanoid Robot Innovation Center, completed a half marathon in 2 hours, 40 minutes, and 42 seconds. Nvidia CEO Jensen Huang has repeatedly stated that the final stage of AI is “physical AI”—in other words, humanoid robots. Nvidia once locked in demand for its GPUs by freely distributing its CUDA software platform, and it has signaled a similar strategy for humanoids. With the launch of its robot training platform COSMOS, Nvidia aims to enable robots to learn and adapt to the physical environment of the real world.
Tesla has set a target of producing 5,000 units of its humanoid robot Optimus this year, with plans to ramp up production tenfold to 50,000 units in 2026. BMW, in partnership with robot developer Figure AI, has also announced plans to supply 100,000 humanoid robots over four years starting in 2025, signaling the dawn of the mass-production era for humanoids.
According to materials released by KB Asset Management citing a Morgan Stanley report, by 2050, up to 75% of jobs in the United States could be replaceable by humanoid robots. The potential market size is estimated at $3 trillion, with the global humanoid robot population projected to reach 63 million units.
Humanoid robots are an orchestration of advanced technologies: sensors that perceive, artificial intelligence that decides, and actuators that execute. Companies supplying hardware, software, and application services all form part of the humanoid value chain. Firms such as Microsoft and Alphabet (Google) provide AI software, Nvidia manufactures GPU semiconductors, and Tesla integrates these components into finished humanoid robots.
Beyond big tech, there are also specialized robotics companies. Intuitive Surgical stands out as a dominant player in medical robotics with its da Vinci surgical system. Universal Robots, acquired by Teradyne, is the global market leader in collaborative robots. On the software side, Aurora, a developer of autonomous driving software, offers sensor fusion and AI-based path-planning technologies that could help humanoid robots navigate both indoor and outdoor environments safely and reliably.
On April 15, three humanoid-themed ETFs were listed simultaneously: KODEX U.S. Humanoid Robots, RISE U.S. Humanoid Robots, and PLUS Global Humanoid Robots Active. KODEX and RISE are passive ETFs focused primarily on U.S. companies, while PLUS is a global active ETF investing across the U.S., Japan, and South Korea.
Although both KODEX and RISE invest in U.S. companies, their portfolio strategies differ significantly. KODEX allocates roughly 57% to big tech names often referred to as the “Magnificent Seven,” including Tesla, Amazon, and Nvidia. RISE places greater weight on robotics specialists such as Intuitive Surgical and Teradyne, structuring its portfolio across hardware, software, and application service providers. PLUS emphasizes geographic diversification, with approximately 19.8% in Korea, 32.5% in Japan, and 40.6% in the United States.
Last December, the listing of the KIWOOM U.S. Quantum Computing ETF, followed by the simultaneous launch of four quantum computing ETFs in March by SOL, RISE, and others, brought renewed attention to the quantum computing theme. In April, ETFs focusing on humanoids and Palantir (PLTR) drew investor interest. From May onward, listings of China tech–related ETFs are also on the horizon.
With tariff disputes triggered by former President Donald Trump and rapidly shifting domestic political dynamics, macroeconomic uncertainty remains high. Rather than reacting to every macro swing, maintaining consistent exposure to promising technologies and industries attracting sustained investment may prove more advantageous for long-term investment performance.