What We Need Isn't Robots, but the Labor They Provide
In 2026, a tempting belief circulates in the tech world: robotaxis will soon replace private cars, and humanoid robots will become the next household necessity. These two narratives are intertwined, pointing to a common conclusion—for most families, cars are no longer essential, while robots will become new "family members." However, this logic chain is facing three tests: cars are not disappearing, but transforming from private assets into mobility services; autonomous driving and robots share a technical foundation, yet the physical world complexity they face is not on the same scale; and robots themselves are more likely to follow an evolution path similar to cars—from private ownership to servitization.
The first challenge lies in the divergence despite shared foundations. Car companies are transferring their intelligent driving capabilities into the robotics domain. Tesla's Optimus shares the end-to-end architecture, AI5 chip, and factory data with FSD, but as of the second quarter of 2026, it remains in small-batch trial production, with mass production postponed to summer 2027. BYD's first humanoid robot "Xiaodi" has entered experience centers, taking on reception and explanation duties. Car manufacturers possess ready-made scenarios, supply chains, and mass production capabilities, but the technical difficulties for cars and robots are vastly different: the former mainly solves mobility in structured road environments, with traffic rules as constraints; the latter must simultaneously handle mobility, manipulation, safety, and economy, making it far more complex than road driving.
The transition from industrial to domestic settings is even more arduous. In 2025, global shipments of humanoid robots were about 18,000 units, and data from the Chinese Electronics Society shows shipments exceeded 40,000 units in the first half of 2026, but household penetration is extremely low, with primary applications still in entertainment, commercial performances, education, research, and data collection. Stanford HAI's "AI Index 2026" report indicates that task completion rates in simulated environments reach 89.4%, but drop sharply to 12% in real homes. Unitree's prospectus confirms this divergence: in 2025, research and education accounted for 73.6% of humanoid robot revenue, while industrial applications were only 9.01%.
The second point is a more likely evolution path from ownership to service. The robotics industry is repeating the route already validated by the automotive industry—shifting from "hardware products" to "service entry points." Agility's Digit adopts a RaaS model in Amazon warehouses, priced at approximately $25 per hour; 1X NEO offers a $499/month home subscription; Tau Robotics provides remote-controlled cleaning services. The domestic robot rental market is also growing rapidly, with platforms like Qingtian Zhu leasing promoting RaaS models; daily rental rates have fallen from a peak of RMB 15,000-20,000 to RMB 3,000-5,000, and applications have expanded from stage performances to industrial trials and commercial marketing. The rental market size exceeded RMB 1 billion in 2025 and is expected to reach RMB 10 billion in 2026.
The economic calculus of robotaxis replacing private cars is not yet settled, and humanoid robots entering homes also face a time window mismatch. Consumers do not need an expensive piece of hardware but the labor service it provides. As Chery Chairman Yin Tongyue predicted, future cars will be replaced by "wheeled robots," and mobility will shift from owning assets to purchasing services. When both cars and robots are no longer personally owned, what truly holds value is not the hardware itself but the underlying embodied AI foundation—models, data, supply chains, and service entry points—which will be the core of the next round of competition.