Terry Fong spent years running NASA's Intelligent Robotics Group, where systems had to function millions of miles from the nearest repair technician. Now watching the humanoid robot race from outside government, he sees American companies making what he considers a strategic error: building robots that excel in investor demos while Chinese manufacturers design for messy reality. The difference, Fong argues, will determine which nation supplies the factories remaking global manufacturing over the next decade. His warning arrives as companies including Tesla, Figure AI, and Apptronik race to commercialize humanoid platforms, each touting impressive capabilities in controlled environments but potentially vulnerable to the chaos of actual production floors.
The performance-versus-adaptability divide reflects fundamentally different design philosophies. Boston Dynamics' Atlas performs gymnastic routines with precision that captivates audiences, while Tesla's Optimus demonstrates consistent part placement in the company's own facilities. These achievements require sophisticated control systems and enormous engineering effort, optimizing robots for specific tasks in structured settings where lighting, flooring, and component positioning remain constant. Chinese robotics firms, by contrast, are reportedly investing in modular architectures that sacrifice peak performance for broader capability. Their robots may never execute a flawless backflip, but engineers design them to switch between welding, assembly, and material handling with software updates rather than hardware replacement. The distinction resembles the difference between a Formula One race car and a pickup truck—one dominates on a purpose-built track, the other handles whatever roads and cargo appear.
Fong's aerospace background informs his perspective on what manufacturing actually demands. NASA's Mars rovers encounter terrain, lighting, and mechanical challenges that mission planners cannot anticipate years in advance, requiring systems designed for adaptation rather than optimal performance in known conditions. Factories face similar unpredictability despite appearing more controlled than planetary surfaces. A semiconductor shortage forces substitution of components with different dimensions. A stamping press breaks down, requiring manual workarounds until replacement parts arrive. Winter holiday demand needs rapid retooling from summer product lines. Robots optimized for perfect execution when every variable aligns may falter when reality intervenes, while adaptable platforms maintain productivity through disruption. Current American humanoid development, focused on demonstrating capabilities that impress investors and engineers, may be optimizing for the wrong metric. The manufacturing customers these companies target care less about a robot's maximum performance than its ability to keep production running when circumstances deviate from plan.
The strategic implications extend beyond individual product capabilities to industry structure and competitive advantage. If adaptable robots prove more valuable in real manufacturing environments, companies that invested billions optimizing for performance may find their platforms poorly suited to customer needs. Chinese manufacturers would gain deployment advantages even if their robots never match American performance benchmarks, because factories would choose reliable flexibility over impressive but brittle capabilities. This dynamic already played out in other technology sectors—Chinese telecommunications equipment gained global market share not by outperforming Western competitors on specifications but by offering acceptable performance at lower cost with better customization for diverse markets. The robotics industry could follow a similar trajectory, with adaptability-focused designs winning manufacturing deployments while performance-optimized humanoids remain confined to demonstration environments and specialized applications. Fong's warning essentially argues that American robotics companies are fighting the last war, competing on metrics that mattered when robots performed narrow tasks in automotive plants, while the future demands platforms that handle the full complexity of modern manufacturing.
What to Watch: Monitor deployment announcements from Chinese humanoid manufacturers including Unitree, UBTECH, and Fourier Intelligence through Q2 2025 for evidence of multi-task installations versus single-purpose applications. Track whether American firms including Figure AI and Apptronik adjust product roadmaps toward modular capabilities rather than performance optimization. Watch for major manufacturers like Foxconn and BYD to specify adaptability requirements in robotics RFPs, signaling which design philosophy the market values.
