Ex-NASA Robotics Chief: US Humanoid Robots Built for Show
Key insights
- Stanford research showed humanoid robots with 90% simulation success rates achieved only 12% completion on real household tasks.
- Ambrose calls for federal interoperability mandates and deployment-focused tax incentives to close the US-China robotics standards gap.
- Chinese firms Unitree, AGIBOT, and Fourier Intelligence are actively scaling factory deployments while US competitors remain pre-commercial.
Why this matters
Industrial robotics standards set now will determine supplier ecosystems, software stacks, and interoperability protocols for the next decade of manufacturing automation. If China's vendors lock in those standards through early-deployment volume, US companies like Figure AI and Boston Dynamics face the same market-access dynamics that Western EV makers now encounter with China's charging infrastructure dominance. The federal policy levers Ambrose identifies, including interoperability mandates and deployment tax incentives, are currently absent from US robotics strategy, and the deployment gap is widening in real time.
Summary
Dr. Robert Ambrose, former head of NASA's Software, Robotics and Simulation Division, is calling out US humanoid robot developers for chasing demo spectacle over deployable systems. A Stanford study he cites found robots hitting 90% simulation success rates completing only 12% of actual household tasks.
The fix Ambrose proposes is federal: specialized tax incentives for robotics deployment and interoperability mandates that force vendors to build compatible systems. Without those, he argues China will lock in global manufacturing standards before US players reach commercial viability.
Essentially: (Figure AI, Hyundai-Atlas) are scaling deployments against (Unitree, AGIBOT, Fourier Intelligence) while the adaptability gap remains unsolved.
- Stanford testing found a 78-point drop from simulation to real-world household task completion rates.
- Ambrose calls for federal interoperability mandates and deployment-focused tax incentives, neither currently part of US robotics policy.
- China's vendors are already in active factory deployments as US competitors remain pre-commercial.
The standards race for humanoid robotics gets won in warehouses and factories, not on demo stages.
Potential risks and opportunities
Risks
- Figure AI and Hyundai-Atlas could face investor pressure in 2026-2027 if factory deployment data surfaces performance gaps matching the Stanford numbers, triggering valuation corrections
- Without interoperability mandates, US robotics vendors risk fragmenting into incompatible proprietary systems, slowing enterprise adoption and giving integrated Chinese platforms a procurement advantage in global manufacturing
- If Unitree and AGIBOT reach industrial deployment scale in 2026, they may set the component and API standards that US manufacturers will have to conform to for global market access
Opportunities
- Simulation-to-real-world transfer specialists including Nvidia Isaac Lab gain strategic leverage as the performance gap Ambrose describes becomes a known enterprise procurement criterion
- Policy consultancies and standards bodies including IEEE and NIST have an opening to draft robotics interoperability frameworks before federal mandates force a rushed, industry-hostile process
- US logistics and contract manufacturers considering domestic robotics deployment, including Amazon Robotics and GXO Logistics, could use the policy vacuum to advocate for procurement standards that structurally favor adaptable systems over demo-optimized ones
What we don't know yet
- Whether the Stanford study's 12% real-world completion figure has been replicated by other institutions or reflects task-set selection bias specific to household environments
- Which specific federal agencies or legislative vehicles Ambrose envisions for enacting interoperability mandates and robotics-specific tax incentives
- Whether current Figure AI and Hyundai-Atlas factory pilot data supports or contradicts the simulation-to-real-world performance gap Ambrose describes
Originally reported by fortune.com
Read the original article →Original headline: Former NASA Robotics Chief: America Is Building the Wrong Kind of Humanoid Robots — Performance Spectacle Over Adaptability Is Ceding Industrial Standards to China