For the first time in decades, having a manufacturing edge is not dependent on an abundant human labor supply like China has lately enjoyed. With AI and automation, America has a once in a generation window of opportunity to rebuild its manufacturing base, even amid labor shortages due to sustained low unemployment.
Enter humanoid robots: intelligent machines that combine AI, automation, advanced sensors and more into a single adaptable workforce. They can breathe new life into our production lines by taking on dark, dirty and dangerous work so that human workers never have to.
Rebuilding a dominant American manufacturing industry with humanoids is possible, but only if federal policy and regulatory bodies create a comprehensive national development and deployment strategy soon.
U.S. robotics policy was built for a world where the hard part was invention. The U.S. presently funds research generously through the National Science Foundation, the Defense Advanced Research Projects Agency and university partnerships, and this has produced genuinely impressive humanoid robots.
But invention isn’t the bottleneck anymore; deployment is America’s new challenge. A mid-sized manufacturer in Ohio or a logistics operator in Georgia can buy a humanoid robot today, but integrating it into an existing production line, retraining staff and proving return on investment is expensive, slow and risky enough that most companies simply don’t try.
The federal government has a long history of using incentives, like the federal research and development tax credit and investment tax credits to push businesses toward adopting new technology. Humanoid robotics deserves the same treatment: a deployment-specific tax incentive that rewards companies for putting humanoids to work on the factory floor, not just for filing patents on new ones. Invention without implementation is a science project; pushing for adoption will boost GDP.
Current examples of humanoid adoption in U.S. manufacturing include installations on BMW, Tesla and Amazon factory floors. But most manufacturers are small and mid-sized businesses that lack in-house robotics engineers to evaluate which humanoids can safely and effectively fit into their operation.
This is a real gap, but the federal government already has a mostly unused solution: The Manufacturing Extension Partnership, a decades-old network of experts in every state that helps smaller manufacturers modernize. The infrastructure already exists to easily expand its mandate to include humanoid deployment experts to help with site assessment, integration planning, and workforce training.
The next bottleneck is no longer safety, its interoperability. We’ve had the answers on how to safely operate humanoids near people for 15 years. At NASA, we created early humanoid safety standards through Robonaut 2, a collaboration with General Motors that deployed the first humanoid in space aboard the International Space Station.
Now, we need to answer an arguably more economically consequential question: Can a humanoid built by one manufacturer work within a factory environment designed around another’s hardware, software and tooling?
Without interoperability standards, every manufacturer that adopts humanoids risks locking itself into a single vendor’s ecosystem, the same dynamic that made early industrial automation expensive and inflexible for decades. The National Institute of Standards and Technology should be tasked not just with developing safety guidelines but devising interoperability standards covering acceptable communication protocols, and approved software interfaces and robot “hand” attachments.
The U.S. may lead in investment in AI for robotics operation but, to truly accelerate domestic manufacturing, we can’t just own the brains; we must own the hands.
Some might argue that the best U.S. strategy is to simply focus on what it already does well: researching and designing the foundational breakthroughs in AI and robotics on which humanoids depend. We could let the messier, more capital- and labor-intensive work of building these machines happen in already established manufacturing ecosystems, just like in semiconductor design versus fabrication.
