The Robot That Doesnt Work Gets a Policy Win Instead
In late April, Figure AI ran a 72-hour warehouse demo at an undisclosed logistics facility. The company says its humanoid robot, Figure 01, sorted 88,000 packages. The video racked up three million views. Nobody disputes the views. Nobody can point to a paying customer.
That gap — between the demo and the deployment — has not stopped Washington from moving. Robots for America, a new industry coalition launched May 8, brought together 17 companies and immediately began lobbying for federal policy frameworks to support domestic humanoid robotics. The Humanoid ROBOT Act, Senate Bill 3275, is already pending. Introduced by Senator Bill Cassidy of Louisiana in November 2025, it would bar federal agencies and their contractors from acquiring humanoid robots produced or developed by companies based in China, Russia, Iran, or North Korea. The coalition wants more: procurement preferences, R&D funding, and a regulatory on-ramp for commercial deployments.
The timing is not accidental.
"We're at an inflection point," said a spokesperson for Robots for America, who asked not to be named ahead of formal coalition announcements. "Policy has to move at the speed of technology, or we lose the window."
But critics say the window being discussed is not about national security — it is about market positioning. The companies pushing hardest for federal support are the same ones whose robots are not yet operating at commercial scale.
Ayanna Howard, a robotics professor at Georgia Tech and former chief science advisor at NASA, has been direct about the state of the art. "The demos are real," she told me. "The question is whether you can run that demo 24 hours a day, seven days a week, in an environment with dust and variable lighting and a human who is tired and making mistakes next to you. That is a different problem."
Howard's concern is not hypothetical. Warehouse logistics requires near-perfect pick-and-place accuracy at scale. Current systems still struggle with item variability: a crumpled box, a soft package, an object partially obscured. The Figure AI demo sorted 88,000 packages over 72 hours. A large e-commerce fulfillment center moves that many packages in roughly four hours with existing automation and human workers.
The procurement angle
S.3275 does not mention Figure AI, 1X, Apptronik, or Agility by name. It does not need to. Its geographic restrictions effectively create a preferred-vendor list: any humanoid robot with an integrated AI system procured with federal funds must come from a US-domiciled company. That list currently includes every major US humanoid startup and excludes Unitree, Fourier Intelligence, and the other Chinese and Asian competitors building in the same space.
The bill's sponsor, Senator Cassidy, framed it as a national security matter. "Humanoid robots with advanced AI capabilities represent a potential vector for foreign surveillance and sabotage," his office said in a November 2025 press release. "We cannot allow adversary nations to embed their technology in federal facilities."
That framing has buy-in from defense hawks and industrial policy advocates. But the practical effect — funneling federal procurement dollars toward US humanoid companies before they have proven commercial deployments — has drawn skepticism from fiscal conservatives and tech policy analysts alike.
"The technology isn't mature enough to know what we're actually procuring," said one congressional aide who works on technology policy and asked not to be named because they were not authorized to speak publicly. "We're writing a blank check based on demos and investor decks."
The aide added that the bill's definition of "humanoid robot with integrated artificial intelligence" is broad enough to potentially cover existing industrial arms and collaborative robots already deployed in federal facilities, which could trigger procurement renegotiations and compliance reviews that slow adoption rather than accelerate it.
What the industry actually wants
Robots for America's 17 member companies span the humanoid robotics stack: hardware manufacturers, motion planning software providers, sensor companies, and logistics integrators. Several are portfolio companies of venture firms actively investing in the space. The coalition's stated goals include expediting federal procurement pathways, establishing testing and certification standards for humanoid robots in commercial and government settings, and funding a joint research program on human-robot collaboration.
Those are not radical asks. Similar coalitions formed around autonomous vehicles in the mid-2010s and around commercial drones in the early 2020s. But both of those industries eventually had to confront deployment failures, accident investigations, and regulatory pushback — moments when policy momentum collided with operational reality.
Humanoid robotics may be approaching that collision faster than the industry wants to admit. Figure AI has disclosed BMW as a commercial partner, but the partnership is structured as a research and development collaboration at BMW's South Carolina plant — not a volume procurement contract. Agility Robotics has deployed Digit in Amazon pilot programs, but Amazon has not announced any scaling commitment. 1X Technologies, backed by OpenAI, has focused primarily on home-assistance applications rather than warehouse logistics.
None of those companies has published data on mean time between failure, pick accuracy rates at scale, or cost-per-unit economics versus human workers. The demos serve a different purpose: they keep the investment thesis alive while the engineering catches up.
The China angle
The geopolitical framing of S.3275 and the Robots for America coalition has one concrete basis in reality: China has moved aggressively to subsidize its own humanoid robotics industry. The Chinese government has designated humanoid robotics as a strategic priority in its latest five-year plan, and companies like Unitree and Fourier Intelligence have deployed robots in Chinese logistics and manufacturing facilities at a pace that US companies have not matched.
Whether that represents a genuine capability lead or simply a different regulatory environment for rapid deployment is contested. Western robotics researchers who study Chinese systems note that many reported deployments involve highly structured environments with significant human oversight and intervention.
"Structured Chinese factory floor, robot does well. Chaotic American fulfillment center, robot does worse," is how one robotics researcher, who asked not to be named because their institution has grant relationships with US humanoid companies, put it. "That's not a permanent moat. It's a current limitation."
But it is a current limitation that US policymakers appear willing to fund around.
What happens next
S.3275 has one cosponsor as of early May 2026: Senator Chris Coons (D-DE), alongside sponsor Senator Bill Cassidy (R-LA). It awaits committee assignment in the Senate Banking, Housing, and Urban Affairs Committee. The bill's path to a floor vote is unclear, and a Senate aide said it is not yet a priority for leadership.
Robots for America is not waiting for legislative action. The coalition has begun direct engagement with the General Services Administration, which manages federal procurement, and with the Defense Logistics Agency, which handles defense supply chain purchasing. Both agencies have existing frameworks for awarding research contracts and pilot procurement agreements to emerging technology vendors — pathways that do not require new legislation.
If those agencies move first, the policy moat gets built without the debate that a bill on the Senate floor would force.
For the humanoid robotics companies, that is the ideal sequence: get the contracts, build the track record, prove the deployment. Whether the robots actually work at scale is a question that federal procurement dollars, by then, may have already answered.