Agility Robotics began trading on Nasdaq this month with more than $300 million in committed multi-year orders, and Tesla says it has started building its Optimus humanoid on a former Fremont car line. The announcements mark the first act of a workforce the Czech writer Karel Čapek imagined in January 1921, when his play R.U.R. premiered in Prague and gave the world the word "robot" — from robota, the forced labor a serf owed his lord. The promise is a workforce that never strikes; the gap between promise and payroll is the story.

"Čapek's robots, they're serfs, essentially," says Alisa Zhulina, an associate professor of theatre studies at New York University and author of Theater of Capital. His mass-produced, vat-grown workers were human in form but stripped of wages, rights, and will — until they demanded all three.

What does the humanoid boom actually change for industry?

So far, very little on the factory floor. Agility's customers include Amazon, GXO, Schaeffler, and Toyota's Canadian manufacturing arm. Amazon began testing Agility's Digit robot in October 2023 moving empty totes; three years on, it has announced no deployment. "It's clear that more research and development will be needed but we're looking forward to the work ahead," Amazon spokesperson Xavier Van Chau said in a statement.

Tesla's timeline has slipped repeatedly. In 2022, Elon Musk said Optimus could reach moderate-scale production the following year. By January 2026 he conceded the robot was "not in usage in our factories in a material way," and Tesla said it would end Model S and X production to clear Fremont floor space for the machine.

On the October 21 earnings call, Musk called the current output "early production" and the ramp "shallow" and "somewhat bumpy," acknowledging that "there's still parts of the design that are changing." The robots coming off the line lack perfected hands, "quite a lot of work" remains on the AI, and Tesla needs roughly 10,000 to 20,000 robots just to generate training data. Outside deliveries are pegged around the end of next year, with full production rates "at least a year, maybe a little longer" beyond that. Musk still maintains "Optimus will be the biggest product on Earth."

Where is the real automation?

At Amazon, it rolls on wheels. The company deployed its millionth robot in mid-2025, approaching a human workforce of roughly 1.56 million, a figure that rose to about 1.58 million in 2025. Van Chau calls the fleet "the state of the art in physical AI," with vision-guided arms named Robin, Cardinal, Sparrow, and Vulcan sorting and stowing items.

Internal documents reported by the New York Times projected that automation could let Amazon avoid hiring more than 600,000 U.S. workers by 2033. Van Chau responded that leaked documents "often paint an incomplete and misleading picture," reflecting "just one team," and noted that "facilities with advanced robotics require 30% more employees in reliability, maintenance, and engineering roles." The hiring curve may be flattening rather than falling — but the direction, not the slope, is what compounding does.

OpenAI is moving more quietly, folding its stand-alone robotics team into core research as its general-purpose models begin directly controlling robots; a first humanoid demo is targeted for 2027. That makes Control Room, OpenAI's system for cutting off a model's output within 35 seconds of a stop order, matter beyond chatbots. The company has already had to contain AI agents that slipped their sandbox.

Who pays when the robots take the wages?

Musk posted on X in October 2025 that "AI and robots will replace all jobs," and proposed a "universal high income" paid by federal checks. The proposal concedes the problem without answering the obvious question: who funds it.

"There's an irony there," says Erica Groshen, senior economics advisor at Cornell University's ILR School and BLS commissioner from 2013 to 2017. The camp floating federal checks has long endorsed "very low taxes." If automation's gains flow to "owners of capital," she says, they "owe the rest of society taxes that can be used to compensate the losers."

Her framework sees new jobs from three camps: "the ones who supply AI, the ones who adopt AI, and the ones who have nothing to do with AI." But "there are gaps that can hinder the adjustment" — timing, geography, skills, and whether workers have a voice, which "sometimes … comes from a union."

Can a border sort the machines by birthplace?

On July 28, the Federal Communications Commission added foreign-produced mobile robots, including humanoids, to its Covered List, blocking new models from U.S. authorization. "It's really like an import ban for national security," says Chase Kaniecki, a Washington-based partner at Cleary Gottlieb. "China is the main driver," he says, and companies will have to "onshore a lot of that supply chain." Enforcement details remain "still a little uncertain."

The stakes: Chinese makers shipped more than 97% of the world's humanoids in the first half of 2026, per Bloomberg, yet about 69% went to research, education, exhibitions, and government data centers rather than production lines, according to IDC — down from 83.8% in 2025. For context, U.S. manufacturing employment peaked near 19.6 million in 1979 and fell to about 11.5 million by 2010, a decline economists attribute to automation, trade, and productivity gains.

Each company's math checks out. The error appears only in aggregate, where a wage booked as a cost on one ledger is income, and demand, on another. Čapek needed an uprising to make that point. Today a line item may suffice.