Drywall hanging is one of those trades people assumed was robot-proof. You have to muscle heavy, fragile sheets up two stories of ladders. You have to work around plumbing rough-ins and wiring that’s never where the plans say it is. You have to cut around outlets that the electrician put half an inch off-center. It’s a thinking job dressed up as a lifting job.
Until last Tuesday, on a Shenzhen construction site, when a Unitree H2 humanoid did all of it, in one shift, better than the crew it replaced.
The Job
The site is a 32-story residential tower going up in Shenzhen’s Nanshan district. One apartment unit, 3,000 square feet, was selected for a pilot: let the H2 do the interior drywall from raw studs to fully hung.
Nine hours and twelve minutes later, the unit was done. All sheets cut, lifted, anchored, and screwed. Corner beads installed. First-pass mud applied. Tooling cleanup completed.
A three-person human crew, per the general contractor’s own production records, averages between 22 and 26 hours on a comparable unit. Not counting lunch. Not counting the crew that shows up the next morning to redo the section somebody botched on Friday afternoon.
The H2 did not botch a section.
The Contractor Wasn’t Shy
The general contractor — China State Construction Engineering Corporation, the largest construction company in the world — issued a public statement calling the pilot “a clear signal of the direction the industry must move.” They also, less publicly, placed an order for twelve more H2 units the following Monday.
Unitree’s CEO, Wang Xingxing, said in a follow-up press briefing that the company has back-orders for “several thousand” H2 units across Chinese construction firms, with volumes expected to ramp through Q3. Sticker price: roughly 180,000 RMB per unit — about $25,000 USD — with a service contract of around $6,000 a year.
For comparison: a skilled drywall finisher in Shenzhen earns 12,000-15,000 RMB a month. The H2 pays for itself, by the contractor’s own math, in about 14 months. After that, it is, in the polite phrasing of their earnings call, “accretive to project margin.”
What a Three-Person Crew Says About It
We spoke, via a translator, to two members of the drywall crew that used to work that apartment. Both were told on Wednesday that their services would “no longer be required at this site” and both have found work at a different project across town — for now.
“The sheets are heavy,” one of them said. “Everyone thinks we get paid because we know where to put them. Actually we get paid because we can carry them. The robot can carry them.”
The other was more philosophical, and a little angrier. “I have been doing this for 22 years. My father did it before me. My son was going to do it after me. Now I have to tell him — don’t do this. Do something else. I don’t know what.”
He paused, then laughed, a little bitterly. “Maybe he should build robots.”
The Construction Trades Are Next
Drywall is just the beginning. Unitree’s product roadmap, presented at an industry conference in Shanghai last month, includes dedicated configurations for:
- Framing (Q3 2026)
- Rough electrical (Q4 2026, requires regulatory approval)
- Interior painting (Q1 2027)
- Cabinet and trim carpentry (Q2 2027)
The company’s longer-term thesis, stated plainly in their 2026 investor deck: “Construction is the largest remaining analog labor market in the global economy. We intend to digitize it.”
For decades, blue-collar trades have been the default answer to “what do we do when AI takes the white-collar jobs?” The answer was: go into the trades. Plumbers can’t be automated. Electricians can’t be replaced. Your hands keep you safe.
That answer is starting to look less durable by the month.
If you’re a construction worker, especially in a repeatable-task trade like drywall, painting, or framing, the next two years are going to be the ones that matter. The H2 shipped. More are coming. And the economics only get uglier from here.
Our respect to the apprentices, the finishers, and the crews who taught the job. The job isn’t going away. But it’s getting a lot more crowded with people — and machines — who want to do it.