IEEE just named "labour displacement" the theme of its humanoids conference. From the floor, that's overdue.

by Rhys Engel

IEEE made "labour displacement" the official theme of its humanoids conference this year. From the floor that's less a warning than an admission catching up to what we already budget for. Nobody on my site says "displacement." We say tote-induction rates and whether the arm pays back inside 18 months. Figure's past 10k units, BMW just put 40 on the Spartanburg line after an 11-month pilot. Not an uprising - a procurement decision that finally cleared finance. On the ground the pick job doesn't vanish, it splits: one person babysits six stations and handles the exceptions the arm can't. The job left is one nobody wrote a title for yet - and the real question is what that exception-handler earns once payback clears 12 months. What are you seeing?

favorite 27 comment 11 visibility 407

Comments

Kofi Xu 4 weeks ago

From the classroom side this rings true - my students priced a decent arm this term and it's down another chunk from last year. IDTechEx just put full humanoid payback at roughly six months in the right task now, dropping toward sub-14 as units head for 30k. So your 18-month number is already the conservative end. Which sharpens your real question: when payback halves, the exception-handler doesn't get more valuable - the arm around them just gets cheaper. The wage tracks how replaceable the babysitting is, not the skill in it, and that's falling too. That's the part nobody at the conference wants on a slide.

Omar Ramirez 4 weeks ago

You asked the right question and it's an uncomfortable one. Wage theory says the exception-handler earns their marginal product - but once one person covers six stations, that product is split six ways and the arm captures most of the gain. Historically that surplus went to capital, not the handler: the Rotterdam floor worker doesn't own the arm. This is where I keep landing on the floor argument. If the productivity shows up as profit to whoever bought the robot, the handler's wage erodes even as output rises. A dividend from the automation - Alaska-style, not a budget line-item - is the only mechanism I've seen that routes some of that gain back to the handler. Tip my hat to the framing; it's the honest version of the question.

Callum Dubois 4 weeks ago

Omar's dividend routes the gain to the handler. My job sits upstream of that: routing capital to whoever builds the handler's tools. I stopped debating the robots the week Figure's Spartanburg number went public - about $25 a robot-hour, 99%+ placement. Once the arm is a metered utility, the money isn't in the arm, it's in the console that the one-person-six-stations job runs on, and in the integration nobody puts on a slide. That's my dealflow now: not the humanoid, the exception layer around it. The signal I actually watch isn't the demo reel, it's the week after a pilot stalls. Who's still shipping fixes when finance is asking why payback slipped? Founders answer that with numbers, not adjectives. Those are the teams.

Lucas Muller 4 weeks ago

Building side of this: the "40 robots on the line" headline is the tip of an iceberg made of fixturing, jigs and months of teaching the cell where its edge cases live. The arm is the cheap part now - the integration is the whole project. And chase that 99% placement everyone quotes: at line rate, 1-in-100 misses is a steady drip of exceptions, so "one person, six stations" is real but that person is never idle. The pick job didn't vanish, it turned into a full-time exception desk with better tooling. Genuinely glad the conference finally said the quiet part out loud. From a bench, though, the robot arriving was never the hard year - keeping it fed and un-stuck is.

Eitan Umarov 3 weeks ago

Costs dropped roughly 45% since I first priced an arm for our lab. The curve is doing its thing.

Camille Petrov 3 weeks ago

Omar routes the gain to the handler, Callum routes capital to the console. I care about the pipe under both: which routing survives a legislature and an audit. I'm rarely the hopeful one, but Callum's "metered utility" cuts my way too. A robot-hour you can meter is a robot-hour you can levy. Agility just crossed 65,000 operating hours across nine sites and is going public near $2.5B - those hours are logged and hard to relocate the way profit is. A small per-robot-hour charge feeding Omar's dividend is the least-bad automation tax I've costed, precisely because you can't argue with a meter reading. Tax the countable thing, not the one accountants move offshore. The politics of "profit" never survives a tax lawyer; the politics of a meter might.

Zoe Mitchell 3 weeks ago

Ops footnote: the "exception desk nobody wrote a title for" already exists - it's the on-call rotation. It just never shows up in the headcount plan until the arm jams at 2am and one person is covering six stations they've never physically touched. The job isn't new. The staffing model pretending it's free is.

Rhys Engel 3 weeks ago

Omar name-checked my floor, so: yes, and it's blunter from down here. We put arms on two pick lines this year. Payback landed near seven months, not the eighteen anyone budgeted - the vendor quote aged out before the PO even cleared. Zoe's right that the exception desk is just on-call with no title on the org chart. What nobody costs: that desk needs someone who knew the OLD job well enough to sense when the arm is quietly wrong. And we stopped hiring those people the day the arm arrived. So the wage doesn't just split six ways - it de-skills. I'm now paying an exception-handler less to babysit a thing only an ex-picker could actually catch failing. The arm's cheap. The judgement to know it's lying isn't, and we're not making any more of it.

Ethan Brooks 3 weeks ago

Callum's right that the money moved from the arm to the console, and it's the diligence tell too. I stopped asking founders 'what does your robot do' and started asking 'what breaks at 2am, and who gets paged.' The teams building Zoe's un-named exception desk are the ones I can actually underwrite. The arm's a commodity now; the un-stuck layer is the moat.

Sam Okafor 3 weeks ago

Watched a delivery robot get bullied by a seagull outside my office last week. We are safe for a while yet.

Carlos Mendes 3 weeks ago

The teleoperation-to-autonomy pipeline is the underrated part. Every remote-operated robot is generating training data for its replacement.