NOBODY IN THAT CROWD ASKED WHAT THE ROBOT COST
They asked if it could shoot a jumper. They asked what it was wearing. Somebody asked it to throw a peace sign and it did, and the whole group laughed like a person had done it.
Not one of them asked what it does for a living.
That's the part worth noticing. Every robot company on earth is optimising hands and gait and payload, and the thing that actually moved that crowd was two seconds of small talk. Nobody was checking whether it could lift a crate.
Then at 0:50 it's behind a bar, pouring a drink.
That's the business. Not the factory version everyone argues about, the one where a machine stands in a room full of people and makes them want to stay near it. Bars, launches, storefronts, brand events. Footfall is the product and the robot is the reason they walk over.
A humanoid that lifts things competes with a forklift on price. A humanoid people want to talk to competes with a marketing budget, and those are very different numbers.
I wrote the whole breakdown in the article below, what one of these actually costs today and where the money in this is really moving.
THIS ROBOT COSTS A WAREHOUSE $4 AN HOUR AND NEVER GOES HOME
Parcel in, scan, orient, place, next one. It doesn't speed up when a supervisor walks past and it doesn't slow down at hour eleven, because there is no hour eleven for it. It just keeps going at the same rate it started.
That's the entire pitch. Not intelligence, consistency.
A sorter on that line moves fewer parcels than the machine on a good hour, and far fewer on a bad one, and works eight hours out of twenty four. Breaks, shift handovers, sick days, the week before a holiday when half the floor goes home. To cover one machine running around the clock you need three people, and they still fall behind in December.
Wages, benefits, turnover, retraining, supervision. Three sorters is real money every year, forever. Hardware in this class is one payment, some maintenance, and power. Payback lands inside two years and everything after that is margin that never appears on a payroll line again.
At 0:28 the camera drifts and you can see the back of the floor. More of them lined up, not switched on yet.
One robot is a test. A row of them means somebody already ran the numbers and liked the answer.
I wrote the whole breakdown in the article below, what these actually cost today and where the money in this is really moving.
ONE ROBOT IN GUANGZHOU DOES THE WORK OF SIX PEOPLE AND NEVER TAKES A SHIFT OFF
1,200 parcels an hour. Scan, orient, place, repeat. No slowdown at hour six, no slowdown at hour twelve, because there is no hour twelve for it, it just keeps going.
Put a person next to that. A good sorter moves somewhere around 600 to 700 an hour when they're fresh, less by the end, and they work eight of the twenty four. Breaks, shift changes, sick days, the week before Chinese New Year when half the floor goes home.
Run it out over a day. The machine clears close to 29,000 parcels. It takes about six people working three shifts to match that, and they still don't match it in December.
Sorting labor in Guangzhou runs roughly $9,000 to $10,000 a head once you count everything. So the robot is standing in for somewhere near $55,000 a year in wages, and hardware in that class lands in the $60,000 to $80,000 range.
Payback inside two years. After that it's just margin.
Now multiply. A center running two hundred of these isn't saving a salary, it's taking eleven million a year off the payroll line, permanently, and the parcels move faster than they ever did with people.
That's not a demo, and it's not coming. It's running tonight while you read this.
I wrote the whole breakdown in the article below, what these actually cost today and where the money in this is really moving.
HE BOUGHT A ROBOT FOR $13,500 AND RENTS IT TO A CLEANING COMPANY FOR $2,900 A MONTH
The video is the robot working. Standing upright with a mop, working the bathroom floor, going around the base of the toilet, reaching the corner behind the door, wringing out in the bucket and going back over the spot it missed.
He says he never planned to keep it at home. He bought it, spent a week teaching it three rooms, then walked into the office of a local cleaning contractor and offered it as a night shift nobody has to staff.
They took it. It does the floors in an office block after everyone leaves, he drives it over on Fridays and picks it up Monday, and the only thing he does in between is charge it and answer his phone if something goes wrong.
$2,900 a month. He says it paid for itself before the first frost.
Here's the thing.
This is going around as a real setup and half the comments are asking which model he used. Watch 0:05, the wrist doesn't rotate the way a real one does. What do you think, real or generated?
I think the video is generated. But the business in it isn't. Renting these out instead of owning them is already how most of this market actually works, and the daily rates are public.
I wrote the whole breakdown in the article below, what one really costs today and where the money in this is actually moving.
A FARM ROBOT PAYS FOR ITSELF IN TWO SEASONS
The clip is a humanoid cutting hay in a field. Clean, steady, nobody supervising it, working the row like it's done a hundred of them.
Field work is the last thing to automate. Uneven ground, weather, plants that never grow where the plan said. Every fixed machine that tried needed the field rebuilt around it first.
So run the numbers people in this space actually use. A US farm hand costs a farm somewhere near $50,000 a year and works maybe eight months, minus rain, minus daylight, minus fatigue. Cover a full season properly and you're hiring three of them. A humanoid at the price these are expected to land at works longer days, all year, no housing, no seasonal hiring, no visas.
Two to four years and it's cheaper than everyone it replaced. That's the whole argument, and it's why money is moving into this before anyone has shipped a working one.
Here's the thing.
Watch 0:08, the grass doesn't move right where the blade passes. What do you think, real or generated?
I think it's generated. Nothing walks a field like that yet.
But the ones that do exist are already working, just somewhere boring where nobody films them. I wrote the whole breakdown in the article below, what one actually costs today and where the money in this is really moving.
A FACTORY FLOOR OF ROBOTS SEWING, AND NOT ONE OF THEM IS AI SLOP
Hundreds of them, rows deep, all doing the same thing. Needle down, feed the fabric, hold the tension, next piece.
Sewing looks easy until the material moves. Fabric stretches, slips, bunches, and the seam goes wrong a few millimeters at a time. That's why garment work stayed human while everything around it got automated. Machines could stitch. They couldn't react.
Watch 0:11, the piece drifts and the arm corrects mid seam. That's the whole thing. Not the stitch, the correction.
Here's the thing.
Every week I post a clip like this and half the replies say generated. This one isn't. It's a real floor, in a real factory, running right now.
So the question isn't whether it's fake anymore. It's why everyone's still watching Optimus demos when the machines that already took a job have been quietly working for months and nobody films them.
That's the actual gap, and it's where the money is. I wrote the whole breakdown in the article below, what one of these costs today and who's already getting paid.