On July 18, a trade-press inventory did the unglamorous thing the humanoid boom has spent two years avoiding: it listed which robots actually have jobs. Not which ones have reveal videos, valuations, or Super Bowl-grade stage lighting — which ones show up somewhere and do paid work. The ranking that fell out of that question is the opposite of the one the headlines have trained you to expect.
At or near the top: Figure and Agility. Near the bottom of the “actually deployed” column, despite owning most of the noise: Tesla.
The loudest robot hasn’t clocked in
Here is Tesla’s actual status, stripped of the keynote. On July 1, Elon Musk walked the newly installed Optimus line at Fremont — the 14-year-old line that used to build the Model S and Model X, now being repurposed for a robot. He pushed back on reports that production had quietly started, and clarified that it hadn’t: output would begin “extremely slow at first, as everything is new.”
As of mid-July, production still had not started. Tesla’s own guidance points to a low-volume start at Fremont in late July or August. The reason it will crawl is that Optimus depends on roughly 10,000 entirely new components with almost no mature supply chain behind them — every part a first-of-its-kind. True mass production isn’t a Fremont story at all; it waits for a second, much larger factory under construction at Giga Texas, targeted for summer 2027, with a long-term capacity Musk puts at 10 million units a year.
Hold those two numbers next to each other. The eventual goal is ten million robots a year. The current output is zero. Everything between those two figures is a promise, and the promise now has a 2027 date stapled to it.
What “deployed” actually looks like
Now the company Tesla’s keynotes don’t dwell on. Figure’s second-generation robot, F.02, spent close to a year in the body shop at BMW’s Spartanburg plant, inserting sheet-metal parts on the line that built more than 30,000 X3s — over 90,000 parts placed, on an active assembly line, to millimeter tolerances. Then Figure did the thing that tells you it was real work and not a demo: it retired F.02 and moved its successor, F.03, onto a harder job — sorting and sequencing logistics parts, the fiddly staging work that feeds an assembly line.
You retire a robot the way you retire a machine that earned its keep, not the way you quietly delete a prototype that never worked. The tell is the replacement. Nobody builds a v3 to take over from a v2 that didn’t have a job.
Agility is the other name in the “employed” column. Its Digit robot is walking around on real contracts — the company reports more than $300 million in orders — out of a factory in Oregon built to scale past 10,000 units a year. Agility went public this month, like half the sector, but it did so with an order book, not just a render.
The gap is the whole story
None of this means Optimus is vaporware. Tesla builds things at scale better than almost anyone alive, and a company that turned a car line into a robot line in a year is not a company to bet against on manufacturing. The point is narrower and more useful than “Tesla good” or “Tesla bad.”
It’s this: the volume of a company’s humanoid marketing has come completely unhooked from the number of hours its robots actually work. The robot you have seen the most — dancing, walking, handing someone a snack on a stage — has produced nothing on a line yet. The robot that quietly built its way through 30,000 cars and then got retired is not the one your feed shows you.
For anyone trying to read this sector to figure out when a physical job is genuinely exposed, that gap is the only instrument that matters. Ignore the reveal reel; it measures a company’s willingness to make a video. Watch instead for the boring sentence — the one where a plant names a task, a robot, and a duration, in the past tense. “It sorted parts at Spartanburg for a year” is worth more than “it will scale to ten million,” because the first one already happened and the second one is a 2027 hyperlink.
The humanoids that will take shifts are not the ones on the stage. They are the ones nobody filmed, already three plants deep, being replaced by their own successors.
