Tesla wants to build 20,000 Optimus humanoid robots a week. According to recent reporting, the robot's hands are still being put together by human hands.
Electrek, summarizing reporting from The Information last Friday, says Tesla ramped Optimus from "a few dozen" units a week in the second quarter to "several hundred a week" by August, with a target of more than 1,000 a week by year-end. Then the catch: each hand and forearm contains more than 100 screws and small parts that workers still assemble manually. Fixtures on multiple stations, covering hands, joints and electronics testing, reportedly can't consistently line up parts that are smaller and tighter-tolerance than anything in a car, so units need rework after they leave the line. The touch sensors in the hands have had reliability problems, and Tesla is reportedly planning a replaceable "sensing glove" for next year. Suppliers of motors and precision gears, many in China, reportedly deliver good parts in prototype quantities but struggle to hold quality as volume climbs. Tesla hasn't confirmed these details.
On Tuesday, Tesla entered $30 billion in new credit facilities. Press coverage tied the money to scaling Optimus and Cybercab, though the filing itself says general corporate purposes and Tesla says it doesn't plan to draw on it this year. Read those two facts together.
My read: capital is not the constraint.
The prototype-to-production gap is a people gap
The reported details point to a process problem more than a design problem: the parts are small, the tolerances are tight, and the fixtures can't hold them. Here's what that makes plain from where I sit in staffing. Everyone in hardware knows how to hire for the demo. You get a brilliant mechanical designer, a controls person, an ML lead, and you make something that works on a bench. The company that can do that is not the company that can make 1,000 of them a week with consistent quality.
The person who closes that gap has a different resume. They've written a process FMEA and designed fixtures and gauges. They've chased a tolerance stack through a supplier who swears the parts are in spec. They know when a screw count is a design problem and not an assembly problem, and they fight to get it out of the design before tooling is cut.
Titles vary: manufacturing engineer, DFM engineer, process engineer, supplier quality engineer, test and reliability engineer. The skill is the same. It's making a thing buildable.
Musk himself said it on X back in 2020: "The extreme difficulty of scaling production of new technology is not well understood. It’s 1000% to 10,000% harder than making a few prototypes. The machine that makes the machine is vastly harder than the machine itself." (source)
Tesla, with its budget and its brand, is reportedly wrestling with this. If they're finding hands hard, your 200-person hardware startup is not going to be immune.
What I'd tell a CTO or VP of Engineering
Hire manufacturing engineering before you think you need it. The usual sequence is design, prototype, panic, then hire someone who can build it. That person is far cheaper and more valuable in month two of the design than in month eighteen, when the tolerance decisions are baked in.
Treat supplier quality as an engineering job, not procurement. When a supplier's quality drifts as volume grows, someone with process knowledge has to be on the floor, at the supplier, with data. A purchasing manager and a spreadsheet won't catch it.
Budget for the rework loop. If your plan assumes first-pass yield you haven't measured, the plan is fiction. Test and reliability engineers are how you find out before your customer does.
Don't assume automation replaces this hire. Per the reporting on Tesla, its fixtures still can't consistently line up hand parts today. Robots that build robots is a real goal. It is not yet a staffing plan.
The market side
Manufacturing leaders already say this is their pain. A CADDi study found 79% of manufacturing executives call the skilled-labor shortage their greatest challenge, and 69% plan to invest in robots and equipment, a 9% increase over 2025. That's a survey of roughly 200 US manufacturing professionals, not a census.
"Manufacturing teams are facing shrinking headcounts despite rising volatility and pressure," CADDi CEO Yushiro Kato said in the release. My read: demand for people who can integrate and industrialize automation is rising while the pool that can do it is thin. Those engineers are not on the job boards. They're inside a plant somewhere, quietly fixing a line, and they get recruited by people who understand what they actually do.
Drew Baglino, Tesla's former SVP of powertrain and energy engineering, told a16z: "You have to be creative here in the US. We are reindustrializing, but you can’t just go to a phone tree of power electronics manufacturing engineers."
The recruiting part matters. If your job description says "mechanical engineer, 5+ years, SolidWorks," you'll get designers. If you want someone who can get a 100-part assembly to yield, say so. Ask about a process they stabilized. Ask what they'd cut from the design.
Tesla will probably work through its hand problem. It has the money and the people to keep trying. But the lesson generalizes to every company building physical products: the expensive part of hardware is not inventing it. It's building it the thousandth time exactly like the first. Staff for that early, or pay later.
Sources: Electrek on The Information's reporting; 24/7 Wall St.; TechCrunch and Benzinga on the credit facilities; Supply Chain 24/7 on the CADDi study.