Tesla's Robotaxi Never Sleeps Now. Musk Wants 20,000 Robots Learning Jobs Too
Tesla's robotaxi service just went 24/7 in every city it operates, and Musk is pointing to an Optimus Academy training up to 20,000 humanoid robots. The real story isn't the schedule, it's the number Tesla still won't give us: how many cars are actually out there.
JUST IN: Tesla's robotaxi fleet doesn't clock out anymore.
Musk posted on X that the service now runs 24/7 across every city where it's live. His framing was classic Elon: a robotaxi "works 168 hours a week." Then, in a second post, he pointed to something called the Optimus Academy, a training site where as many as 20,000 humanoid robots will pick up new tasks.
Two announcements. One message. Tesla is selling hours now, not just cars.
The Clock Comes Off
Here's the sequence. First, the robotaxi service dropped its overnight pause. That pause wasn't decorative. It covered charging windows, cleaning, maintenance, the boring stuff that keeps a fleet alive. It's gone now.
Then Musk moved the conversation from wheels to hands. The Optimus Academy is a training ground. Up to 20,000 robots, learning tasks in one shared facility instead of a single unit shuffling around a lab.
And just like that, the pitch shifted from "we built a self-driving car" to "we built a labor force that learns."
What Actually Changes
Unit economics, mostly. A car that sits idle from midnight to 6am is dead capital. Six dead hours out of 24 is 25% of the day doing nothing. If Tesla recovers even half of that, revenue per vehicle climbs without a single new car hitting the road.
That's the whole game. Waymo has been grinding toward more operating hours for the same reason. Every idle hour is a competitor's hour.
The Optimus piece is a different bet, and a bolder one. A shared academy means shared learning. One robot figures out a task, the fleet inherits it. That's a data flywheel, and flywheels compound a lot faster than factories do.
But here's the thing. Tesla still won't say how many robotaxis are actually operating, how many cities count as "every city," or what the safety record looks like at 3am. Musk's 168 hours is a scheduling claim, not proof of scale.
My take: Musk is promising a robot workforce before he's finished proving the robotaxi one. That's not a knock. It's just the order he's chosen, and investors keep letting him get away with it.
So does a car that never sleeps make more money than one that does? Yes. But only if someone's paying for the ride at 4am in a city where the fleet is three cars deep.
The market's verdict: Tesla shares have been trading on the autonomy story for two years already. This is another brick in that wall, not a new wall.
What To Watch
Watch the city count. One city running 24/7 is an experiment. Twenty is a business. Tesla's next earnings call needs a fleet number, not another slogan.
Watch Optimus too. 20,000 robots in a training academy sounds massive until you ask what they do after class. Musk has floated humanoid production at scale before. The Academy only matters if it feeds a real assembly line, and soon.
And watch the payment rails. Autonomous fleets and robot labor both need machine-to-machine settlement, and that's where crypto actually earns its keep. Ride payments, robot wages, agent-to-agent transfers. If Tesla ever opens that layer to outside developers, the crypto angle stops being a thought experiment and becomes infrastructure.
Traders are watching closely. The 24/7 switch looks small. The signal isn't.
