Tesla has set a target of building 20,000 Optimus robots a week, according to Gizmodo. That number reads like a product launch. It is closer to a wish list, and the detail that undercuts it is not the size of the target but the part it depends on: the hand.
The hand is the bottleneck
Gizmodo reports that automotive-style, large-scale manufacturing of intricate robot hands remains a major unresolved challenge for Tesla's Optimus scale-up plan. That detail matters more than the headline figure. Car factories are built to stamp metal, weld frames and assemble modules at volume. They are not built for parts with dozens of small actuators, tendons and sensors packed into a human-sized hand, which is what current humanoid designs, Optimus included, rely on for dexterous manipulation.
Hands are widely considered the hardest part of a humanoid to manufacture at scale: the most moving parts, the least tolerance for failure, and no automotive-line precedent for building them by the thousand. Until Tesla shows it has solved that specific problem, the 20,000-a-week figure is a target, not a delivery schedule. Buyers tracking the platform can check its announced specs against rivals on the Optimus Gen 2 page.
Agility Robotics hedges the form factor
While Tesla wrestles with hands, Agility Robotics is questioning a more basic assumption: legs. The Robot Report reports that Agility, maker of the Digit humanoid, is exploring wheeled robot designs alongside its existing bipedal platform, part of a broader effort to cover a range of designs and form factors for different operating environments and customer needs.
Digit is not a concept demo: it is Agility's existing, shipped bipedal platform, not a prototype. That a company iterating on a real product is still testing wheels alongside legs suggests the industry has not settled on one answer for what a humanoid should look like on the job. Legs handle stairs and uneven terrain; wheels are faster, more stable and mechanically simpler on flat floors. For many deployments, the biped form factor may turn out to be a means to an end rather than the point of the product, a distinction worth checking on the Digit page.
A third data point: state-backed humanoid ambitions
Humanoid robotics is also drawing interest well beyond warehouses and factories. Interesting Engineering reports that former Ukrainian Defense Minister Mykhailo Fedorov has unveiled a defense technology initiative aiming to field an army of humanoid robots to carry out risky operations alongside drones. The Decoder adds that this private-sector "Army of Robots" initiative envisions robots handling casualty evacuation, mine clearance and combat tasks.
That is a different use case from moving totes, but it points to the same underlying condition: manufacturing and design choices are still unsettled, whether the buyer is a logistics operator or a government.
What this means for buyers
Three data points, one conclusion. Tesla's production target is bold but bottlenecked by a hand-manufacturing problem the company has not yet solved. Agility Robotics, despite shipping a real platform, is still testing whether bipedal is the right form factor for every job. New entrants, including state-backed initiatives, are adding demand to a market that has not finished settling its basic engineering questions.
For anyone deciding what to buy, deploy or wait for, the practical filter is manufacturing readiness and form-factor fit for the task at hand, not the size of an announced target or the polish of a demo video. A robot that walks well on stage is not the same as a robot whose hands can be built by the thousand, or whose form factor matches the floor it will work on. Track deployment numbers and component-level manufacturing claims, not headline goals: the humanoid spec sheet is the place to check them against each other.