American humanoid robots

The $5k is just the parts: this is a printer-and-solder project, not a purchase, so budget the weeks of building and the RL expertise it takes to make it walk.

This is a research kit, not a robot: for roughly $6,000 you get servos, a printed chassis, and Python code, then you assemble it, print the body, add your own Jetson and battery, and train the policies yourself. It runs about 19 minutes before overheating hurts stability, tops out at 0.25 m/s, and the mechanical design is licensed non-commercial. Buy it to do ML research; do not buy it expecting anything to work on power-up.

You are buying a wheeled, gripper-handed laundry-and-tidying appliance that leans on remote human operators when it gets stuck, ships to California first in fall 2026, and cannot be programmed, so treat it as an early-access home service, not an autonomous humanoid you own outright.

Behind the friendly demo, most real chores are done by a remote human piloting the robot through your home, so you are buying an early data-collection platform, not an autonomous butler.

This is a lightweight, gentle developer platform, not a worker: the 1 DoF grippers and ~2 kg-class payload mean light interaction (toys, gestures, guiding people), so budget $50K expecting to write your own applications on top of the built-in behaviors, and know EU support is not spelled out post-Amazon.

It does more real work per day than almost any rival, but at about $200k plus deployment and a yearly software fee (or $8,500 a month as RaaS), with paddle grippers instead of hands, it is an operations contract, not a robot you buy and tinker with.

You will never own this robot or see a price tag: Aeo is a subscription service commissioned by Aeolus's team for care facilities, so unless you run an institution in Japan, Hong Kong, or Taiwan (or can convince a distributor to pilot one), you cannot actually buy it.

There is no order form and no price: Apollo is an enterprise pilot program, so for anyone outside a Fortune 500 supply chain it is a look-but-cannot-buy machine.

It is the most famous humanoid alive and you still cannot have one: every 2026 unit is spoken for by its owner and Google, with outside orders not even opening until 2027.

You cannot buy this at any price: it is a rent-by-the-hour enterprise service and an invite-only home alpha, so for a normal buyer it teaches what is still out of reach.

You are not buying a robot, you are signing up for a roughly $100k-per-year supervised pilot with a young defense-leaning startup: it walks and manipulates under human-in-the-loop control today, but full hands-off autonomy, untethered battery runtime, and open developer access are still works in progress, not things in the crate.

Do not read '175 lb payload' as a robot that carries 175 lb: that number traces to a bicep-curl demo, the CEO's stated static payload is about 30 kg (66 lb), and as of mid-2026 the MK-2 is a fresh announcement, not a unit you can take delivery of. You are looking at a roughly USD 100k-per-year-class lease of a defense-leaning platform whose real MK-2 wins are ruggedness and manufacturability, not open access or hands-off autonomy.

You cannot buy this: K-Scale Labs ran out of runway and shut down in November 2025, cancelling all preorders and refunding deposits after shipping only two units. What is left is the open-source release (CAD, K-OS, K-Sim, K-VLA), so the K-Bot is now a build-it-yourself project with no company behind it, not a product.

You are signing up for a managed pilot of an early wheeled manipulator that is still slower than a human at real tasks (Kinisi's own head of product said so during its first glass-sorting deployment), and the company was just acquired by Bear Robotics, so what you are really buying is a pilot on a platform whose roadmap now belongs to Bear.

This is not a robot you can buy: it is an early-stage enterprise pilot platform sold through direct sales and Robot-as-a-Service, with published specs still partial and figures like payload and price varying by source, so budget for a negotiated deployment engagement, not a fixed sticker price.

You cannot buy this, there is no price, and it has grippers not hands: what you can actually get is an application into a roughly 50-household free beta, and the real story is the training method (a low-cost glove that captured about 10 million chore episodes from real homes) more than the hardware you take home.

Ignore the $20k-30k dream number. There is no way to order one, the dates have slipped repeatedly, the V3 has not even been shown, and as of Tesla's own July 2026 update the production lines are still being installed.