Home Companion Robots Are Actually Shipping Now. The Specs Tell You Why They Are Small.

While most humanoid robot coverage focuses on warehouses and factories, a quieter thing happened this month: consumer units started arriving at actual houses. UBTech’s UWORLD U1 companion series began its first consumer deliveries in mid-September, against more than 13,000 pre-orders.

The interesting part isn’t that they shipped. It’s what the specs reveal about the design constraints of putting a robot in a home versus a warehouse.

Small is a safety decision, not a cost decision

Look at the PrimeBOT Q1, shown at IFA 2026 and going on sale in China later this month: 88 centimeters tall, explicitly designed around portability and domestic safety, targeting academic research labs and home companionship.

Compare that to Tesla’s Optimus at 173 cm and 57 kg, built to deadlift 68 kg. That’s a machine that can seriously injure someone if a control loop fails at the wrong moment. An 88 cm robot physically cannot do the same damage — the mass and reach ceiling caps the worst case.

This is the same reasoning behind why UBTech positioned the U1 for home companionship rather than heavy labor. When the deployment environment contains children, pets, stairs, and unpredictable furniture, and there’s no safety officer or defined exclusion zone, the honest engineering answer is to reduce what the robot is physically capable of rather than trying to make the software perfect. Capability reduction is a much more reliable safety control than capability plus guardrails.

What “companion” actually means right now

Worth being blunt about scope, because the marketing language runs ahead of function. A companion robot in 2026 is broadly: presence, basic interaction, some mobility, some vision-based awareness of its environment. It is not doing your dishes, folding laundry, or handling anything requiring fine bimanual manipulation in an unstructured space.

That gap isn’t a failure of any one product. Manipulation in cluttered, variable environments remains the hard unsolved problem in robotics — considerably harder than locomotion, which is largely why the impressive demos you see are usually walking, running, or lifting rather than tidying a real room.

For context on how far locomotion has come versus manipulation: at the second World Humanoid Robot Games in Beijing in August, drawing 2,056 robots from 666 teams, Tiangong Ultra ran 100m in 8.64 seconds — faster than Usain Bolt’s 9.58 second human record — and cleared over 3 meters in high jump. Athletic performance is genuinely superhuman in narrow tasks. Picking up an unfamiliar object off a messy counter still isn’t solved.

Who these are actually for

Based on how manufacturers are positioning them, current home units make sense for three groups:

• Researchers and academic labs wanting an affordable, safe platform to develop on — explicitly PrimeBOT’s stated focus, and a real market
• Developers building for the platform ahead of broader capability, where early ecosystem position matters more than current functionality
• Consumers buying presence and novelty with clear expectations about what it does — which is a legitimate purchase, as long as the expectation is calibrated

Who these aren’t for: anyone expecting meaningful household labor reduction. That’s not cynicism about the category, it’s a description of where the technology currently is.

The price reality

Manufacturer-published prices across the broader humanoid market run from about $13,500 for the Unitree G1 to $100,000 for the H2 Plus. Consumer companion units sit at the lower end of that range, which is what makes 13,000 pre-orders plausible in a way it wouldn’t be at industrial pricing.

Tesla’s projected $20,000-$30,000 for Optimus at scale is worth holding next to those numbers — and worth remembering it’s a target at volumes that don’t exist yet, not a current price.

What to watch

The signal to track isn’t unit counts or pre-orders, which are easy to announce and hard to verify. It’s whether the second year of ownership looks like the first. Consumer robotics has a long history of products that demo well, ship, and then quietly stop being used within months because the novelty exhausts before the utility arrives.

The manufacturers that publish real retention and daily-use data, rather than shipment totals, will be the ones worth believing. Nobody is doing that yet.

Official reference: UBTech delivery reporting via Hindustan Times/SAG data (September 2026), BGR’s IFA 2026 humanoid coverage for PrimeBOT Q1 specifications, and manufacturer-published pricing compiled by TheresARobotForThat. Pre-order figures are company-reported and not independently verified — treat them as directional.