Does the Unitree G1 cost less than an iPhone? The price truth

No: a Unitree G1 does not cost less than a normal iPhone. The claim works as a social-media hook because both products are described as “consumer technology,” but the official numbers are not close. Unitree lists the G1 from $13,500, explicitly excluding tax and shipping. An iPhone comparison also changes by model and country, so the only defensible conclusion is that the robot costs many times more than a mainstream phone.

The more interesting question is why $13,500 still attracts attention. For a walking humanoid development platform, it is unusually low. For a household purchase, it is extremely high—and the sticker price is only the beginning.

What the base price buys

Unitree lists the G1 at about 1.32 metres tall and 35 kilograms, with a folded size of 690 × 450 × 300 millimetres. The base configuration has 23 degrees of freedom. Perception combines an Intel RealSense depth camera and Livox 3D lidar. A roughly 9,000 mAh quick-release battery is rated for about two hours.

Those are meaningful specifications for universities, robotics startups and developers. The form factor can enter spaces designed for people, and the sensors provide a foundation for mapping and interaction. But hardware specifications are not a promise that the robot can cook, clean or work unsupervised.

The product page separates the standard G1 from higher configurations. More advanced computing, dexterous hands and additional joints can raise capability and cost. Unitree also notes that some demonstrated functions remain under development. In other words, “from $13,500” is the entrance price, not the price of every robot seen in a polished video.

The costs hidden by a viral headline

Shipping a 35-kilogram robot is not equivalent to delivering a phone. Freight, customs duties, VAT or sales tax can materially change the invoice. A lab may also need spare batteries, replacement parts, protective space, computing hardware and staff time for programming and supervision.

Support is another variable. A consumer phone has mature repair networks and predictable apps. A humanoid development platform can require specialist maintenance and may have different warranty or service arrangements by region. Before buying, request a written quotation that specifies the exact model, hands, compute module, warranty, delivery and training.

There is also a safety cost. A 35-kilogram moving machine has enough mass to damage objects or injure someone. Research teams use controlled zones, emergency stops and risk assessments. A low headline price does not turn experimental robotics into an ordinary gadget.

What the videos do—and do not—prove

G1 footage demonstrates impressive balance and dynamic motion. That is evidence of mechanical control. It is not automatically evidence of autonomous decision-making. A sequence may be preprogrammed, teleoperated, trained for one environment or selected from many attempts.

For any demo, look for an uncut recording, a clear statement about autonomy, repeated trials and recovery from failure. Ask whether the robot recognizes a new object, plans the action and executes it without a human. Those criteria also guide our 2026 humanoid comparison.

Who should consider a G1?

The G1 can make sense for a funded lab, a company building embodied-AI software, or an advanced educational programme that understands the integration work. It does not make financial sense for someone who simply wants floors cleaned or laundry folded. A specialist appliance will be cheaper, safer and more reliable.

The truthful headline is less sensational: Unitree has pushed a humanoid research platform into the price range of a used car, not a smartphone. That is still important. Lower hardware costs allow more teams to experiment, which can accelerate software and safety research. It just should not be confused with a finished domestic robot. For that market, compare the practical options in which home robots really do chores, not a viral price analogy.