A humanoid robot for the price of a used car
When the Unitree G1 was announced in 2024 with a starting price of $16,000, a lot of people in the robotics world did a double-take. That's not a typo. Sixteen thousand dollars — roughly what you'd pay for a decent second-hand car — for a 1.3-metre tall, fully articulated humanoid robot that can walk, manipulate objects, and yes, do a backflip.
For context: Boston Dynamics' Spot costs around $75,000. Figure's humanoid is aimed at enterprise customers who will pay hundreds of thousands per unit. Tesla hasn't announced a price for Optimus but has suggested it could eventually cost $20,000-$30,000 — at scale, years from now.
Unitree shipped the G1 for $16,000. Now. To researchers, developers, and anyone else willing to buy one.
That's a significant moment. Not because the G1 is the most capable humanoid robot ever built — it isn't — but because price is often what separates "impressive demo" from "actual technology wave." And Unitree just moved the price floor dramatically.
Who is Unitree?
Unitree Robotics is a Chinese company founded in 2016, based in Hangzhou. They're not a household name in the West, but in robotics circles they're well known for doing something unusual: building capable robots and selling them at prices that undercut Western competitors by a wide margin.
Their quadruped robots — four-legged dogs similar to Spot — have been available for a few thousand dollars while Boston Dynamics' equivalent costs many times more. This has made Unitree popular with university research labs, hobbyists, and developers who want to work with capable hardware without a six-figure budget.
The G1 follows the same philosophy applied to humanoids. Build something genuinely capable. Price it aggressively. Get it into as many hands as possible.
What the G1 can actually do
The G1 stands 1.3 metres tall and weighs around 35 kilograms. It has 23 degrees of freedom — meaning 23 independently controllable joints — with an upgraded version offering 43. It's fully electric, with a battery that gives roughly two hours of operation.
The demo videos Unitree released are genuinely impressive. The G1 can:
- Walk over uneven terrain and recover from being pushed
- Crouch, kneel, and get back up from the floor
- Use its hands to pick up objects, pour liquids, and handle tools
- Execute a standing backflip
- Operate at a walking speed of around 2 metres per second
The hand dexterity is particularly notable. Unitree showed the G1 cracking eggs, soldering circuit boards, and tying knots. These are tasks that require fine motor control — the kind of thing that separates a genuinely useful robot from one that can only handle large, predictable objects.
Why do robotics companies keep showing backflips? Because a backflip requires the robot to leave the ground, rotate its entire body, and land safely — all without external support. It's a demanding test of dynamic balance, motor coordination, and control software. If a robot can do a backflip reliably, it has a foundation of physical capability that translates to many useful tasks. It's a benchmark, not a party trick.
How Unitree gets the price so low
The honest answer involves several factors. China has a mature, cost-efficient supply chain for the components that go into robots — motors, sensors, batteries, structural parts. Unitree can source these at prices that companies manufacturing in the US or Europe simply can't match.
Unitree's engineering approach is also pragmatic rather than perfectionist. They build robots that are good enough to be genuinely useful, iterate quickly, and ship. Western robotics companies — particularly Boston Dynamics — have historically optimised for capability and demonstration value over cost and availability.
Unitree also accepts lower margins in exchange for market penetration. They're playing a long game: get their hardware into the hands of developers and researchers worldwide, build an ecosystem, and benefit from the software and applications that community creates.
What it's actually being used for
Because the G1 is already shipping, we have real-world data on what people are doing with it. Research labs are using it as a platform for reinforcement learning experiments — training robots to perform tasks through trial and error. The relatively low cost means a university lab can buy several units without exhausting its entire equipment budget.
Developers are building software stacks for it. Unitree provides an SDK and ROS (Robot Operating System) support, which makes it accessible to anyone with robotics software experience. The result is a growing community of people extending what the G1 can do.
Some early commercial applications are emerging in warehousing and light manufacturing — the same categories that Figure, Agility, and others are targeting with much more expensive hardware.
The limitations you should know about
The G1 is impressive for its price. It is not the most capable humanoid robot available. Some honest limitations:
Battery life. Two hours is short for industrial use. A robot that needs recharging every two hours is difficult to integrate into a continuous production environment.
Payload capacity. The G1 can carry around 2 kilograms. That's enough for light assembly work and many warehouse tasks, but it rules out applications involving heavy lifting.
Software maturity. Boston Dynamics has spent decades refining the software that makes Atlas move with such fluency. The G1's movements are capable but not yet at the same level of robustness across unpredictable real-world conditions.
Support ecosystem. Buying a Unitree robot means you're more on your own than with an enterprise product from a company with a large customer success team. For researchers and developers, that's fine. For a business deploying robots in a critical operation, it's worth considering.
The G1 isn't trying to beat Boston Dynamics at their own game. It's trying to make the game accessible to everyone — and in doing that, it's changing what "accessible" means for humanoid robotics.
Why the price matters more than the specs
There's a pattern in technology where capability and cost move in opposite directions over time. Early computers filled rooms and cost millions. Early smartphones were expensive, fragile, and limited. In both cases, the technology became transformative not when it became maximally capable, but when it became accessible enough for a critical mass of people to build on it.
Humanoid robotics is at an early stage of that curve. Right now, the most capable robots are also the most expensive, which means only well-funded organisations can work with them. That concentrates development in a small number of places and slows the overall pace of innovation.
When a capable humanoid robot costs $16,000, a university lab can buy three. A startup can afford to experiment. A developer anywhere in the world can buy one and build something. The breadth of people who can contribute to advancing the technology expands enormously.
That's not a small thing. The history of technology suggests it may be the most important thing.
The bottom line
The Unitree G1 is a real humanoid robot at a price that seemed impossible two years ago. It's not perfect, and it won't replace Boston Dynamics Atlas for cutting-edge research. But it will put capable humanoid hardware into the hands of thousands of developers, researchers, and companies who couldn't access it before.
The robots that will actually change how we live and work probably won't come exclusively from the companies with the biggest budgets. They'll be built by people who had access to affordable hardware, experimented relentlessly, and found applications that the original designers never imagined. The G1 is a tool that makes that possible.
Watch this space.