25th August 2026
The second World Humanoid Robot Games in Beijing are certainly spectacular.
More than 2,000 humanoid robots from 16 countries are competing, and Chinese machines have already beaten human records in events including the 100 metres and standing high jump. One robot ran 100 metres in 9.39 seconds, faster than Usain Bolt's 9.58-second world record.
But running fast is not necessarily the important breakthrough.
The real question is whether a robot can walk into a factory it has never seen before, pick up a variety of objects, recognise what needs doing, use ordinary tools, cope when something goes wrong and work for eight hours without constant human supervision.
That is where humanoid robots could eventually become much more important than today's familiar industrial robots.
Why factories already use robots
Modern factories already contain enormous numbers of robots, particularly robotic arms.
These machines can weld the same point thousands of times, paint a car body with remarkable consistency, move components from one position to another and perform repetitive operations at speeds humans cannot match.
They are extremely good at one defined job.
The problem is that they normally require the factory to be designed around them.
A robotic arm might be bolted to the floor, surrounded by safety equipment and programmed to perform a particular sequence. Change the product significantly and the robot may need reprogramming or even replacement.
That is where the humanoid idea becomes interesting.
A human-shaped robot doesn't necessarily need the factory to be redesigned around it.
It could potentially use the same doors, stairs, workbenches, tools, shelves and machinery as the human worker it replaces.
That could make automation possible in factories where installing conventional robotics would be too expensive or inflexible.
The human body is actually quite an ingenious piece of machinery
We sometimes think of the human shape as an unnecessary complication for robots.
Why give a machine two legs, two arms and hands when wheels and a robotic arm might be better?
There is a very good argument for that.
But there is another way of looking at it.
Factories have been designed for humans.
The shelves are at human height.
Tools are designed for human hands.
Buttons and switches are positioned for humans.
Components are placed where workers can reach them.
Doors, corridors and stairs are designed around the human body.
A humanoid robot could therefore potentially enter an existing workplace without requiring the enormous infrastructure changes associated with conventional automation.
That could be its greatest advantage.
And artificial intelligence changes everything
The really important development isn't the legs.
It is the combination of robotics and AI.
A conventional industrial robot might be told:
"Move your arm from position A to position B 500 times."
A future humanoid might be told:
"Take these components from that shelf and assemble them according to this specification."
The difference is enormous.
The robot would have to see the objects, understand what they are, decide how to manipulate them, plan its movements and react if something isn't where it expected it to be.
That is the problem researchers are trying to solve with what is increasingly called embodied AI.
China's robotics industry is putting enormous resources into this. Reuters reports that Chinese companies are now moving beyond demonstrations towards logistics, manufacturing and other practical applications, with robots already being shown sorting parcels and assembling electronic components.
But we shouldn't get carried away
This is where the robot games can be misleading.
A robot running 100 metres faster than Usain Bolt is an extraordinary engineering achievement.
But it doesn't mean that robot can replace Usain Bolt — or a factory worker.
In fact, some of the machines taking part in the games still fall over, crash into barriers or struggle with relatively simple physical tasks.
And the commercial problem is even more difficult.
A factory doesn't care whether a robot can do a spectacular backflip.
It wants to know:
How much does it cost? How many hours can it work? How often does it fail? How much maintenance does it require? And does it produce more value than employing a human?
That is the test humanoid robotics now has to pass.
Reuters reports that the industry is increasingly moving away from spectacular demonstrations towards productivity, autonomy and return on investment. Current humanoids can cost roughly 300,000–500,000 yuan, which remains a significant barrier to mass deployment.
The first big breakthrough may not be humanoids everywhere
I suspect we will see something more subtle.
Factories are likely to use different types of robots for different jobs.
A robotic arm will remain better than a humanoid for a highly repetitive operation.
A wheeled autonomous robot may be better for moving materials around a warehouse.
A specialised machine may be better for welding.
And a humanoid could become useful for the jobs that are difficult to automate because they require flexibility.
That is already happening in Britain.
The Government has highlighted a UK-developed robotic system from Rivelin Robotics that gives industrial robots human-like precision and dexterity for finishing 3D-printed components. It is being used to automate work that has traditionally required people to manually clean and finish parts.
That is potentially more significant economically than a robot winning a running race.
The extraordinary possibility is the "general-purpose" robot
Imagine a factory where production changes during the week.
On Monday a robot loads components into a machine.
On Tuesday it moves finished products.
On Wednesday it uses a power tool.
On Thursday it helps package goods.
Today, that would generally require several different automated systems.
A sufficiently capable humanoid might eventually be able to do all four simply by being given different instructions.
That is the promise.
It would turn a robot from a machine designed for a task into something closer to a worker capable of learning tasks.
And that is why the development is attracting such enormous investment.
Xpeng's robotics division, for example, has just raised more than $900 million and is valued at over $6.3 billion, with plans to produce 1,000 humanoids a month by the end of 2026 and target industrial and retail applications.
China may have an important advantage
China isn't simply developing robots.
It already has an enormous manufacturing ecosystem capable of producing motors, batteries, sensors, electronics and mechanical components at scale.
That matters enormously because humanoid robots contain large numbers of precision components.
Reuters reports that China currently dominates humanoid robot shipments, accounting for around 82% of the global market.
China also has a huge manufacturing base in which to test them.
That creates a potentially powerful feedback loop.
More robots go into factories.
Factories generate data.
Data improves the AI.
Better AI makes robots more useful.
More useful robots create more demand.
More demand brings down production costs.
Lower costs encourage still more factories to adopt them.
That is how technologies can suddenly move from being expensive curiosities to ordinary industrial equipment.
But the UK has an opportunity too
This is something I think Britain should take seriously.
The UK doesn't necessarily have to manufacture millions of humanoid robots to benefit.
It could develop the software, sensors, specialist robotics, control systems and applications that make them useful.
There are already encouraging examples.
The Government-backed Rivelin technology is one.
And Siemens has partnered with British companies to establish a UK manufacturing capability for autonomous mobile robots, designed to move materials around factories and warehouses without fixed tracks.
The UK's problem is that it has historically been relatively weak at industrial automation compared with some competitors.
A 2025 Royal College of Art white paper examining the UK's robotics sector found a relatively small domestic ecosystem, dominated by SMEs.
That means the arrival of humanoid robotics could either become another technology that Britain imports or an opportunity to develop a new domestic industry.
And there is a huge labour-market question
This is perhaps where the subject connects with our recent discussion about falling graduate vacancies and employers reducing recruitment.
If robots eventually become capable of doing a much wider range of physical jobs, the effect won't be limited to factory workers.
Warehouse workers, delivery staff, cleaners, agricultural workers, construction workers and eventually some service occupations could be affected.
But there is an important distinction.
The first effect may be to fill vacancies rather than eliminate workers.
If a business cannot find enough people to work unpleasant, repetitive or physically demanding jobs, a robot may allow it to expand without recruiting another 50 workers.
That is quite different from a company employing 500 people today deciding to replace 500 people with robots tomorrow.
The transition could therefore be much slower than some headlines suggest.
There is another fascinating possibility
Humanoid robots could actually make Britain more competitive without necessarily creating mass unemployment.
Suppose a Scottish factory currently cannot compete with a Chinese factory because labour costs are much higher.
If both eventually use highly automated production, the importance of wage differences could diminish.
A Scottish factory might then be able to manufacture products competitively even though Scottish workers are paid considerably more.
That could potentially encourage reshoring — bringing manufacturing back to Britain.
And for places such as the Highlands, that could be significant.
A highly automated factory doesn't necessarily need thousands of workers.
It may need a smaller number of highly skilled people maintaining robots, managing production, designing products and handling logistics.
That could make manufacturing viable in places where a traditional labour-intensive factory would struggle.
The real robot revolution may therefore be much less dramatic than the films suggest
We probably won't wake up one morning to discover that millions of humanoids have replaced human workers.
It is much more likely that they will gradually appear in particular jobs where they make economic sense.
A robot might initially work alongside a person.
Then it might perform the most repetitive part of that person's job.
Eventually one person might supervise several robots.
And only later might businesses decide that certain jobs no longer need a human worker at all.
The important development will therefore not be "robots can walk like humans."
It will be:
"Robots can understand what needs doing and work out how to do it."
That is a much harder problem.
And interestingly, even the head of Unitree, one of China's leading robotics companies, has suggested that a major "ChatGPT moment" for robotics could still be two to ten years away.
So I wouldn't conclude that humanoid robots are about to sweep through British factories.
But I would conclude that something important is happening.
The robotic arm transformed manufacturing by making machines extremely good at individual tasks.
The humanoid robot is attempting something much more ambitious: making the machine adaptable enough to perform many different tasks in an environment designed for humans.
If that succeeds, the economic consequences could be much larger than anything demonstrated at the Robot Games.
And perhaps the most interesting race isn't actually between China and America.
It is between the countries that learn how to use these machines productively and those that merely watch them perform impressive tricks on television.