A humanoid robot is making Dairy Queen Blizzards in a working restaurant in Shanghai. It is an impressive demonstration of what physical AI can now do. It is not evidence that humanoid robots are ready to transform hospitality.
Understanding the difference between those two statements may be far more important than the robot itself.
On 29 August, robotics company Sharpa and CFB Group, which operates Dairy Queen in China, opened a robot restaurant on Wujiang Road in Shanghai.
Sharpa says its North humanoid robot can complete the sequence required to make and serve a Blizzard using the restaurant’s existing ingredients, equipment and processes rather than machinery or a workspace specifically designed for automation.
According to Sharpa, each order involves between 50 and 60 consecutive manipulation steps. The robot has to handle cups and spoons, open cupboards, add ingredients, operate equipment and complete Dairy Queen’s familiar upside-down presentation of the finished product.
That is technologically interesting because hospitality environments were designed around people.
Industrial robots have transformed manufacturing partly because factories can be designed around highly specialised machines performing predictable tasks. Restaurants, hotels and commercial kitchens present a much more complicated environment.
Doors need opening. Objects differ slightly every time they are picked up. Cups deform when gripped. People walk through workspaces. Equipment designed for human hands needs operating.
A robot capable of navigating that environment without the environment first being redesigned around the robot represents a potentially important step.
But potentially is an important word.
Behind the robotics headlines
Only two days before the Shanghai restaurant opened, Reuters published an extensive investigation into China’s rapidly growing humanoid robotics industry.
Its findings provide useful context.
China now has more than 150 humanoid robot companies. According to figures cited by Reuters, Chinese manufacturers accounted for around 95% of approximately 20,000 humanoids shipped worldwide last year.
Chinese national, provincial and local government bodies also spent at least $230 million buying humanoids and associated equipment during the first half of 2026.
Money, manufacturing capability and ambition are therefore not in short supply.
Practical capability is another matter.
Following visits to factories and robot training centres, examination of government procurement records and interviews with executives, investors and researchers, Reuters found a substantial gap between the ambitions surrounding humanoid robotics and what the machines can presently achieve in productive workplaces.
Dexterity, reliability, intelligence and cost remain significant barriers.
Robots that can perform impressive, choreographed demonstrations can still struggle when something unexpected happens during an apparently simple working task.
That distinction matters enormously to hospitality.
A restaurant does not need a robot capable of successfully making an ice cream once.
It needs one capable of doing so repeatedly, safely and economically, hundreds or thousands of times while dealing with all the small variations that constitute an ordinary working day.
The measure of success is ultimately not whether something can be demonstrated.
It is whether it can be relied upon.
Investment is running ahead of commercial reality
That does not mean humanoid robotics should be dismissed.
Quite the opposite.
Reuters’ investigation suggests China is deliberately building manufacturing capability, driving down component costs and creating the enormous quantities of training data required for embodied AI to improve.
Competition is already reducing prices.
The underlying bet is that producing and deploying large numbers of imperfect robots today will generate the experience and data necessary to develop much more capable robots tomorrow.
Whether that happens quickly is uncertain.
Reuters cites industry estimates suggesting the sector currently has around 500,000 hours of high-quality robot training data against a possible requirement of 100 million hours to achieve much more general physical intelligence.
Capital alone cannot immediately close that gap.
Yet capital continues to arrive.
DealStreetAsia reported on 31 August that Sharpa has raised more than RMB4.5 billion, approximately US$669 million, from investors including Alibaba, Meituan, Tencent, JD.com and others.
That is not proof that its robots will become commercially successful.
It is evidence of the resources now being committed to finding out.
Why hospitality should pay attention
For hospitality operators, therefore, the interesting question is probably not when humanoid robots will replace people.
There is little evidence today to suggest widespread replacement is imminent.
A better question is which individual pieces of work will become economically sensible to automate first.
Reuters found that industry participants expect near-term humanoid applications to remain concentrated around dangerous, repetitive and structured work.
That sounds remarkably relevant to hospitality.
Hotels, restaurants and catering businesses contain significant quantities of repetitive physical work, much of it carried out behind the scenes and much of it increasingly difficult or expensive to recruit people to perform.
Some of that work will be better suited to specialist machines than humanoids.
Some may never make economic sense to automate.
And some may eventually favour general-purpose robots precisely because hospitality workplaces were built for human beings.
This is why Sharpa’s experiment in Shanghai is worth watching without overstating what it represents.
The important aspect is not that a robot can make an ice cream.
It is that a robot is being tested doing a relatively complicated sequence of physical tasks, in front of customers, inside a real hospitality workplace using equipment designed for people.
Now comes the difficult part.
Can it do those things reliably?
Can it recover when something unexpected happens?
Can it work for sufficiently long periods without human intervention?
And, ultimately, can it perform useful work at a cost that makes commercial sense?
Those answers are not yet available.
From robots taking jobs to robots doing jobs
In February 2024, H&C News considered this subject under the deliberately provocative headline Bloody robots, coming over ‘ere, nicking our jobs.
The argument was that the most valuable role for robotics in hospitality might not be replacing hospitality professionals, but taking repetitive, mundane or physically demanding tasks away from them.
Two and a half years later, that still looks a more useful way of looking at the subject than asking whether robots are coming for people’s jobs.
What has changed is the amount of money, engineering effort and artificial intelligence now being applied to the problem.
Humanoid robotics remains considerably further from widespread commercial reality than some of its more spectacular demonstrations might suggest.
But development is happening.
The Shanghai Dairy Queen should therefore be viewed neither as a gimmick nor as evidence of an imminent robotic hospitality revolution.
It is something much more useful.
It is an experiment taking place in the real world.
For hospitality, watching what happens when the demonstration ends and the work begins could tell us far more about the future.
============================================
By Denis Sheehan, Publisher, H&C News
Reference Sources:
Reuters — China can build kung fu-fighting robots. But it can’t get them to do factory work
Sharpa — North at Work, Shanghai Dairy Queen deployment
DealStreetAsia — Sharpa funding and Shanghai robot restaurant
============================================
If intelligence becomes abundant, what happens to hospitality?
