A Robot and Ice Cream

In the heart of Shanghai, on Wujian Street in the Jing’an district, a Dairy Queen café has had a robot working behind the counter since August 29. This is not an exhibition or a staged demonstration for cameras — the robot is actually working a full 10 a.m. to 10 p.m. shift, serving customers in a real queue.

The robot is called North and was developed by the Chinese company Sharpa. It moves on wheels, has two arms and performs what sounds like a simple but is actually fairly complex task: preparing Dairy Queen’s signature Blizzard. This is the ice cream that is traditionally turned upside down before being served to demonstrate that it does not fall out of the cup.

To prepare a single Blizzard, the robot has to perform around 55 consecutive actions. Sharpa describes the process as a chain of dozens of manipulations: taking the order, picking up the cup, adding ice cream and toppings, placing it in the mixer, mixing it, removing it, turning it upside down and handing it to the customer. On the company’s website, the number of operations is described as 50 to 60.

But the most interesting part of the experiment is not the number of movements. Sharpa emphasizes that the café did not have to redesign its kitchen to accommodate the robot. It uses standard equipment, regular cups and the existing workplace layout. In other words, the restaurant was not adapted to the robot — the robot was designed to work in an existing human environment.

The developers consider this a major step toward the commercialization of robotics. In their view, real success is not about making a robot perform a complicated task once for an impressive video. It is about being able to repeat that task reliably dozens or hundreds of times in a real-world environment, surrounded by customers, queues, noise and conditions that are far from a laboratory.

There is, however, one obvious limitation: North is currently slower than a human. It takes approximately 6–7 minutes to prepare one Blizzard, while an experienced employee can do the same job about four minutes faster. So for now, the robot is considerably slower than a person. Despite this, according to the company, reservations to see the robot in action have already been booked for the coming weeks.

And this raises an interesting question: why might a slow robot be more important than a fast one? Videos of robots are usually designed to be spectacular — robots run, jump, dance and perform complicated movements within seconds. But such videos are often carefully edited. North represents a completely different approach: the robot works 12-hour shifts in front of real customers, while the company openly admits that humans are still faster at the same task.

The key question is no longer simply, “Can the robot do it?” It is “How many times can it do it correctly in a row?” Reliability and the ability to operate continuously in an unchanged real-world environment may ultimately prove much more important than impressive speed.

Behind North are Sharpa’s three founders — David Li, Xiang Shaoqing and Sun Kai. Before entering robotics, they founded Hesai, one of the world’s largest manufacturers of lidar sensors for the automotive industry. Now their focus has shifted from automotive sensors to robotic arms.

There are, however, some important caveats. The figures regarding preparation time and the human speed advantage come from Sharpa through Chinese media; the company’s website does not provide detailed statistics on order completion times or success rates. It is also unclear how many robots are operating at the café and whether a human employee is standing nearby in case something goes wrong.

Still, the experiment is revealing. A robot that makes ice cream more slowly than a human may actually be far more interesting than a robot that dances beautifully for 30 seconds. The future of robotics will not be decided on a stage or in viral videos, but in places where robots have to perform the same task every day — again, again and again. And preferably without confusing a Blizzard with a mop bucket.