A robotic arm extracts espresso, adds a customized design, and serves the finished coffee through a glass window in less than a minute.
At the World Robot Conference in Beijing, the Cofe+ coffee robot from Hi-Dolphin Robot Technology drew attention with a machine nearly two meters tall and about 1.5 meters wide.
Founded in 2018, Hi-Dolphin was among the earlier Chinese companies to enter the coffee robot market. Its development work began in 2017 with a question: what technology would address the coffee industry’s needs? Over eight years, the product has gone through eight iterations.
Hi-Dolphin’s answer rests on the cost of running a cafe. Rent, labor, and renovation costs account for most of a conventional shop’s expenses, it said. The company’s robots can reportedly deliver comparable coffee quality and latte art with about 70% of the investment required to open a conventional store.
The company said its products have been deployed at nearly 4,000 locations worldwide, including Harvard University, Massachusetts General Hospital, the National University of Singapore, Denver International Airport, Kuala Lumpur International Airport, and Google’s headquarters. Its robots have also appeared at the World Economic Forum Annual Meeting in Davos and China’s annual Two Sessions political meetings. Altogether, they have produced about six million beverages.
Building the components for a robot cafe
Hi-Dolphin’s decision to build a coffee robot in 2018 started with the economics of running a cafe.
“We weren’t looking for a coffee use case for robots. The coffee industry itself needed a more automated way of operating,” said Han Feizi, co-founder of Hi-Dolphin.
Han ran the numbers. At chains such as Starbucks and Luckin Coffee, rent, labor, and the amortization and depreciation of store renovations account for 60–70% of total costs, he said, and those expenses continue to rise. Baristas who can consistently produce latte art are also expensive to train and tend to have high turnover.
Coffee production, meanwhile, is highly standardized. The drink is consumed frequently, production processes are mature, and aspects of flavor can be quantified. Equipment for grinding beans, extracting espresso, and frothing milk is already well developed, giving the industry a foundation for automating some human tasks.
In the first half of 2019, Hi-Dolphin installed its first prototype in an office building in Shanghai’s Changning District. Problems soon emerged: customers did not know how to use it, which drink was theirs, or how long they would have to wait after ordering.
The company also lacked a ready-made supply chain.
Cup lids were one problem. Baristas had traditionally capped takeaway coffee by hand, leaving little demand for an automatic lid-fitting machine. Hi-Dolphin also had to work out how its syrup dispensers and ice makers could precisely control ingredient ratios.
It had little choice but to design its own components and find factories to manufacture them.
“Developing components in-house was extremely painful, but once you make it work, it becomes a moat,” Han said. “We have applied for more than 120 patents, including invention patents, and built up more than 1,200 pieces of technical know-how and over 2,800 sets of process parameters.”
After resolving its supply chain problems, Hi-Dolphin focused on improving the product and customer experience. By the sixth generation, the robot could produce latte art and customized designs based on virtually any image.
“Whether latte art succeeds is not a question of how precisely the movements are controlled. It is a fluid dynamics problem,” Han said.
The robot uses fluid dynamics modeling to calculate how milk foam flows, spreads, and builds up on espresso, translating latte art into measurable physical variables.
When a customer uploads an image or text, an algorithm works backward in real time to calculate how to control the milk flow. The robotic arm then reproduces the design on the coffee’s surface.
The latest, seventh-generation robot further refines grinding and extraction settings to improve flavor while maintaining consistency.
At the Hongqiao Robot Coffee Competition in May, professional judges rated the seventh-generation product’s coffee at least as highly as that of human coffee masters on taste.
Five ways to sell a robot cafe
Hi-Dolphin said it has deployed its products at more than 4,000 locations worldwide. Only 40 are company-operated stores, used for product testing and customer feedback. Business partners operate the rest locally.
The company offers five commercial models to suit different customers:
- Under the agency model, customers use the Hi-Dolphin brand, with machines mainly deployed at airports, tourist attractions, and similar locations.
- Under the purchase model, customers buy the equipment and install it at properties they own.
- The customization model is aimed at major brands and large companies. Customers can use their own branding and customize the robot’s exterior.
- The leasing model targets financial institutions, government agencies, public institutions, and other organizations that are not permitted to engage in commercial activities outside their primary functions.
- Under the joint-operation model, designed for high-profile locations such as landmark buildings, Hi-Dolphin funds the equipment.
Hi-Dolphin has also integrated its supply chain for ingredients including coffee beans, milk products, and syrups. It adjusts flavors for different markets and develops new products, using chains such as Luckin Coffee as benchmarks for taste and variety.
“At its core, a robotic cafe is still a coffee service,” Han said. “Taste is what matters most.”
Hi-Dolphin adjusts the price of each cup to local coffee prices. It said the cost of opening a location remains about 70% of that of a coffee chain store.
Adapting to overseas infrastructure and standards
Hi-Dolphin began expanding overseas in 2021. Its first prospective order came from the organizers of Expo 2020 Dubai, who planned to buy 50 coffee robots.
The pandemic complicated that effort. International travel to and from China involved 28-day quarantines, equipment export procedures were cumbersome, and setting up machines for use overseas was difficult.
Han decided to proceed, despite the prospect of losing money on the Dubai expo.
“It was a rare opportunity to verify the product in real-world settings in other countries,” he said. “That included testing how it would adapt to different climates, voltages, payment systems, and consumer habits.”
Hi-Dolphin used the opportunity to establish a dedicated overseas team. Over the next three years, it deployed small batches of products in different countries to test their reliability and collect operating data.
Han listed some of the questions the company needed to answer:
- In a humid country such as Brazil, what happens when circuits and materials are exposed to moisture?
- In the dry, static-prone environment of the Middle East, could precision sensors be disrupted?
- In Japan, where stores can be small and doors and passageways narrow, can a wheeled machine move freely in and out? Could differences in water hardness and mineral content affect the final taste of the coffee?
Once the machines could operate reliably in different environments, the necessary national certifications were secured, and a remote after-sales support system was in place, Han concluded that Hi-Dolphin was ready to expand overseas on a larger scale.
Emerging markets in Southeast Asia and the Middle East were among its first targets.
Coffee consumption is growing in these regions, but infrastructure can complicate expansion. Some Middle Eastern countries still rely heavily on cash, while mobile payment platforms in Southeast Asia differ by country. Entering each new market may require a connection to a different payment system.
In many areas, communications networks range from 2G to 4G and can be slow or unstable. Some do not support internet-of-things devices, restricting the operation of robots that depend on remote maintenance and data transmission.
Hi-Dolphin responded by hiring specialists and developing an artificial intelligence-powered microcommunications system to manage connectivity, aiming to make its robots work across markets.
Mature markets including Europe, the US, Japan, and South Korea present different challenges. Consumers may be more comfortable with unmanned equipment, but they expect higher coffee quality, and compliance standards are stricter.
Food contact materials are one example. Extraction heads, milk tubes, water tanks, cup holders, and other components that touch coffee, milk, or water must meet the relevant requirements of the US Food and Drug Administration, the European Union’s EC 1935/2004 regulation, and Japan’s Food Sanitation Act. They must also undergo testing for substances that could migrate from the materials into beverages.
Cleaning is another challenge. Milk lines need standardized high-temperature sanitization procedures and traceable cleaning records. Unmanned equipment also requires automatic cleaning functions and reminders when cleaning is due.
Locations such as international airports and hospitals impose especially strict food hygiene and fire safety controls. Equipment may need additional safeguards against contamination, disease transmission, and fire.
Hi-Dolphin said it accounts for these requirements during product design, rather than seeking certification only after entering a market.
The company designs its coffee products to meet the regulatory requirements for professional cafes, with traceable operations, consumables, and cleaning. Production data for every cup can be traced to specific material batches and equipment settings.
As of August, Hi-Dolphin said its Cofe+ robots had reached 70 countries and regions and more than 200 international cities.
Han divides localization into two layers:
- The first covers software and content: interface languages, menu recipes, payment methods, brand colors, latte art designs, and beverage varieties. These changes leave the machine’s underlying structure intact and are relatively easy to implement.
- The second covers interfaces and compliance requirements, including voltage, plugs, local certifications, plumbing, water quality, and consumables. These must meet local standards, while components such as motion control, fluid dynamics algorithms, and mechanical architecture remain standardized globally.
“If you start customizing the underlying structure, the company will inevitably fall into chaos later on, and you will create huge hidden risks for the after-sales system,” Han said.
Hi-Dolphin plans to simplify its products by reducing the number of components, support more product categories, improve the customer experience, and add more intelligent functions. Overseas, it aims to adapt more closely to local markets and refine operations.
“Hi-Dolphin is targeting the new economy, new business formats, and new models,” he said. “We are using robotics technology to reshape the cost structure and consumer experience of the traditional coffee business.”
KrASIA features translated and adapted content that was originally published by 36Kr. This article was written by Feng Yaling for 36Kr.