A robot at a Haidilao hot pot restaurant in China got stuck in celebration mode — and simply would not stop dancing. The machine, designed to perform a brief birthday song-and-dance routine for diners, malfunctioned and kept looping its festive gyrations long after the moment had passed. Staff couldn’t shut it down. Customers watched, some amused, some unnerved, as the bot continued its relentless, joyful performance at a table where nobody was celebrating anymore.
The incident, first reported by Business Insider, quickly went viral on Chinese social media platforms and spread to X (formerly Twitter), where users posted clips of the dancing robot with a mix of delight and existential dread. One widely shared post described it as “the most dystopian birthday party ever.” Another simply said: “This is the future we were promised?”
Funny as the video is, the episode touches something deeper than a software glitch. It lands squarely in the middle of an accelerating global push to automate the restaurant industry — a push that Haidilao, China’s largest hot pot chain, has been leading for years. And it raises a question that restaurant operators, technologists, and investors are increasingly grappling with: what happens when the machines designed to enhance customer experience start detracting from it?
Haidilao has long been famous for its over-the-top service. Free manicures while you wait for a table. Complimentary snacks. Staff who will fish noodles out of the broth for you. The chain, founded in 1994 in Sichuan province and now operating more than 1,400 locations worldwide, built its brand on a level of human hospitality that borders on theatrical. So when it began deploying robots — for food delivery, noodle-making, and yes, birthday celebrations — it wasn’t replacing warmth with cold efficiency. It was trying to layer automation on top of an already intense service culture.
The birthday robot is part of that layering. At many Haidilao locations, when staff learn it’s a diner’s birthday, a robot rolls to the table, lights up, and performs a choreographed routine complete with music and LED displays. It’s kitschy. It’s Instagram-friendly. Under normal circumstances, it lasts about 90 seconds.
This time, it lasted considerably longer.
According to the Business Insider report, the robot became stuck in its celebration loop, repeating the same dance sequence over and over. Restaurant employees attempted to intervene manually but were unable to override the programming on the spot. The machine eventually had to be physically removed from the dining area — still dancing, apparently, as it was wheeled away.
The footage is inherently comedic. But strip away the humor and you’re looking at a real operational failure in a high-traffic commercial environment. A piece of customer-facing automation that staff could not control. That’s not a punchline. That’s a liability question.
Haidilao’s automation ambitions are substantial and well-documented. The company opened what it called a “smart restaurant” in Beijing as early as 2018, featuring robotic arms that prepared soup bases and a system of automated conveyor trays that delivered dishes directly to tables. The goal was to reduce labor costs — which in China’s restaurant sector have been climbing steadily — while maintaining the brand’s signature service flair. Robots handle the repetitive tasks; humans handle the emotional ones. That was the theory.
In practice, the line between those two categories has proven blurry. A birthday celebration is, by nature, an emotional moment. Assigning it to a robot was always a gamble on customer tolerance — a bet that diners would find the novelty charming rather than hollow. For the most part, that bet paid off. The birthday robots became a social media sensation in their own right, with customers filming the routines and sharing them widely. Haidilao got free marketing. Diners got content.
But a robot that can’t stop celebrating flips the equation. Instead of charming novelty, you get uncanny persistence. The machine doesn’t read the room. It can’t sense that the moment has passed, that the laughter has turned to discomfort, that the table wants to go back to eating. It just keeps going.
This is a known problem in human-robot interaction research, sometimes called the “persistence error” — when an automated system continues executing a task beyond the point of social appropriateness because it lacks the contextual awareness to know when to stop. In industrial settings, persistence errors can cause physical harm. In a restaurant, the damage is reputational. But it’s damage nonetheless.
The broader restaurant automation industry is watching incidents like this carefully. Chains across the United States, Europe, and Asia have been deploying service robots at an accelerating pace since the pandemic exposed the fragility of labor-dependent operations. Companies like Bear Robotics, Pudu Robotics, and Richtech Robotics supply machines that bus tables, deliver food, and greet customers at thousands of restaurants worldwide. The global restaurant robotics market is projected to exceed $1.5 billion by 2028, according to multiple industry estimates.
Yet adoption hasn’t been without friction. In 2023, several U.S. restaurant chains quietly pulled back on robot deployments after finding that the machines slowed down service during peak hours and confused elderly diners. Chili’s parent company Brinker International tested Servi robots from Bear Robotics, then scaled back. The problem wasn’t that the robots didn’t work. It was that they didn’t work well enough in the messy, unpredictable reality of a busy restaurant floor.
Haidilao’s dancing robot incident is a different flavor of the same fundamental challenge. The technology functions — it dances, it plays music, it lights up. But it lacks the fail-safe mechanisms and contextual intelligence needed to handle edge cases gracefully. And in customer-facing roles, edge cases aren’t rare. They’re constant. A child grabs the robot. A table asks it to stop. The Wi-Fi drops mid-routine. The software enters a loop. Every one of these scenarios requires either sophisticated onboard decision-making or a simple, reliable manual override. The Haidilao robot apparently had neither.
For Haidilao specifically, the stakes are higher than for most chains experimenting with automation. The company’s stock, listed on the Hong Kong exchange, has been volatile in recent years as it expanded aggressively, then retrenched, then expanded again. Investor confidence hinges partly on the narrative that Haidilao can scale its famously labor-intensive service model through smart use of technology. A viral video of a malfunctioning robot doesn’t destroy that narrative. But it chips away at it.
The company has not issued a detailed public statement about the incident as of this writing. Haidilao’s communications team has historically been adept at turning potential PR problems into brand-building moments — responding with humor, offering compensation, leaning into the absurdity. It would be surprising if they didn’t attempt something similar here.
And to be fair, the internet’s reaction has been more amused than angry. On Weibo and Douyin, the dominant sentiment was laughter, not outrage. Many users expressed sympathy for the restaurant staff, who were visibly flustered in the videos. Some joked that the robot was simply “too dedicated to its job.” The meme potential, in other words, is high. The reputational risk, for now, is low.
But memes fade. The underlying technical and operational questions don’t. If you’re deploying customer-facing robots at scale — across hundreds or thousands of locations, in environments with spilled soup and screaming toddlers and Wi-Fi dead zones — you need those machines to fail gracefully. A robot that gets stuck in a loop and can’t be easily overridden by floor staff is a robot that isn’t ready for prime time, no matter how cute its dance moves are.
The restaurant industry’s automation push isn’t slowing down. Labor shortages persist across major markets. Wage pressures continue to mount. And a new generation of diners, raised on self-checkout kiosks and delivery apps, is broadly comfortable interacting with machines. The question isn’t whether robots will become a permanent fixture in restaurants. They already are.
The question is whether the companies deploying them — and the companies building them — are investing enough in the unglamorous engineering work that makes automation reliable in chaotic real-world conditions. Sensor fusion. Fail-safe protocols. Manual override systems that actually work when a 22-year-old server is panicking because the birthday robot won’t shut up. That’s the hard part. That’s where the money and attention need to go.
A dancing robot is easy to build. A dancing robot that knows when to stop? That’s the real engineering challenge. And until the industry solves it, we’ll keep getting videos like the one from Haidilao — entertaining, viral, and quietly revealing of just how far restaurant automation still has to go.
WebProNews is an iEntry Publication