Don’t Be Blinded by the Cool Demos of Humanoid Robots: Sun Ying, CEO of Hengdong Technology, Reveals the Truth About the Robot Business – It’s All About Making Money
Hello everyone, I’m your financial journalist.
What’s been the hottest topic in the tech world lately? Definitely humanoid robots. There have been countless press conferences and demo videos showing robots dancing, folding clothes, and even making coffee, which has thrilled everyone and made them think, “The future has arrived.”
But today, I’m going to pour some cold water on the hype, or rather, offer you a cup of “reality tea.”
I carefully analyzed an in-depth interview with Sun Ying, the CEO of Zhejiang Hengdong Intelligent Technology. Sun Ying isn’t the kind of entrepreneur who just makes grand promises in PPTs; she comes from a sales background at Huawei and has experience in robot vision. Now, she leads a company with an annual output value of 1 billion yuan, focusing on the practical implementation of robots. Her views are quite sharp and even a bit “counter-trendy”: Don’t just focus on how impressive the technology is; first, see who’s paying for it and how the math adds up.
If we compare embodied intelligence (the technology that allows robots to perceive and act like humans) to a marathon, the current state of the industry is like many people still at the starting line, competing over who has the best running posture, while Sun Ying’s team is already thinking about how to supply the runners, repair the roads, and ensure they can finish the race.
Below, I’ll break down the key points of this interview into five easy-to-understand points to help you understand why most robot companies don’t survive and where the real business opportunities lie:
---
1. Technological Leadership Doesn’t Equal Survival: Don’t Mistake “Showoff” for “Survival”
Many entrepreneurs have a misconception: as long as my technology is ahead of others, I’ll win. Sun Ying’s first failed startup experience shattered this myth.
1. Technology is a Ticket, Not a Moat
In her first startup, her team was impressive; they helped establish national standards and were among the first to use Intel’s depth cameras in robots, with cutting-edge technology. But what happened? The company still went bankrupt. Why? Because technological leadership didn’t solve practical problems like how to sell the products, how to coordinate the team, or where the market was. Technology just gives you a place at the table; whether you win depends on your skills and cash flow.
**2. “Stopping Losses” is More Difficult and Important than “Perseverance”
At Hengdong Group, Sun Ying has a strict rule for new businesses: if the direction is wrong or the results don’t meet expectations, stop investing immediately. For example, they tried robot dogs for factory patrols, but since they weren’t profitable, they’re now only maintaining the operation without further investment.
In plain terms: Many bosses are reluctant to let go of sunk costs, thinking, “One more investment and it’ll work.” But in the business world, “being able to do something” doesn’t mean it’s “worth doing.” Cutting off unprofitable projects and focusing resources on profitable areas is what mature companies do.
3. Capital is an Accelerator, Not a Lifesaver
Sun Ying believes that introducing capital (financing) when your product and market strategy aren’t clear can be dangerous. Capital forces you to expand quickly and spend money recklessly, leading to reckless decisions before you’re even established.
In plain terms: Money can help you move faster only when you’re sure there’s a market for your product and it can be sold repeatedly. If you haven’t even figured that out, money will only make things worse.
---
2. The Context Determines Survival: Ask “Who’s Paying” Before Discussing the “Technology Route”
This is the core idea of the entire interview. Sun Ying says: If you pick the wrong context, even the best technology is useless; if you pick the right one, you can improve your engineering capabilities over time.
1. What Makes a Good Context? Look at ROI (Return on Investment)
She gave a rather unappealing example: a pig farm.
Sounds boring, right? But pig farms are one of the best applications for embodied intelligence.
- Real Pain Points: The environment in pig farms is harsh (high temperature, humidity, corrosion), and people can get sick easily; frequent entry and exit can disrupt biosecurity (e.g., introducing viruses).
- Clear Cost Savings: Robots can inspect the farms, saving on disinfection, isolation, and labor costs. If robots can also detect diseases early, they can save on medication.
- Compared to Humanoid Robots: Humanoid robots are still in the lab; who’s paying for them? No one. Pig farm robots may not look attractive, but they replace specific operational costs, and owners are willing to pay for that.
2. Four Questions to Evaluate a Context:
Sun Ying provided a general framework for assessing any AI project:
1. Who’s paying? Is it the boss’s whim (one-time purchase) or the business department to save money (repeated purchases)? The latter is a good business model.
2. Is the ROI feasible? How much labor can robots save, and how long will it take to recoup the costs?
3. Is the environment tolerant of errors? Current robots make mistakes. If the task allows some errors (e.g., missing a pig during inspection), it’s suitable; if it requires 100% accuracy (e.g., surgery), current technology isn’t up to the task.
4. Can data be accumulated? If each project is unique, you can’t apply past experience to the next; you’ll only be able to do project-based contracting.
3. Logistics: Building on Existing Foundations
Besides pig farms, Sun Ying also mentioned logistics. Instead of chasing the hottest trends, they used their experience in industrial robots and the needs of Yiwu’s small commodity distribution center to develop intelligent warehousing.
In plain terms: Don’t reinvent the wheel. Use your existing technology and real-world needs; that’s the fastest path to success. Customers don’t care about the algorithms you use; they care about whether you can save them money and handle more goods in the same space.
---
3. A Working Demo Doesn’t Mean a Successful Product: Three Hurdles from “Demos” to “Actual Delivery”
Many robot companies fail at the demo stage. The robots look great on stage, but in the real world, they fail. Why? There are three major barriers:
1. Cost
Demos can use the best sensors and expensive chips, but products must be cost-effective. If costs can’t be reduced, customers won’t buy them, or if maintenance is too high, the business won’t be viable.
2. Stability in the Real World
The lab has a flat floor and uniform lighting, but factories, warehouses, and pig farms have dirty floors, changing lighting, and high temperatures. Can robots work continuously for thousands of hours without failing in such conditions?
3. Data Feedback
After deployment, can the data collected by robots improve the algorithms, making them smarter? If not, they’re just expensive machinery.
In plain terms: A demo shows you can do something, but a product must be stable, affordable, and repeatable. A robot company’s product is a comprehensive solution that includes the machine, software, services, and data iteration.
---
4. The Balance Between In-house Development and Outsourcing
There’s a common debate in the robotics industry: should all components be developed in-house?
1. In-house Development Depends on Value
Sun Ying’s approach is practical: Master core capabilities and buy standardized equipment. For example, in intelligent warehouses, stackers and shuttles are cheaper and more stable to buy. But core software, vision algorithms, and custom robotic arms must be developed in-house.
In plain terms: Ask yourself, “Is it worth developing it?” If it doesn’t give you a competitive advantage, outsource it. Focus on building your strengths.
2. Where’s the Real Barrier?
Sun Ying points out that relying solely on motion control or basic models is insufficient for survival.
- Without context-specific data, you’ll end up in price wars.
- A model alone is useless without practical applications.
The real barrier is a combination of core technology, context data, specialized models, and system integration. You don’t need everything, but you need to establish irreplaceable value in one of these areas.
3. The World Model: From “Understanding” to “Predicting
Sun Ying’s view on “world models” (technologies that help AI understand the physical world) has changed. She used to think they were too distant, but now she sees their usefulness.
In plain terms: Previous robots reacted to what they saw; world models allow robots to predict and act proactively. For example, they can avoid collisions by predicting the path of a rolling ball.
---
5. Team and Strategy: Generalization is the Goal, Not the Starting Point
Finally, Sun Ying offers advice for entrepreneurs and investors, as well as her ultimate judgment on the industry’s path:
1. Team Selection: Look for Ability to Handle Challenges, Not Just Resumes
When evaluating projects, she looks for founders who’ve faced real-world challenges.
In plain terms: Robotics is a long-term battle. The first year might be smooth, but the second year brings delays, customer issues, supply chain problems, and more. A team’s ability to stay calm and make informed decisions is crucial. Teams that only do demos in labs often can’t handle the challenges of the industry.
2. General vs. Vertical: What Are You Betting On?
Sun Ying’s controversial statement: “Generalization is the result, not the starting point.”
- Generalized approaches (like Tesla’s Optimus) bet on a technological breakthrough for all scenarios, but the risk is that the technology might never materialize, killing the company.
- Vertical approaches (like Hengdong and Figure) focus on surviving in specific contexts, accumulating data and cash flow before expanding. The risk is getting stuck in one area.
In plain terms: There’s no right or wrong choice; it depends on your risk tolerance. Ask yourself: What risk are you willing to take and do you have the resources to handle it? If you have funds and top-tier technology, go for generalization; if not, focus on a niche and survive first.
---
Summary: Insights for Everyone
1. For Investors/Observers: Don’t be fooled by cool videos. Ask three questions: Who’s paying? Is the investment long-term or one-time? Can the robot work reliably in harsh environments? If you can’t answer these, it’s likely just a PPT-based company.
2. For Entrepreneurs: Let go of the dream of changing the world; focus on practical, profitable scenarios. Calculate the costs and make your products stable; technology advancement is secondary.
3. For the Industry: The future of embodied intelligence won’t be dominated by one company. It’ll be a collaboration where some develop basic models and hardware, and others apply them in real-world contexts. The former sets the ceiling, while the latter ensures survival.
In Sun Ying’s words:
For embodied intelligence to move from demos to actual business, you need to prove not just that you can do it, but that it’s worth doing, profitable, and can be delivered repeatedly.
In this泡沫y era, who’s paying is the unavoidable and honest question.