Summary of Key Points
This interview, based on the firsthand observations of Sun Pengfei, CEO of Zhenxiao Intelligence (who failed in his robotics startup in 2013), reveals the “truth” about the commercialization of the robotics industry, particularly in the areas of embodied intelligence and humanoid robots: Many orders are merely “pseudo-demands.” The sign of an industry downturn lies not in the cessation of financing but in the lack of customer repeat purchases or the ability to replicate sales. Business customers (B-side) are interested in “business outcomes” rather than technical concepts. Companies that develop “general-purpose robot brains” have limited prospects; it is industry-specific integrated solutions that are the key to profitability. Robots are only considered truly successful when they are included in customers’ cost calculations and acceptance reports.
1. The Failure Twelve Years Ago: The Wrong Order of Steps
When Sun Pengfei started developing a child companionship robot in 2013, he made a common mistake: he focused on technology by first comparing his product to competitors (such as Ubot) before identifying market demand. The first-generation product cost 3,000 yuan and was quickly abandoned by children after half an hour of play; the second generation, which included video call functionality (a first in China), sold for 4,000 to 5,000 yuan, but still failed to gain market acceptance. Later, when he shifted to commercial service robots, he still couldn’t find a model that customers were willing to pay for regularly, leading to the company’s closure.
His lesson is clear: “Don’t just do what others are doing; instead, ask yourself, ‘Do customers really need this? Are they willing to pay for it?’” As a young entrepreneur, he always wanted to create the product first, but now he understands that “whether it’s worth doing” is more important than “whether it can be made.”
2. The Signs of an Industry Downturn: Financing Cessation is a Symptom, Not the Cause
Many believe that the industry declines when funding stops, but Sun Pengfei believes the downturn begins with issues on the customer side. The sequence is as follows:
1. Problems arise on the customer side: No repeat purchases after small-scale tests (POCs); a benchmark customer cannot be replicated; robots are installed but rarely used.
2. Companies try to change their focus: They use the same technology for different applications (e.g., from catering to healthcare to education) in hopes of securing new orders.
3. Financing dries up: Investors realize there is no commercial progress and stop investing; the media shifts to new hot topics; talent and supply chains leave the industry.
Therefore, even when press conferences are still bustling, the industry may already be in a slump. The real test is whether orders can be replicated and sustained.
3. The “False PMFs” in the B-Side Market
The industry often sees claims of “thousands of orders” and “multi-million yuan contracts,” but Sun Pengfei believes many of these are “false PMFs” (products that do not match real market demand). Here are four critical questions to ask:
1. Where do the orders come from? Are they based on relationships with founders or investors, or are they from anonymous customers? Relationships can get you one order, but not ten.
2. Who is paying? Are the payments coming from government grants or corporate innovation budgets (using other people’s money), or from business departments using their own funds? The latter represents real demand.
3. Is the acceptance process rigorous? Real customers will ask: “Who is responsible for any issues? Has the efficiency met expectations? How many people are needed for maintenance?” If the acceptance process is merely formalistic, it indicates that no one is truly accountable for the results.
4. Can orders be converted into continuous business? Selling one robot to a single workplace does not equal establishing a sustainable market; a product must be useful across multiple customers.
Another factor is the long service lifecycle for B-side customers, which can last several years or even decades. Companies with stable cash flows and teams are more likely to secure long-term contracts.
4. The Future of General-Purpose Robot “Brains”
This year, “world models” (robot “brains”) have gained popularity, but Sun Pengfei is skeptical about the prospects of independent “brain” companies:
1. The barriers to developing general-purpose models are high: Creating a single model for all scenarios is challenging both in terms of algorithms and data collection.
2. Manufacturers prefer in-house development: Large companies in the electronics and automotive industries tend to develop their own software. Robot manufacturers will likely follow this trend as well.
The future will see a two-tier structure: a few companies will provide basic “brains,” while many others will develop industry-specific integrated solutions. The money will flow towards these latter companies, as they can offer valuable services such as scenario development, delivery, and maintenance. Pure software solutions are not profitable; they must be integrated into hardware and services to generate revenue.
5. Industrial Customers Buy “Definite Business Outcomes”
Sun Pengfei learned that industrial customers buy robots to solve specific problems, such as increasing production efficiency (OEE) from 92% to 94%. They analyze bottlenecks and the time required for cost recovery. Robots are a means to achieve these goals, not an end in themselves.
Another practical consideration in industrial settings is that robots must be reliable; even if they are fast, they must not cause delays in other parts of the production line. For example, if the slowest step in a production line takes 10 minutes, the rest cannot operate faster until that step is resolved. Therefore, while companies may talk about the capabilities of robots, industrial buyers are more interested in the tangible benefits they can bring.
Industrial customers also have practical requirements for “intelligence”: How often does the system need human intervention after 100 hours of operation? How long does it take to complete tasks? Can errors be traced? A successful press conference is not enough; industrial systems need to be foolproof. Large models will not directly control equipment; multiple layers of safety measures (such as algorithm boundaries and PLC backups) are necessary.
In summary, robots must be integrated into customers’ cost structures and acceptance processes to be considered a true success in the industry. This interview emphasizes that no matter how advanced the technology, it must always be grounded in business realities: Who is paying? Why are they paying? How can they continue to pay? These are essential questions that the robotics industry must address.