Author: Wanxiang Blockchain

From April 20th to 23rd, the 2026 Hong Kong Web3 Carnival, co-hosted by Wanxiang Blockchain Labs and HashKey Group, was successfully held at the Hong Kong Convention and Exhibition Centre. On April 21st, on the stage of the "One Wind" venue, Alven Lin, Partner of Fintech and Blockchain at Summer Capital, moderated a roundtable discussion with Ben Zhai, CEO of Robo.ai, Chen Jianqiu, Executive Director of Hong Kong Robotics Group , and Sun Zexuan, CTO of Dreame Technology's AI Hardware Division, on the theme of "From 'Selling Products' to 'Selling Services': Exploring the Subscription Economy and Assetization of Smart Hardware ." The following is a summary of the discussion, with some omissions that do not affect the original meaning.

Alven: Welcome to the Hong Kong Web3 Carnival. I'm Alven, the moderator of this roundtable and a partner in the Finance and Blockchain division at Summer Capital. Summer Capital is a multi-strategy investment firm in Hong Kong that has been investing in the crypto market since 2017. We are also one of the major investors and founders of Solana Company, the second-largest DAT company in Solana, and we held a joint event here this morning. Today, we are honored to have three heavyweight guests from the smart hardware device field to discuss the subscription economy model and potential development opportunities of smart hardware with us. First, could the three guests give a brief self-introduction?

Ben: My name is Ben, and I come from a company headquartered in Dubai and listed on Nasdaq. Our company has undergone several transformations since its IPO, especially in the last year or two, which aligns well with today's theme and the conference theme. I'll share more details about that later. Our company is called Robo.ai.

Chen Jianqiu: Hello everyone, I'm Chen Jianqiu from Gangzai Robotics. Gangzai Robotics primarily provides one-stop robot solutions, covering the entire chain from product development and scenario expansion to operational services and subsequent RWA asset tokenization. I'll share more details with you later, thank you.

Sun Zexuan: Hello everyone, my name is Sun Zexuan (Spur), from Dreame Technology. Dreame Technology primarily focuses on robotic vacuum cleaners. Previously, I was a researcher in the blockchain lab at École Polytechnique in Paris, where I successfully launched a rocket. Currently, I am the technical lead for the AI ​​hardware department, possessing expertise in both AI hardware and blockchain. So, I'd like to share some insights with you today. Thank you.

Alven: Thank you to our three guests for their self-introductions. Our guests today have very diverse backgrounds, including robotics, AI, B2B hardware sales experience, and B2C product line experience. The subscription economy business model for hardware is common in the software industry, often referred to as the SaaS model. However, it's not as prevalent in the hardware sector. There are some innovative models, such as the battery swapping services offered by some new energy vehicles, which are essentially a hardware subscription mechanism. But not all hardware devices commonly adopt this subscription approach.

Today, I'd like to discuss the following topics with our three guests: past experiences and challenges in smart hardware sales; explorations and reflections on the transformation from selling products to selling services; the practice of assetization and RWA derived from the hardware subscription model; and new trends that the hardware industry may face under the current AI model. Let's start with the topic of hardware sales, which is familiar to everyone. Your companies all have extensive experience in selling products to both B2B and B2C, covering robots, aircraft, and diverse C-end products. In this process, I would like to ask each guest to share some of your successful experiences or challenges in traditional hardware sales. Did these challenges prompt you to gradually shift towards selling services? Mr. Jianqiu, please begin.

Chen Jianqiu: Okay, thank you, Alven. From the perspective of selling services, this is actually a very rigid demand right now. There's still a significant gap between current robot products and genuine user services. We've contacted many B2B clients who truly want solutions to problems in their scenarios, such as having robots patrol or provide health guidance. They want services; the product is merely a tool to fulfill those services. But currently, the gap is still quite large. Clients might approach robot manufacturers, but manufacturers are unwilling to provide services, only willing to provide products. Therefore, we've done a lot of exploration, including our self-developed cloud-based brain, which is also designed to serve operators. It requires an operations service personnel to help scenario providers implement these functions.

Here's an interesting point: it's difficult to educate customers and turn them into robotics experts. However, we can build an operational platform where operators acquire the necessary knowledge and tools, directly empowering the scenario providers. The ultimate effect of the service is cost reduction and efficiency improvement, which directly generate economic benefits. Once economic benefits are achieved, many issues, such as product pricing, are easily resolved. That's about it, thank you.

Ben: I've been involved in several startups with subscription service elements over the years. The earliest subscription or SaaS model actually started with Salesforce in the late 1990s. At that time, Salesforce was enjoying great success in the capital markets, and everyone felt that the SaaS model was becoming incredibly popular. Later, this model gradually expanded from software services to hardware and other fields. As the host just mentioned, I was also fortunate enough to participate in the founding of several new energy vehicle companies in China more than a decade ago. One of them was very influential; this company can be said to have made a revolutionary innovation in the entire automotive industry. Today, new energy vehicles account for half of the annual new vehicle sales in mainland China. But twelve or thirteen years ago, people had a strong fear of new energy vehicles because of range anxiety and safety concerns.

Our founder, Mr. Li Bin of NIO, took a revolutionary step in addressing this user anxiety: separating the vehicle from the battery, since the battery accounts for about 40% of the total cost. To reduce concerns about range anxiety, safety anxiety, and long-term value degradation, he created a completely new model called BaaS. We know that SaaS stands for Software as a Service; Mr. Li Bin created BaaS, which stands for Battery as a Service. This model can be considered the first subscription-like model used in truly large-scale intelligent equipment.

I also worked for another new energy vehicle company, which later went public on Nasdaq. At the time, they claimed to be a company truly offering subscription-based services. So, it's fair to say that several companies started experimenting over a decade ago, but there were many challenges, which I believe the host will mention later. However, I personally feel that whether it's subscription-based, BaaS, SaaS, or the blockchain-based transformation of various smart devices in the future, they are essentially the same. The core intention is to lower the barrier to entry for users, reduce heavy asset investment, and maintain a continuous cash flow. The ideas are the same; it's just that the products we face are becoming increasingly different. Next, let's hear from Dreame and several other guests.

Alven: Dreame is now one of the leading companies in China and even globally in terms of shipments of smart home appliances and smart hardware devices. Could you share your most successful experience? In the process of selling so many product lines and through international sales channels, what have been your most successful experiences? What challenges do you foresee you might face?

Sun Zexuan: Actually, I come from Dreame's AI hardware department, and our main product is currently the Dreame AI Ring. So today, I'll mainly be answering your questions about subscriptions.

Currently, in the AI ​​ring field, there's a leading player called Ora Ring, whose business model is primarily subscription-based, and it currently has approximately 2 million subscribers. This indirectly proves that Ora is essentially a software company rather than a hardware company. Our overall strategy for pursuing AI rings is to attack Ora's subscription-based competitors by offering users free basic services through a one-time purchase. We hope to quickly attract price-sensitive users who dislike subscriptions, especially in the Chinese market.

However, in the long run, the overall value of AI hardware tends to stabilize. Therefore, AI-enhanced services combined with high-quality data will continue to grow in value. So our judgment on the future is very clear: the future will definitely be based on AI service-oriented business models. As for why Dreame iterates so quickly, this is mainly due to Mr. Yu's strategy and vision. He empowers every entrepreneur, every business unit leader, and every product manager with greater drive and gives us more opportunities for innovation, resulting in our very rapid iteration speed. That's about it.

Alven: Okay, thank you Spur for sharing. Through the presentations of the three speakers, you might have realized that hardware sales and subscription models are not mutually exclusive or mutually exclusive. In many cases, acquiring customers through hardware sales is the most basic service behavior, while subscription service models are used to improve user stickiness or extend the paid lifecycle.

Sun Zexuan: Let me add some details about our strategy. We use high-value hardware as an entry point to acquire high-quality data and build long-term relationships and trust with users. In the future, we will leverage cutting-edge AI and high-quality data to build deeply personalized SaaS services that users are willing to continuously pay for. We are also driven by software.

Alven: I understand. In this process, hardware subscriptions and hardware sales are essentially integrated. Compared to traditional pure software subscriptions, what are the subtle differences in operation between this model and other methods? In this process, what do you think are the barriers for users to purchase subscription services? Or, what is the ultimate motivation for users? What potential challenges have you faced?

Sun Zexuan: I'll still use the AI ​​ring as an example, because it's our main business. Our AI ring will launch in October 2025 and will be featured on CCTV's Spring Festival Gala in 2026. After AWE 2026, we received 100 million RMB in orders, with monthly sales of approximately 5 to 10 million RMB. We are currently expanding into overseas markets.

Our mission and vision are: I believe a ring is the best sensor entry point. Our future goal is to create a next-generation AI interaction interface. A ring has two advantages: first, it possesses a clear ability to perceive bodily data; second, it can integrate AI, possessing powerful AI interaction capabilities and the ability to understand the user. Based on this, it is an excellent sensor and entry point. Therefore, more users are willing to accept such devices, interact with them, obtain their personal bodily data, and thus better serve themselves.

Chen Jianqiu: Actually, our scenarios are somewhat different because our products have a higher value. For example, the charging patrol robot we're currently developing costs tens or hundreds of thousands of yuan per unit. Many clients feel that the initial installation cost is too high, potentially requiring an investment of hundreds of millions of yuan. So we consider this aspect carefully. Our goal is to provide excellent service to operators. In addition to the cloud-based brain mentioned earlier, we are also exploring one-stop operational services, including operational platforms and personnel. Furthermore, we are actively exploring the dedicated RWA track in the robotics field. We have already launched our first robot RWA platform and will soon launch our first robot RWA project. This is actually helping project owners shift from a one-time purchase to a monthly subscription model. From a first-principles perspective, people are more accustomed to paying monthly fees. For example, the biggest application market in the future is robot nannies; people might be more accustomed to paying a few thousand yuan per month for its service, rather than spending tens or hundreds of thousands of yuan upfront to buy it outright. Therefore, subscription models make it easier for high-value robots to enter every household.

Ben: Mr. Merlin just gave an introduction; their product is called Arkreen. This company is a Singaporean blockchain company that we co-invested in with Hash Global. I can use our investment in Arkreen to explain your question. Like the cars you mentioned, costing 300,000 to 400,000, or even 500,000 to 600,000, the battery accounts for 40% of the cost. Removing that 40% significantly lowers the barrier to entry. Your robots are also very expensive, often costing hundreds of thousands. Arkreen's individual product price isn't high, perhaps only a few hundred to 500 US dollars. But they are now considering expanding to Africa. For African merchants or end-users, 500 US dollars is not a small amount. So, to some extent, whether it's the SaaS model, the BaaS model, or the future RaaS model for robots, the ultimate goal is to lower the barrier to entry. If charging costs one cent for one minute or 10 cents for half an hour, this approach lowers the barriers to use and emergency situations for everyone. Therefore, although the SaaS or subscription model has existed for nearly thirty years, I believe that with the continuous emergence of services, hardware, software, AI, and various robots and new products, especially in new scenarios where people are unfamiliar with the products and lack understanding of security and durability, the subscription model or variations of the traditional SaaS model is a very good model. This is also related to the next step of tokenization—totalization is essentially breaking down large things into smaller ones.

Alven: Yes, so the guests have raised another topic, which is also what I'd like to ask, or discuss with everyone: When the subscription model brings a new cash flow mechanism, how do we assetize it under this mechanism, or even implement a concept that's currently very popular in Hong Kong's Web3—tokenization, also known as asset tokenization or RWA? I've seen a lot of discussion on this topic at this conference. How did you build the entire RWA architecture? From the legal structure to sales, and then to the user interaction platform, could you share your current explorations or practices from your respective perspectives? Legal structure is one aspect, then how do we securitize assets, and then tokenization; after tokenization, how do we interact with Crypto Native DeFi protocols, and how do we interact with the community? What experiences can you share here?

Ben: Let me take Arkreen, which we invested in last year, as an example. When we went public in 2021, we listed on Nasdaq with the concept of new energy vehicles and intelligent vehicles, and we have since undergone several adjustments. In fact, last year we believed that the integration of smart assets and intelligent assets was a very viable way out. So starting last year, we made different investments and attempts in blockchain companies, DePIN companies, and the tokenization of smart hardware.

I've always maintained that our investment in Arkreen, for a Nasdaq-listed company that evolved from traditional new energy vehicles, is an entry ticket, or rather, a tuition fee, or an advertising fee, to the Web3, blockchain, and digital currency markets. Last year, besides investing in Arkreen, we also tried to transform our existing new energy vehicles into smart cars, then into driverless cars, and further extend this to other unmanned and intelligent equipment. Last year, we also announced Robo.ai in Dubai, claiming it to be the first smart car with a digital wallet and digital ID. A digital wallet and digital ID mean that each device has a unique identifier. We are currently discussing collaborations with Arkreen and other partners, exploring how to transform all future driverless cars and robots into devices with independent digital wallets and digital IDs. As payments between people gradually shift to payments between machines, how can we quickly realize the so-called "machine economy" of the future? I believe it's not far off. Therefore, we are making various attempts, and today we hope to use this opportunity to find more partners.

Speaking of the concept of RWA (Real World Architecture), this term has actually been around for four or five years. Initially, it mainly referred to real estate projects. But today, through attending the Hong Kong Web3 Festival these past two days, I've sensed a significant change—compared to last year, although our scale this year may be slightly smaller due to various factors, everyone is clearly placing greater emphasis on hardware and the real world. Last year, 90% of the discussions were still about the traditional approach. But this year, look, our first opening speaker today is the CEO of Lotus Cars, a traditional automaker. Although I don't think his content has many substantial innovations related to Web3, having him as the opening speaker sends a very positive signal: everyone is moving from the original focus on the virtual world, digital assets, and smart assets towards integration with real-world assets.

Alven: Okay, thank you. Jianqiu, do you have any experience to share? About your products and your RWA division.

Chen Jianqiu: Okay. RWA is indeed very popular right now, and it's a major topic of this conference. Currently, RWA exists more as a financial tool, and the core requirement is that the underlying assets are of high quality. In the early stages, the return on real estate-related RWA may not be particularly high, but now, with the maturity of robotics technology and products, the return on robotics and its reach will become increasingly larger.

Let me give you an example, using one of our main charging projects as an example. We're doing destination charging, which involves deploying mobile charging robots—charging vehicles—in underground parking lots of residential communities and office buildings. This is actually a great example. First, from the user's perspective, we can integrate with stablecoins, allowing them to pay for electricity with stablecoins. On the RWA side, after a user purchases a share of the robot, their stablecoins are credited to the contract and transferred directly to the user according to their share. The entire process can be completely transparent and open; it's a very natural combination.

Secondly, our robot products themselves generate significant cash flow. We've calculated that the return on investment is actually very high. We also offer operational services and cloud-based AI services, where all data can be encrypted on-chain – a direction we're exploring in our collaboration with Arkreen. This includes data from the charging process, all of which can be recorded on the blockchain. This is a very good point of integration between robotics and Web3. We're not limited to our own products; we welcome partners like Ben and Dreame to join us in exploring the Robotic Robotics RWA (Resource-Based Automation) field. We also welcome more partners to participate in discussions at this event. Thank you.

Alven: I'd like to ask if Dreame has any plans or explorations in the Web3 field? What are your views on RWA under the subscription economy model, or the possibilities of RWA?

Sun Zexuan: Our business unit isn't currently focusing too much on hardware assetization; we're mainly concentrating on data assetization and data privacy. However, I think whether hardware highly tied to data can undergo RWA (Responsive Wireless Assessment) transformation is a direction worth considering. That's about it.

Alven: Yes. Just now I thought the convenience that blockchain brings to assets isn't just transparency, but also fairness and accessibility. At the same time, a lot of data can be on-chain, making data flow clearer and data sovereignty more complete. Furthermore, there's a unique economic model in the Web3 space called token incentives. Many Web3 projects use tokens to incentivize user behavior. If hardware subscriptions are also a long-term user payment behavior, could a "incentive + user subscription" approach create a flywheel effect in this field? Encouraging users to actively provide feedback on their usage behavior, even on-chain data, and then receiving incentives to unlock more advanced services and products. Do you think this mechanism is feasible?

Sun Zexuan: I think it's very feasible. I believe the biggest points of convergence between AI hardware and blockchain are twofold: first, trust, and second, incentives. Regarding trust, we hope to protect user privacy data through privacy-preserving computations in blockchain cryptography—that's the first point. The second point is tokenomics, or tokenization. In China, this can be understood as a points system. Because the AI ​​ring can monitor users' physical data in real time, including sleep data, heart rate, and exercise data, we hope to incentivize users with tokens when they engage in positive behavior. In the future, these tokens can be used to offset subscription fees, thereby enabling access to more advanced AI-enhanced services. Therefore, tokenomics has a strong connection and convergence with blockchain and fintech. I particularly agree that blockchain or Web3 has two technological pillars: first, cryptography, and second, fintech. These two will undoubtedly drive the entire Web3 forward and accelerate the efficiency of human operations.

Alven: Okay, thank you. Jianqiu , what are your thoughts?

Chen Jianqiu: Actually, this is a lot of fun. Incentives can encourage our users to participate more in the ecosystem. For example, in the health field mentioned earlier, users can upload more data, which means their health data is recorded more completely; and for our organization, it also allows us to train better models. For the robotics business, incentives can encourage more users to participate in the robotics ecosystem, achieving human-machine coexistence. For example, when charging, users might need to unplug or plug in the charging gun. Users can participate in this action and receive token rewards, which they can use when charging themselves.

On the other hand, if a robot encounters temporary problems and users help resolve them, it can be integrated into the ecosystem, thus reducing our operation and maintenance costs. So this is actually a very good technology and behavioral model.

Ben: I think there's a paradox in subscription models, tokenization, and the lowering of barriers to entry that I just mentioned. What is this paradox? On the one hand, tokenization, SaaS models, and BaaS models lower the barriers to entry and break down the value; but on the other hand, to truly use this model effectively, the barriers to entry are actually higher.

I had the privilege of serving as the Chairman of the NIO User Trust for several years. NIO excels in user service, resulting in extremely high user loyalty. Nearly 70% of NIO's new car repurchases come from referrals by existing users. This is an unimaginable figure in the automotive industry. However, the effort involved is immense. I believe that all subscription-based, tokenization-based, or value-decomposing models, or SaaS models, face three hurdles:

First, capital. You need a huge amount of capital investment.

Second, consistency. The user experience must be consistent for me to continue renewing my subscription—whether it's next year or next month.

Third, trust. As Dreame just mentioned, for service providers or hardware providers, it's our commitment. What I'm doing isn't a one-off deal.

Only when capital, consistency, and commitment are all met can a true moat be built. Otherwise, if the moat isn't built well, one could quickly drown. So I don't think this is that simple. Whether it's the logic of Web3 or blockchain, at its core, it's not much different from traditional business. At its core, it's about human nature.

Alven: Yes, I think the next stage of Web3 development should return to the traditional essence of business. The design of token economics cannot be purely treated as a utility token or incentive token, as we discussed before. It must return to how to empower projects or companies, creating value and stickiness in the process of providing services to users. The guests have repeatedly mentioned one word—AI. The development of AI in the past few years has been remarkable. Underlying basic models, embodied intelligence, even new world models, and the combination of AI and crypto are all being widely discussed. I would like to ask the three of you to introduce your companies' current AI strategies or product lines. In the AI ​​era, what significant driving force do you see for existing hardware, products, or services? Or are you also facing many challenges, requiring transformation and reform? In this process, will AI change users' subscription habits and behaviors?

Ben: We changed our listed company name to Robo.ai last August. I think this is a very interesting decision we made as a traditional new energy vehicle listed company, and it's also a signal. But the latter part of the question is more important. After changing our name, we restructured our business model, team, and board of directors. What's left? What's left is truly returning to the essence of business. We have real cash flow, real revenue, real profits, a real team, and our own unique technology. So for us, the challenge is enormous. That's why I really hope to take this opportunity to cooperate with partners like Hong Kong Robotics and Dreame. Actually, Dreame was also preparing to make cars last year and approached me; their boss went to Dubai several times. I hope to use this occasion to have truly deep, practical, and tangible cooperation with everyone. I also want to thank Arkreen for providing us with this platform, which is also a small advertisement for our company. Thank you.

Chen Jianqiu: AI has been very popular in recent years, and the technology is advancing rapidly. The term I've heard most often this year is probably "intelligent agent." We have several business segments, one is robots, another is robots similar to those used for physical examinations, which is more of a health project, perhaps similar to Dreame. We are also working on intelligent agents for both traditional Chinese medicine and Western medicine. As you can see, our Hong Kong-based robot's TCM intelligent agent recently achieved the top score globally, but there are many hardships behind it. It's more about exploring user needs and delving deeper into the technology itself.

In the AI ​​era, creating an intelligent agent isn't difficult, but creating a good one is extremely challenging. This is especially true in a fiercely competitive environment. Therefore, we need to focus more on the user's perspective and help them solve real-world problems. There are many interesting aspects to this, which we can discuss later. For example, in health-related intelligent agents, there are some common issues. We all know what a healthy daily routine should look like, but due to social obligations or business commitments, we can't always live according to our ideal lifestyle. For instance, if I have to socialize in the evening, should I drink baijiu, wine, or spirits? There are many interesting points here. We've also monitored this using various devices; for example, drinking beer might lead to better rest, but drinking wine might leave you feeling weak the next day. These can all be supported by data to demonstrate how we can maintain better health even with less-than-ideal lifestyles. This is the change that AI and intelligent agents are bringing to our lives. Thank you.

Sun Zexuan: I think AI is the most important thing for AI hardware. In the future, I believe most hardware companies will be based on AI services, and our company will definitely be based on a subscription model for AI services. This is our fundamental business model. So we are now focusing all our efforts on AI.

Alven: From what I've heard, AI will spawn many new service paradigms in your existing hardware products. These paradigms will bring many new user subscriptions, whether it's the health agent you just mentioned, or other features in your products. My last question is a bit open-ended, still revolving around the three points of convergence: hardware, AI, and crypto. I'd like to quickly ask the three of you, where do you think the most likely convergence points will be?

Chen Jianqiu: I think the most likely application is for AI to help you make money. When AI acquires cryptocurrencies, it gains trust, credentials, and payment capabilities. In many situations, it can help you make money, make quick decisions, and earn some extra cash. I think this is a very direct application. Going deeper, for example, in future health scenarios, many things can combine AI and cryptocurrencies; there are actually many possibilities.

Sun Zexuan: For our business unit, I think the most important point is that AI serves as the next-generation entry point for human-computer interaction. The best application of AI is that it can solve a large number of scenarios, such as combining NFC payment with blockchain to form a payment scenario.

Ben: I 100% agree with payments. Payments are definitely one of the most difficult and sexy prospects, and also one of the most promising parts of the machine economy for us.

Alven: It seems there's still a lot of room for imagination in this direction. We were limited by time today and didn't get to discuss many topics, but I'm very grateful to our three guests for sharing their long-term experience in their respective companies and fields. I think AI combined with crypto and smart hardware has tremendous potential. Thank you all for your time and participation.